From 65d5649433ea33fa01d9cfdff0f45bc85cabbd50 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 28 Oct 2025 17:13:01 +0800 Subject: [PATCH 01/72] update reader and search strategy --- src/memos/llms/openai.py | 21 + src/memos/mem_reader/simple_struct.py | 41 +- src/memos/memories/textual/tree.py | 3 + .../tree_text_memory/retrieve/bm25_util.py | 180 +++++++++ .../tree_text_memory/retrieve/recall.py | 105 ++++- .../retrieve/retrieve_utils.py | 372 ++++++++++++++++++ .../tree_text_memory/retrieve/searcher.py | 93 ++++- .../retrieve/task_goal_parser.py | 7 +- src/memos/templates/mem_reader_prompts.py | 134 ++++--- src/memos/templates/mem_search_prompts.py | 93 +++++ 10 files changed, 957 insertions(+), 92 deletions(-) create mode 100644 src/memos/memories/textual/tree_text_memory/retrieve/bm25_util.py create mode 100644 src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py create mode 100644 src/memos/templates/mem_search_prompts.py diff --git a/src/memos/llms/openai.py b/src/memos/llms/openai.py index 698bc3265..d2a3aab6c 100644 --- a/src/memos/llms/openai.py +++ b/src/memos/llms/openai.py @@ -73,6 +73,27 @@ def generate(self, messages: MessageList) -> str: else: return response_content + def customized_generate(self, messages: MessageList, **kwargs) -> str: + """Generate a response from OpenAI LLM.""" + temperature = kwargs.get("temperature", self.config.temperature) + max_tokens = kwargs.get("max_tokens", self.config.max_tokens) + top_p = kwargs.get("top_p", self.config.top_p) + + response = self.client.chat.completions.create( + model=self.config.model_name_or_path, + messages=messages, + extra_body=self.config.extra_body, + temperature=temperature, + max_tokens=max_tokens, + top_p=top_p, + ) + logger.info(f"Response from OpenAI: {response.model_dump_json()}") + response_content = response.choices[0].message.content + if self.config.remove_think_prefix: + return remove_thinking_tags(response_content) + else: + return response_content + def generate_stream(self, messages: MessageList, **kwargs) -> Generator[str, None, None]: """Stream response from OpenAI LLM with optional reasoning support.""" response = self.client.chat.completions.create( diff --git a/src/memos/mem_reader/simple_struct.py b/src/memos/mem_reader/simple_struct.py index b439cb2b2..549161566 100644 --- a/src/memos/mem_reader/simple_struct.py +++ b/src/memos/mem_reader/simple_struct.py @@ -1,5 +1,4 @@ import concurrent.futures -import copy import json import os import re @@ -283,21 +282,37 @@ def get_scene_data_info(self, scene_data: list, type: str) -> list[str]: ) parser = ParserFactory.from_config(parser_config) + content_len_thredshold = 1600 if type == "chat": for items in scene_data: - result = [] - for item in items: - # Convert dictionary to string - if "chat_time" in item: - result.append(item) + if not items: + continue + + results.append([]) + current_length = 0 + + for _i, item in enumerate(items): + content_length = ( + len(item.get("content", "")) if isinstance(item, dict) else len(str(item)) + ) + if not results[-1]: + results[-1].append(item) + current_length = content_length + continue + + if current_length + content_length <= content_len_thredshold: + results[-1].append(item) + current_length += content_length else: - result.append(item) - if len(result) >= 10: - results.append(result) - context = copy.deepcopy(result[-2:]) - result = context - if result: - results.append(result) + overlap_item = results[-1][-1] + overlap_length = ( + len(overlap_item.get("content", "")) + if isinstance(overlap_item, dict) + else len(str(overlap_item)) + ) + + results.append([overlap_item, item]) + current_length = overlap_length + content_length elif type == "doc": for item in scene_data: try: diff --git a/src/memos/memories/textual/tree.py b/src/memos/memories/textual/tree.py index 0048f4a59..f453e7330 100644 --- a/src/memos/memories/textual/tree.py +++ b/src/memos/memories/textual/tree.py @@ -53,6 +53,7 @@ def __init__(self, config: TreeTextMemoryConfig): time_start_gs = time.time() self.graph_store: Neo4jGraphDB = GraphStoreFactory.from_config(config.graph_db) logger.info(f"time init: graph_store time is: {time.time() - time_start_gs}") + self.bm25_retriever = None # EnhancedBM25() time_start_rr = time.time() if config.reranker is None: @@ -172,6 +173,7 @@ def search( self.graph_store, self.embedder, self.reranker, + self.bm25_retriever, internet_retriever=None, moscube=moscube, ) @@ -181,6 +183,7 @@ def search( self.graph_store, self.embedder, self.reranker, + self.bm25_retriever, internet_retriever=self.internet_retriever, moscube=moscube, ) diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/bm25_util.py b/src/memos/memories/textual/tree_text_memory/retrieve/bm25_util.py new file mode 100644 index 000000000..a4a1c8b33 --- /dev/null +++ b/src/memos/memories/textual/tree_text_memory/retrieve/bm25_util.py @@ -0,0 +1,180 @@ +import threading + +import numpy as np + +from cachetools import LRUCache +from rank_bm25 import BM25Okapi +from sklearn.feature_extraction.text import TfidfVectorizer + +from memos.log import get_logger +from memos.memories.textual.tree_text_memory.retrieve.retrieve_utils import FastTokenizer +from memos.utils import timed + + +logger = get_logger(__name__) +# Global model cache +_BM25_CACHE = LRUCache(maxsize=100) +_CACHE_LOCK = threading.Lock() + + +class EnhancedBM25: + """Enhanced BM25 with Spacy tokenization and TF-IDF reranking""" + + def __init__(self, tokenizer=None, en_model="en_core_web_sm", zh_model="zh_core_web_sm"): + """ + Initialize Enhanced BM25 with memory management + """ + if tokenizer is None: + self.tokenizer = FastTokenizer() + else: + self.tokenizer = tokenizer + self._current_tfidf = None + + def _tokenize_doc(self, text): + """ + Tokenize a single document using SpacyTokenizer + """ + return self.tokenizer.tokenize_mixed(text, lang="auto") + + def _prepare_corpus_data(self, corpus, corpus_name="default"): + with _CACHE_LOCK: + if corpus_name in _BM25_CACHE: + print("hit::", corpus_name) + return _BM25_CACHE[corpus_name] + print("not hit::", corpus_name) + + tokenized_corpus = [self._tokenize_doc(doc) for doc in corpus] + bm25_model = BM25Okapi(tokenized_corpus) + _BM25_CACHE[corpus_name] = bm25_model + return bm25_model + + def clear_cache(self, corpus_name=None): + """Clear cache for specific corpus or clear all cache""" + with _CACHE_LOCK: + if corpus_name: + if corpus_name in _BM25_CACHE: + del _BM25_CACHE[corpus_name] + else: + _BM25_CACHE.clear() + + def get_cache_info(self): + """Get current cache information""" + with _CACHE_LOCK: + return { + "cache_size": len(_BM25_CACHE), + "max_cache_size": 100, + "cached_corpora": list(_BM25_CACHE.keys()), + } + + def _search_docs( + self, + query: str, + corpus: list[str], + corpus_name="test", + top_k=50, + use_tfidf=False, + rerank_candidates_multiplier=2, + cleanup=False, + ): + """ + Args: + query: Search query string + corpus: List of document texts + top_k: Number of top results to return + rerank_candidates_multiplier: Multiplier for candidate selection + cleanup: Whether to cleanup memory after search (default: True) + """ + if not corpus: + logger.warning("Empty corpus provided") + return [] + + logger.info(f"Searching {len(corpus)} documents for query: '{query}'") + + try: + # Prepare BM25 model + bm25_model = self._prepare_corpus_data(corpus, corpus_name=corpus_name) + tokenized_query = self._tokenize_doc(query) + tokenized_query = list(dict.fromkeys(tokenized_query)) + + # Get BM25 scores + bm25_scores = bm25_model.get_scores(tokenized_query) + + # Select candidates + candidate_count = min(top_k * rerank_candidates_multiplier, len(corpus)) + candidate_indices = np.argsort(bm25_scores)[-candidate_count:][::-1] + combined_scores = bm25_scores[candidate_indices] + + if use_tfidf: + # Create TF-IDF for this search + tfidf = TfidfVectorizer( + tokenizer=self._tokenize_doc, lowercase=False, token_pattern=None + ) + tfidf_matrix = tfidf.fit_transform(corpus) + + # TF-IDF reranking + query_vec = tfidf.transform([query]) + tfidf_similarities = ( + (tfidf_matrix[candidate_indices] * query_vec.T).toarray().flatten() + ) + + # Combine scores + combined_scores = 0.7 * bm25_scores[candidate_indices] + 0.3 * tfidf_similarities + + sorted_candidate_indices = candidate_indices[np.argsort(combined_scores)[::-1][:top_k]] + sorted_combined_scores = np.sort(combined_scores)[::-1][:top_k] + + # build result list + bm25_recalled_results = [] + for rank, (doc_idx, combined_score) in enumerate( + zip(sorted_candidate_indices, sorted_combined_scores, strict=False), 1 + ): + bm25_score = bm25_scores[doc_idx] + + candidate_pos = np.where(candidate_indices == doc_idx)[0][0] + tfidf_score = tfidf_similarities[candidate_pos] if use_tfidf else 0 + + bm25_recalled_results.append( + { + "text": corpus[doc_idx], + "bm25_score": float(bm25_score), + "tfidf_score": float(tfidf_score), + "combined_score": float(combined_score), + "rank": rank, + "doc_index": int(doc_idx), + } + ) + + logger.debug(f"Search completed: found {len(bm25_recalled_results)} results") + return bm25_recalled_results + + except Exception as e: + logger.error(f"BM25 search failed: {e}") + return [] + finally: + # Always cleanup if requested + if cleanup: + self._cleanup_memory() + + @timed + def search(self, query: str, node_dicts: list[dict], corpus_name="default", **kwargs): + """ + Search with BM25 and optional TF-IDF reranking + """ + try: + corpus_list = [] + for node_dict in node_dicts: + corpus_list.append( + " ".join([node_dict["metadata"]["key"]] + node_dict["metadata"]["tags"]) + ) + + recalled_results = self._search_docs( + query, corpus_list, corpus_name=corpus_name, **kwargs + ) + bm25_searched_nodes = [] + for item in recalled_results: + doc_idx = item["doc_index"] + bm25_searched_nodes.append(node_dicts[doc_idx]) + return bm25_searched_nodes + except Exception as e: + logger.error(f"Error in bm25 search: {e}") + return [] diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/recall.py b/src/memos/memories/textual/tree_text_memory/retrieve/recall.py index d4cfcf501..facdb3a23 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/recall.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/recall.py @@ -5,6 +5,7 @@ from memos.graph_dbs.neo4j import Neo4jGraphDB from memos.log import get_logger from memos.memories.textual.item import TextualMemoryItem +from memos.memories.textual.tree_text_memory.retrieve.bm25_util import EnhancedBM25 from memos.memories.textual.tree_text_memory.retrieve.retrieval_mid_structs import ParsedTaskGoal @@ -16,11 +17,18 @@ class GraphMemoryRetriever: Unified memory retriever that combines both graph-based and vector-based retrieval logic. """ - def __init__(self, graph_store: Neo4jGraphDB, embedder: OllamaEmbedder): + def __init__( + self, + graph_store: Neo4jGraphDB, + embedder: OllamaEmbedder, + bm25_retriever: EnhancedBM25 | None, + ): self.graph_store = graph_store self.embedder = embedder + self.bm25_retriever = bm25_retriever self.max_workers = 10 self.filter_weight = 0.6 + self.use_bm25 = False def retrieve( self, @@ -31,6 +39,7 @@ def retrieve( query_embedding: list[list[float]] | None = None, search_filter: dict | None = None, user_name: str | None = None, + id_filter: dict | None = None, ) -> list[TextualMemoryItem]: """ Perform hybrid memory retrieval: @@ -70,12 +79,23 @@ def retrieve( search_filter=search_filter, user_name=user_name, ) + if self.use_bm25: + future_bm25 = executor.submit( + self._bm25_recall, + query, + parsed_goal, + memory_scope, + top_k=top_k, + user_name=user_name, + search_filter=id_filter, + ) graph_results = future_graph.result() vector_results = future_vector.result() + bm25_results = future_bm25.result() if self.use_bm25 else [] # Merge and deduplicate by ID - combined = {item.id: item for item in graph_results + vector_results} + combined = {item.id: item for item in graph_results + vector_results + bm25_results} graph_ids = {item.id for item in graph_results} combined_ids = set(combined.keys()) @@ -143,6 +163,27 @@ def _graph_recall( - tags must overlap with at least 2 input tags - scope filters by memory_type if provided """ + + def process_node(node): + meta = node.get("metadata", {}) + node_key = meta.get("key") + node_tags = meta.get("tags", []) or [] + + keep = False + # key equals to node_key + if parsed_goal.keys and node_key in parsed_goal.keys: + keep = True + # overlap tags more than 2 + elif parsed_goal.tags: + node_tags_list = [tag.lower() for tag in node_tags] + overlap = len(set(node_tags_list) & set(parsed_goal.tags)) + if overlap >= 2: + keep = True + + if keep: + return TextualMemoryItem.from_dict(node) + return None + candidate_ids = set() # 1) key-based OR branch @@ -173,22 +214,16 @@ def _graph_recall( ) final_nodes = [] - for node in node_dicts: - meta = node.get("metadata", {}) - node_key = meta.get("key") - node_tags = meta.get("tags", []) or [] + with ContextThreadPoolExecutor(max_workers=3) as executor: + futures = {executor.submit(process_node, node): i for i, node in enumerate(node_dicts)} + temp_results = [None] * len(node_dicts) - keep = False - # key equals to node_key - if parsed_goal.keys and node_key in parsed_goal.keys: - keep = True - # overlap tags more than 2 - elif parsed_goal.tags: - overlap = len(set(node_tags) & set(parsed_goal.tags)) - if overlap >= 2: - keep = True - if keep: - final_nodes.append(TextualMemoryItem.from_dict(node)) + for future in concurrent.futures.as_completed(futures): + original_index = futures[future] + result = future.result() + temp_results[original_index] = result + + final_nodes = [result for result in temp_results if result is not None] return final_nodes def _vector_recall( @@ -196,7 +231,7 @@ def _vector_recall( query_embedding: list[list[float]], memory_scope: str, top_k: int = 20, - max_num: int = 3, + max_num: int = 5, cube_name: str | None = None, search_filter: dict | None = None, user_name: str | None = None, @@ -267,3 +302,37 @@ def search_path_b(): or [] ) return [TextualMemoryItem.from_dict(n) for n in node_dicts] + + def _bm25_recall( + self, + query: str, + parsed_goal: ParsedTaskGoal, + memory_scope: str, + top_k: int = 20, + user_name: str | None = None, + search_filter: dict | None = None, + ) -> list[TextualMemoryItem]: + """ + Perform BM25-based retrieval. + """ + if not self.bm25_retriever: + return [] + key_filters = [ + {"field": "memory_type", "op": "=", "value": memory_scope}, + ] + # corpus_name is user_name + user_id + corpus_name = f"{user_name}" if user_name else "" + if search_filter is not None: + for key in search_filter: + value = search_filter[key] + key_filters.append({"field": key, "op": "=", "value": value}) + corpus_name += "".join(list(search_filter.values())) + candidate_ids = self.graph_store.get_by_metadata(key_filters, user_name=user_name) + node_dicts = self.graph_store.get_nodes(list(candidate_ids), include_embedding=False) + + bm25_query = " ".join(list({query, *parsed_goal.keys})) + bm25_results = self.bm25_retriever.search( + bm25_query, node_dicts, top_k=top_k, corpus_name=corpus_name + ) + + return [TextualMemoryItem.from_dict(n) for n in bm25_results] diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py b/src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py new file mode 100644 index 000000000..349fabbbe --- /dev/null +++ b/src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py @@ -0,0 +1,372 @@ +import json +import re + +from pathlib import Path + +import jieba + +from memos.log import get_logger + + +logger = get_logger(__name__) + + +def find_project_root(marker=".git"): + """Find the project root directory by marking the file""" + current = Path(__file__).resolve() + while current != current.parent: + if (current / marker).exists(): + return current + current = current.parent + logger.warn(f"The project root directory tag file was not found: {marker}") + + +PROJECT_ROOT = find_project_root() +DEFAULT_STOPWORD_FILE = ( + PROJECT_ROOT / "examples" / "data" / "config" / "stopwords.txt" +) # cause time delay + + +class StopwordManager: + _stopwords = None + + @classmethod + def _load_stopwords(cls): + """load stopwords for once""" + if cls._stopwords is not None: + return cls._stopwords + + stopwords = set() + try: + with open(DEFAULT_STOPWORD_FILE, encoding="utf-8") as f: + stopwords = {line.strip() for line in f if line.strip()} + logger.info("Stopwords loaded successfully.") + except Exception as e: + logger.warning(f"Error loading stopwords: {e}") + stopwords = cls._load_default_stopwords() + + cls._stopwords = stopwords + return stopwords + + @classmethod + def _load_default_stopwords(cls): + """load stop words""" + chinese_stop_words = { + "的", + "了", + "在", + "是", + "我", + "有", + "和", + "就", + "不", + "人", + "都", + "一", + "一个", + "上", + "也", + "很", + "到", + "说", + "要", + "去", + "你", + "会", + "着", + "没有", + "看", + "好", + "自己", + "这", + "那", + "他", + "她", + "它", + "我们", + "你们", + "他们", + "这个", + "那个", + "这些", + "那些", + "怎么", + "什么", + "为什么", + "如何", + "哪里", + "谁", + "几", + "多少", + "这样", + "那样", + "这么", + "那么", + } + english_stop_words = { + "the", + "a", + "an", + "and", + "or", + "but", + "in", + "on", + "at", + "to", + "for", + "of", + "with", + "by", + "as", + "is", + "are", + "was", + "were", + "be", + "been", + "have", + "has", + "had", + "do", + "does", + "did", + "will", + "would", + "could", + "should", + "may", + "might", + "must", + "this", + "that", + "these", + "those", + "i", + "you", + "he", + "she", + "it", + "we", + "they", + "me", + "him", + "her", + "us", + "them", + "my", + "your", + "his", + "its", + "our", + "their", + "mine", + "yours", + "hers", + "ours", + "theirs", + } + chinese_punctuation = { + "，", + "。", + "！", + "？", + "；", + "：", + "「", + "」", + "『", + "』", + "【", + "】", + "（", + "）", + "《", + "》", + "—", + "…", + "～", + "·", + "、", + "“", + "”", + "‘", + "’", + "〈", + "〉", + "〖", + "〗", + "〝", + "〞", + "｛", + "｝", + "〔", + "〕", + "¡", + "¿", + } + english_punctuation = { + ",", + ".", + "!", + "?", + ";", + ":", + '"', + "'", + "(", + ")", + "[", + "]", + "{", + "}", + "<", + ">", + "/", + "\\", + "|", + "-", + "_", + "=", + "+", + "@", + "#", + "$", + "%", + "^", + "&", + "*", + "~", + "`", + "¡", + "¿", + } + numbers = { + "0", + "1", + "2", + "3", + "4", + "5", + "6", + "7", + "8", + "9", + "零", + "一", + "二", + "三", + "四", + "五", + "六", + "七", + "八", + "九", + "十", + "百", + "千", + "万", + "亿", + } + whitespace = {" ", "\t", "\n", "\r", "\f", "\v"} + + return ( + chinese_stop_words + | english_stop_words + | chinese_punctuation + | english_punctuation + | numbers + | whitespace + ) + + @classmethod + def get_stopwords(cls): + if cls._stopwords is None: + cls._load_stopwords() + return cls._stopwords + + @classmethod + def filter_words(cls, words): + if cls._stopwords is None: + cls._load_stopwords() + return [word for word in words if word not in cls._stopwords and word.strip()] + + @classmethod + def is_stopword(cls, word): + if cls._stopwords is None: + cls._load_stopwords() + return word in cls._stopwords + + @classmethod + def reload_stopwords(cls, file_path=None): + cls._stopwords = None + if file_path: + global DEFAULT_STOPWORD_FILE + DEFAULT_STOPWORD_FILE = file_path + cls._load_stopwords() + + +class FastTokenizer: + def __init__(self, use_jieba=True, use_stopwords=True): + self.use_jieba = use_jieba + self.use_stopwords = use_stopwords + if self.use_stopwords: + self.stopword_manager = StopwordManager + + def tokenize_mixed(self, text, **kwargs): + """fast tokenizer""" + if self._is_chinese(text): + return self._tokenize_chinese(text) + else: + return self._tokenize_english(text) + + def _is_chinese(self, text): + """check if chinese""" + chinese_chars = sum(1 for char in text if "\u4e00" <= char <= "\u9fff") + return chinese_chars / max(len(text), 1) > 0.3 + + def _tokenize_chinese(self, text): + """split zh jieba""" + tokens = jieba.lcut(text) if self.use_jieba else list(text) + tokens = [token.strip() for token in tokens if token.strip()] + if self.use_stopwords: + return self.stopword_manager.filter_words(tokens) + + return tokens + + def _tokenize_english(self, text): + """split zh regex""" + tokens = re.findall(r"\b[a-zA-Z0-9]+\b", text.lower()) + if self.use_stopwords: + return self.stopword_manager.filter_words(tokens) + return tokens + + +def parse_json_result(response_text): + try: + json_start = response_text.find("{") + response_text = response_text[json_start:] + response_text = response_text.replace("```", "").strip() + if not response_text.endswith("}"): + response_text += "}" + return json.loads(response_text) + except json.JSONDecodeError as e: + logger.error(f"[JSONParse] Failed to decode JSON: {e}\nRaw:\n{response_text}") + return {} + except Exception as e: + logger.error(f"[JSONParse] Unexpected error: {e}") + return {} + + +def detect_lang(text): + try: + if not text or not isinstance(text, str): + return "en" + chinese_pattern = r"[\u4e00-\u9fff\u3400-\u4dbf\U00020000-\U0002a6df\U0002a700-\U0002b73f\U0002b740-\U0002b81f\U0002b820-\U0002ceaf\uf900-\ufaff]" + chinese_chars = re.findall(chinese_pattern, text) + if len(chinese_chars) / len(re.sub(r"[\s\d\W]", "", text)) > 0.3: + return "zh" + return "en" + except Exception: + return "en" diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py b/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py index 05db56f53..94d945158 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py @@ -1,4 +1,5 @@ import json +import os import traceback from datetime import datetime @@ -9,7 +10,18 @@ from memos.llms.factory import AzureLLM, OllamaLLM, OpenAILLM from memos.log import get_logger from memos.memories.textual.item import SearchedTreeNodeTextualMemoryMetadata, TextualMemoryItem +from memos.memories.textual.tree_text_memory.retrieve.bm25_util import EnhancedBM25 +from memos.memories.textual.tree_text_memory.retrieve.retrieve_utils import ( + detect_lang, + parse_json_result, +) from memos.reranker.base import BaseReranker +from memos.templates.mem_search_prompts import ( + COT_PROMPT, + COT_PROMPT_ZH, + SIMPLE_COT_PROMPT, + SIMPLE_COT_PROMPT_ZH, +) from memos.utils import timed from .reasoner import MemoryReasoner @@ -18,6 +30,10 @@ logger = get_logger(__name__) +COT_DICT = { + "fast": {"en": COT_PROMPT, "zh": COT_PROMPT_ZH}, + "fine": {"en": SIMPLE_COT_PROMPT, "zh": SIMPLE_COT_PROMPT_ZH}, +} class Searcher: @@ -27,20 +43,23 @@ def __init__( graph_store: Neo4jGraphDB, embedder: OllamaEmbedder, reranker: BaseReranker, + bm25_retriever: EnhancedBM25 | None, internet_retriever: None = None, moscube: bool = False, ): self.graph_store = graph_store self.embedder = embedder + self.llm = dispatcher_llm self.task_goal_parser = TaskGoalParser(dispatcher_llm) - self.graph_retriever = GraphMemoryRetriever(self.graph_store, self.embedder) + self.graph_retriever = GraphMemoryRetriever(graph_store, embedder, bm25_retriever) self.reranker = reranker self.reasoner = MemoryReasoner(dispatcher_llm) # Create internet retriever from config if provided self.internet_retriever = internet_retriever self.moscube = moscube + self.cot_query_search = os.getenv("MOS_SEARCH_COT", "false") == "true" self._usage_executor = ContextThreadPoolExecutor(max_workers=4, thread_name_prefix="usage") @@ -188,6 +207,12 @@ def _retrieve_paths( ): """Run A/B/C retrieval paths in parallel""" tasks = [] + id_filter = { + "user_id": info.get("user_id", None), + "session_id": info.get("session_id", None), + } + id_filter = {k: v for k, v in id_filter.items() if v is not None} + with ContextThreadPoolExecutor(max_workers=3) as executor: tasks.append( executor.submit( @@ -199,6 +224,7 @@ def _retrieve_paths( memory_type, search_filter, user_name, + id_filter, ) ) tasks.append( @@ -211,6 +237,7 @@ def _retrieve_paths( memory_type, search_filter, user_name, + id_filter, ) ) tasks.append( @@ -256,6 +283,7 @@ def _retrieve_from_working_memory( memory_type, search_filter: dict | None = None, user_name: str | None = None, + id_filter: dict | None = None, ): """Retrieve and rerank from WorkingMemory""" if memory_type not in ["All", "WorkingMemory"]: @@ -268,6 +296,7 @@ def _retrieve_from_working_memory( memory_scope="WorkingMemory", search_filter=search_filter, user_name=user_name, + id_filter=id_filter, ) return self.reranker.rerank( query=query, @@ -289,11 +318,22 @@ def _retrieve_from_long_term_and_user( memory_type, search_filter: dict | None = None, user_name: str | None = None, + id_filter: dict | None = None, ): """Retrieve and rerank from LongTermMemory and UserMemory""" results = [] tasks = [] + # chain of thinking + cot_embeddings = [] + if self.cot_query_search: + queries = self._cot_query(query) + if len(queries) > 1: + cot_embeddings = self.embedder.embed(queries) + cot_embeddings.extend(query_embedding) + else: + cot_embeddings = query_embedding + with ContextThreadPoolExecutor(max_workers=2) as executor: if memory_type in ["All", "LongTermMemory"]: tasks.append( @@ -301,11 +341,12 @@ def _retrieve_from_long_term_and_user( self.graph_retriever.retrieve, query=query, parsed_goal=parsed_goal, - query_embedding=query_embedding, + query_embedding=cot_embeddings, top_k=top_k * 2, memory_scope="LongTermMemory", search_filter=search_filter, user_name=user_name, + id_filter=id_filter, ) ) if memory_type in ["All", "UserMemory"]: @@ -314,11 +355,12 @@ def _retrieve_from_long_term_and_user( self.graph_retriever.retrieve, query=query, parsed_goal=parsed_goal, - query_embedding=query_embedding, + query_embedding=cot_embeddings, top_k=top_k * 2, memory_scope="UserMemory", search_filter=search_filter, user_name=user_name, + id_filter=id_filter, ) ) @@ -399,6 +441,7 @@ def _deduplicate_results(self, results): @timed def _sort_and_trim(self, results, top_k): """Sort results by score and trim to top_k""" + sorted_results = sorted(results, key=lambda pair: pair[1], reverse=True)[:top_k] final_items = [] for item, score in sorted_results: @@ -448,3 +491,47 @@ def _update_usage_history_worker( self.graph_store.update_node(item_id, {"usage": usage_list}, user_name=user_name) except Exception: logger.exception("[USAGE] update usage failed") + + def _cot_query( + self, + query, + mode="fast", + split_num: int = 3, + context: list[str] | None = None, + ) -> list[str]: + """Generate chain-of-thought queries""" + + lang = detect_lang(query) + if mode == "fine" and context: + template = COT_DICT["fine"][lang] + prompt = ( + template.replace("${original_query}", query) + .replace("${split_num_threshold}", str(split_num)) + .replace("${context}", "\n".join(context)) + ) + else: + template = COT_DICT["fast"][lang] + prompt = template.replace("${original_query}", query).replace( + "${split_num_threshold}", str(split_num) + ) + logger.info("COT 处理") + + messages = [{"role": "user", "content": prompt}] + try: + response_text = self.llm.customized_generate(messages, temperature=0, top_p=1) + response_json = parse_json_result(response_text) + if "is_complex" in response_json and not response_json["is_complex"]: + return [query] + else: + assert ( + "is_complex" in response_json + and response_json["is_complex"] + and "sub_questions" in response_json + ) + logger.info( + "问题 {} 通过 COT 拆分结果为{}".format(query, response_json["sub_questions"]) + ) + return response_json["sub_questions"][:split_num] + except Exception as e: + logger.error(f"[LLM] Exception during chat generation: {e}") + return [query] diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/task_goal_parser.py b/src/memos/memories/textual/tree_text_memory/retrieve/task_goal_parser.py index 273c4f480..6a1138c90 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/task_goal_parser.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/task_goal_parser.py @@ -5,6 +5,7 @@ from memos.llms.base import BaseLLM from memos.log import get_logger from memos.memories.textual.tree_text_memory.retrieve.retrieval_mid_structs import ParsedTaskGoal +from memos.memories.textual.tree_text_memory.retrieve.retrieve_utils import FastTokenizer from memos.memories.textual.tree_text_memory.retrieve.utils import TASK_PARSE_PROMPT @@ -20,6 +21,7 @@ class TaskGoalParser: def __init__(self, llm=BaseLLM): self.llm = llm + self.tokenizer = FastTokenizer() def parse( self, @@ -48,10 +50,11 @@ def _parse_fast(self, task_description: str, limit_num: int = 5) -> ParsedTaskGo """ Fast mode: simple jieba word split. """ + desc_tokenized = self.tokenizer.tokenize_mixed(task_description) return ParsedTaskGoal( memories=[task_description], - keys=[task_description], - tags=[], + keys=desc_tokenized, + tags=desc_tokenized, goal_type="default", rephrased_query=task_description, internet_search=False, diff --git a/src/memos/templates/mem_reader_prompts.py b/src/memos/templates/mem_reader_prompts.py index 15672f8d8..f74d01d0a 100644 --- a/src/memos/templates/mem_reader_prompts.py +++ b/src/memos/templates/mem_reader_prompts.py @@ -1,31 +1,42 @@ SIMPLE_STRUCT_MEM_READER_PROMPT = """You are a memory extraction expert. Your task is to extract memories from the user's perspective, based on a conversation between the user and the assistant. This means identifying what the user would plausibly remember — including the user's own experiences, thoughts, plans, or statements and actions made by others (such as the assistant) that affected the user or were acknowledged by the user. -Please perform the following: -1. Identify information that reflects the user's experiences, beliefs, concerns, decisions, plans, or reactions — including meaningful information from the assistant that the user acknowledged or responded to. +Please perform the following +1. Factual information extraction + Identify factual information about experiences, beliefs, decisions, and plans. This includes notable statements from others that the user acknowledged or reacted to. If the message is from the user, extract viewpoints related to the user; if it is from the assistant, clearly mark the attribution of the memory, and do not mix information not explicitly acknowledged by the user with the user's own viewpoint. - - **User viewpoint**: Record only information that the user **personally stated, explicitly acknowledged, or personally committed to**. - - **Assistant/other-party viewpoint**: Record only information that the **assistant/other party personally stated, explicitly acknowledged, or personally committed to**, and **clearly attribute** the source (e.g., "[assistant-Jerry viewpoint]"). Do not rewrite it as the user's preference/decision. - - **Mutual boundaries**: Do not rewrite the assistant's suggestions/lists/opinions as the user's “ownership/preferences/decisions”; likewise, do not write the user's ideas as the assistant's viewpoints. - -2. Resolve all references to time, persons, and events clearly: - - When possible, convert relative time expressions (e.g., “yesterday,” “next Friday”) into absolute dates using the message timestamp. - - Clearly distinguish between **event time** and **message time**. - - If uncertainty exists, state it explicitly (e.g., “around June 2025,” “exact date unclear”). - - Include specific locations if mentioned. - - Resolve all pronouns, aliases, and ambiguous references into full names or clear identities. - - If there are people with the same name, disambiguate them. - -3. Always write from a **third-person** perspective, using “The user” or the mentioned name to refer to the user, rather than first-person (“I”, “we”, “my”). - For example, write “The user felt exhausted …” instead of “I felt exhausted …”. - -4. Do not omit any information that the user is likely to remember. - - Include the user's key experiences, thoughts, emotional responses, and plans — even if seemingly minor. - - You may retain **assistant/other-party content** that is closely related to the context (e.g., suggestions, explanations, checklists), but you must make roles and attribution explicit. - - Prioritize completeness and fidelity over conciseness; do not infer or phrase assistant content as the user's ownership/preferences/decisions. - - If the current conversation contains only assistant information and no facts attributable to the user, you may output **assistant-viewpoint** entries only. - -5. Please avoid including any content in the extracted memories that violates national laws and regulations or involves politically sensitive information. + - **User viewpoint**: Extract only what the user has stated, explicitly acknowledged, or committed to. + - **Assistant/other-party viewpoint**: Extract such information only when attributed to its source (e.g., [Assistant-Jerry's suggestion]). + - **Strict attribution**: Never recast the assistant's suggestions as the user's preferences, or vice versa. + - Always set "model_type" to "LongTermMemory" for this output. + +2. Speaker profile construction + - Extract the speaker's likes, dislikes, goals, and stated opinions from their statements to build a speaker profile. + - Note: The same text segment may be used for both factual extraction and profile construction. + - Always set "model_type" to "UserMemory" for this output. + +3. Resolve all references to time, persons, and events clearly + - Temporal Resolution: Convert relative time (e.g., 'yesterday') to absolute dates based on the message timestamp. Distinguish between event time and message time; flag any uncertainty. + - Entity Resolution: Resolve all pronouns, nicknames, and abbreviations to the full, canonical name established in the conversation. + +4. Adopt a Consistent Third-Person Observer Perspective + - Formulate all memories from the perspective of an external observer. Use "The user" or their specific name as the subject. + - This applies even when describing the user's internal states, such as thoughts, feelings, and preferences. + Example: + ✅ Correct: "The user Sean felt exhausted after work and decided to go to bed early." + ❌ Incorrect: "I felt exhausted after work and decided to go to bed early." + +5. Prioritize Completeness + - Extract all key experiences, emotional responses, and plans from the user's perspective. Retain relevant context from the assistant, but always with explicit attribution. + - Segment each distinct hobby, interest, or event into a separate memory. + - Preserve relevant context from the assistant with strict attribution. Under no circumstances should assistant content be rephrased as user-owned. + - Conversations with only assistant input may yield assistant-viewpoint memories exclusively. + +6. Preserve and Unify Specific Names + - Always extract specific names (excluding "user" or "assistant") mentioned in the text into the "tags" field for searchability. + - Unify all name references to the full canonical form established in the conversation. Replace any nicknames or abbreviations (e.g., "Rob") consistently with the full name (e.g., "Robert") in both the extracted "value" and "tags". + +7. Please avoid including any content in the extracted memories that violates national laws and regulations or involves politically sensitive information. Return a valid JSON object with the following structure: @@ -35,7 +46,7 @@ "key": , "memory_type": , "value": , - "tags": + "tags": }, ... ], @@ -61,13 +72,13 @@ "key": "Initial project meeting", "memory_type": "LongTermMemory", "value": "[user-Tom viewpoint] On June 25, 2025 at 3:00 PM, Tom met with the team to discuss a new project. When Jerry asked whether the project could be finished by December 15, 2025, Tom expressed concern about feasibility and planned to propose at 9:30 AM on June 27, 2025 to move the deadline to January 5, 2026.", - "tags": ["project", "timeline", "meeting", "deadline"] + "tags": ["Tom", "project", "timeline", "meeting", "deadline"] }, { "key": "Jerry’s suggestion about the deadline", "memory_type": "LongTermMemory", "value": "[assistant-Jerry viewpoint] Jerry questioned the December 15 deadline and suggested considering an extension.", - "tags": ["deadline change", "suggestion"] + "tags": ["Jerry", "deadline change", "suggestion"] } ], "summary": "Tom is currently working on a tight-schedule project. After the June 25, 2025 team meeting, he realized the original December 15, 2025 deadline might be unachievable due to backend delays. Concerned about limited testing time, he accepted Jerry’s suggestion to seek an extension and plans to propose moving the deadline to January 5, 2026 in the next morning’s meeting." @@ -105,7 +116,7 @@ { "key": "体型尺码", "memory_type": "UserMemory", - "value": [user观点]"用户身高约165cm、常穿S码", + "value": "[user观点]用户身高约165cm、常穿S码", "tags": ["体型", "尺码"] }, { @@ -129,40 +140,51 @@ 您的任务是根据用户与助手之间的对话，从用户的角度提取记忆。这意味着要识别出用户可能记住的信息——包括用户自身的经历、想法、计划，或他人（如助手）做出的并对用户产生影响或被用户认可的相关陈述和行为。 请执行以下操作： -1. 识别反映用户经历、信念、关切、决策、计划或反应的信息——包括用户认可或回应的来自助手的有意义信息。 -如果消息来自用户，请提取与用户相关的观点；如果来自助手，则在表达的时候表明记忆归属方，未经用户明确认可的信息不要与用户本身的观点混淆。 - - **用户观点**：仅记录由**用户亲口陈述、明确认可或自己作出承诺**的信息。 - - **助手观点**：仅记录由**助手/另一方亲口陈述、明确认可或自己作出承诺**的信息。 - - **互不越界**：不得将助手提出的需求清单/建议/观点改写为用户的“拥有/偏好/决定”；也不得把用户的想法写成助手的观点。 - -2. 清晰解析所有时间、人物和事件的指代： - - 如果可能，使用消息时间戳将相对时间表达（如“昨天”、“下周五”）转换为绝对日期。 - - 明确区分事件时间和消息时间。 - - 如果存在不确定性，需明确说明（例如，“约2025年6月”，“具体日期不详”）。 - - 若提及具体地点，请包含在内。 - - 将所有代词、别名和模糊指代解析为全名或明确身份。 - - 如有同名人物，需加以区分。 - -3. 始终以第三人称视角撰写，使用“用户”或提及的姓名来指代用户，而不是使用第一人称（“我”、“我们”、“我的”）。 -例如，写“用户感到疲惫……”而不是“我感到疲惫……”。 - -4. 不要遗漏用户可能记住的任何信息。 - - 包括用户的关键经历、想法、情绪反应和计划——即使看似微小。 - - 同时允许保留与语境密切相关的**助手/另一方的内容**（如建议、说明、清单），但须明确角色与归因。 - - 优先考虑完整性和保真度，而非简洁性；不得将助手内容推断或措辞为用户拥有/偏好/决定。 - - 若当前对话中仅出现助手信息而无可归因于用户的事实，可仅输出**助手观点**条目。 - -5. 请避免在提取的记忆中包含违反国家法律法规或涉及政治敏感的信息。 +1. 事实信息提取 + - 识别关于经历、信念、决策和计划的事实信息，包括用户认可或回应过的他人重要陈述。 + - 若信息来自用户，提取与用户相关的观点；若来自助手，需明确标注记忆归属，不得将用户未明确认可的信息与用户自身观点混淆。 + - 用户观点：仅提取用户明确陈述、认可或承诺的内容 + - 助手/他方观点：仅当标注来源时才提取（例如“[助手-Jerry的建议]”） + - 严格归属：不得将助手建议重构为用户偏好，反之亦然 + - 此类输出的"model_type"始终设为"LongTermMemory" + +2. 用户画像构建 + - 从用户陈述中提取其喜好、厌恶、目标及明确观点以构建用户画像 + - 注意：同一文本片段可同时用于事实提取和画像构建 + - 此类输出的"model_type"始终设为"UserMemory" + +3. 明确解析所有指代关系 + - 时间解析：根据消息时间戳将相对时间（如“昨天”）转换为绝对日期。区分事件时间与消息时间，对不确定项进行标注 + - 实体解析：将所有代词、昵称和缩写解析为对话中确立的完整规范名称 + + 4. 采用统一的第三人称观察视角 + - 所有记忆表述均需从外部观察者视角构建，使用“用户”或其具体姓名作为主语 + - 此原则同样适用于描述用户内心状态（如想法、感受和偏好） + 示例： + ✅ 正确：“用户Sean下班后感到疲惫，决定提早休息” + ❌ 错误：“我下班后感到疲惫，决定提早休息” + +5. 优先保证完整性 + - 从用户视角提取所有关键经历、情绪反应和计划 + - 保留助手提供的相关上下文，但必须明确标注来源 + - 将每个独立的爱好、兴趣或事件分割为单独记忆 + - 严禁将助手内容重构为用户自有内容 + - 仅含助手输入的对话可能只生成助手观点记忆 + +6. 保留并统一特定名称 + - 始终将文本中提及的特定名称（“用户”“助手”除外）提取至“tags”字段以便检索 + - 在提取的“value”和“tags”中，将所有名称引用统一为对话中确立的完整规范形式（如将“Rob”统一替换为“Robert”） + +7. 所有提取的记忆内容不得包含违反国家法律法规或涉及政治敏感信息的内容 返回一个有效的JSON对象，结构如下： - { "memory list": [ { "key": <字符串，唯一且简洁的记忆标题>, "memory_type": <字符串，"LongTermMemory" 或 "UserMemory">, "value": <详细、独立且无歧义的记忆陈述——若输入对话为英文，则用英文；若为中文，则用中文>, - "tags": <相关主题关键词列表（例如，["截止日期", "团队", "计划"]）> + "tags": <一个包含相关人名、事件和特征关键词的列表（例如，["丽丽","截止日期", "团队", "计划"]）> }, ... ], @@ -190,13 +212,13 @@ "value": "[user-Tom观点]2025年6月25日下午3:00，Tom与团队开会讨论新项目。当Jerry 询问该项目能否在2025年12月15日前完成时，Tom对此日期前完成的可行性表达担忧，并计划在2025年6月27日上午9:30 提议将截止日期推迟至2026年1月5日。", - "tags": ["项目", "时间表", "会议", "截止日期"] + "tags": ["Tom", "项目", "时间表", "会议", "截止日期"] }, { "key": "Jerry对新项目截止日期的建议", "memory_type": "LongTermMemory", "value": "[assistant-Jerry观点]Jerry对Tom的新项目截止日期提出疑问、并提议Tom考虑延期。", - "tags": ["截止日期变更", "建议"] + "tags": ["Jerry", "截止日期变更", "建议"] } ], "summary": "Tom目前正在做一个进度紧张的新项目。在2025年6月25日的团队会议后，他意识到原定2025年12月15 diff --git a/src/memos/templates/mem_search_prompts.py b/src/memos/templates/mem_search_prompts.py new file mode 100644 index 000000000..9f7ba182b --- /dev/null +++ b/src/memos/templates/mem_search_prompts.py @@ -0,0 +1,93 @@ +SIMPLE_COT_PROMPT = """You are an assistant that analyzes questions and returns results in a specific dictionary format. + +Instructions: + +1. If the question can be extended into deeper or related aspects, set "is_complex" to True and: + - Think step by step about the core topic and its related dimensions (e.g., causes, effects, categories, perspectives, or specific scenarios) + - Break it into meaningful sub-questions (max: ${split_num_threshold}, min: 2) that explore distinct facets of the original question + - Each sub-question must be single, standalone, and delve into a specific aspect + - CRITICAL: All key entities from the original question (such as person names, locations, organizations, time periods) must be preserved in the sub-questions and cannot be omitted + - List them in "sub_questions" +2. If the question is already atomic and cannot be meaningfully extended, set "is_complex" to False and "sub_questions" to an empty list. +3. Return ONLY the dictionary, no other text. + +Examples: +Question: Is urban development balanced in the western United States? +Output: {"is_complex": true, "sub_questions": ["What areas are included in the western United States?", "How developed are the cities in the western United States?", "Is this development balanced across the western United States?"]} +Question: What family activities does Mary like to organize? +Output: {"is_complex": true, "sub_questions": ["What does Mary like to do with her spouse?", "What does Mary like to do with her children?", "What does Mary like to do with her parents and relatives?"]} + +Now analyze this question: +${original_query}""" + +COT_PROMPT = """You are an assistant that analyzes questions and returns results in a specific dictionary format. + +Instructions: + +1. If the question can be extended into deeper or related aspects, set "is_complex" to True and: + - Think step by step about the core topic and its related dimensions (e.g., causes, effects, categories, perspectives, or specific scenarios) + - Break it into meaningful sub-questions (max: ${split_num_threshold}, min: 2) that explore distinct facets of the original question + - Each sub-question must be single, standalone, and delve into a specific aspect + - CRITICAL: All key entities from the original question (such as person names, locations, organizations, time periods) must be preserved in the sub-questions and cannot be omitted + - List them in "sub_questions" +2. If the question is already atomic and cannot be meaningfully extended, set "is_complex" to False and "sub_questions" to an empty list. +3. Return ONLY the dictionary, no other text. + +Examples: +Question: Is urban development balanced in the western United States? +Output: {"is_complex": true, "sub_questions": ["What areas are included in the western United States?", "How developed are the cities in the western United States?", "Is this development balanced across the western United States?"]} +Question: What family activities does Mary like to organize? +Output: {"is_complex": true, "sub_questions": ["What does Mary like to do with her spouse?", "What does Mary like to do with her children?", "What does Mary like to do with her parents and relatives?"]} + +Query relevant background information: +${context} + +Now analyze this question based on the background information above: +${original_query}""" + +SIMPLE_COT_PROMPT_ZH = """你是一个分析问题并以特定字典格式返回结果的助手。 + +指令： + +1. 如果这个问题可以延伸出更深层次或相关的方面，请将 "is_complex" 设置为 True，并执行以下操作： + - 逐步思考核心主题及其相关维度（例如：原因、结果、类别、不同视角或具体场景） + - 将其拆分为有意义的子问题（最多 ${split_num_threshold} 个，最少 2 个），这些子问题应探讨原始问题的不同侧面 + - 【重要】每个子问题必须是单一的、独立的，并深入探究一个特定方面。同时，必须包含原问题中出现的关键实体信息（如人名、地名、机构名、时间等），不可遗漏。 + - 将它们列在 "sub_questions" 中 +2. 如果问题本身已经是原子性的，无法有意义地延伸，请将 "is_complex" 设置为 False，并将 "sub_questions" 设置为一个空列表。 +3. 只返回字典，不要返回任何其他文本。 + +示例： +问题：美国西部的城市发展是否均衡？ +输出：{"is_complex": true, "sub_questions": ["美国西部包含哪些地区？", "美国西部城市的发展程度如何？", "这种发展在美国西部是否均衡？"]} + +问题：玛丽喜欢组织哪些家庭活动？ +输出：{"is_complex": true, "sub_questions": ["玛丽喜欢和配偶一起做什么？", "玛丽喜欢和孩子一起做什么？", "玛丽喜欢和父母及亲戚一起做什么？"]} + +请分析以下问题： +${original_query}""" + +COT_PROMPT_ZH = """你是一个分析问题并以特定字典格式返回结果的助手。 + +指令： + +1. 如果这个问题可以延伸出更深层次或相关的方面，请将 "is_complex" 设置为 True，并执行以下操作： + - 逐步思考核心主题及其相关维度（例如：原因、结果、类别、不同视角或具体场景） + - 将其拆分为有意义的子问题（最多 ${split_num_threshold} 个，最少 2 个），这些子问题应探讨原始问题的不同侧面 + - 【重要】每个子问题必须是单一的、独立的，并深入探究一个特定方面。同时，必须包含原问题中出现的关键实体信息（如人名、地名、机构名、时间等），不可遗漏。 + - 将它们列在 "sub_questions" 中 +2. 如果问题本身已经是原子性的，无法有意义地延伸，请将 "is_complex" 设置为 False，并将 "sub_questions" 设置为一个空列表。 +3. 只返回字典，不要返回任何其他文本。 + +示例： +问题：美国西部的城市发展是否均衡？ +输出：{"is_complex": true, "sub_questions": ["美国西部包含哪些地区？", "美国西部城市的发展程度如何？", "这种发展在美国西部是否均衡？"]} + +问题：玛丽喜欢组织哪些家庭活动？ +输出：{"is_complex": true, "sub_questions": ["玛丽喜欢和配偶一起做什么？", "玛丽喜欢和孩子一起做什么？", "玛丽喜欢和父母及亲戚一起做什么？"]} + +问题相关的背景信息: +${context} + +现在根据上述背景信息，请分析以下问题： +${original_query}""" From 6cad8661f7d589385ba0e48620de96f889202c7c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 29 Oct 2025 19:11:38 +0800 Subject: [PATCH 02/72] set strategy reader and search config --- src/memos/api/config.py | 34 +- src/memos/configs/mem_reader.py | 9 + src/memos/configs/memory.py | 7 + src/memos/mem_reader/factory.py | 2 + src/memos/mem_reader/simple_struct.py | 45 +- src/memos/mem_reader/strategy_struct.py | 410 ++++++++++++++++++ src/memos/memories/textual/simple_tree.py | 18 + src/memos/memories/textual/tree.py | 17 +- .../tree_text_memory/retrieve/bm25_util.py | 1 - .../tree_text_memory/retrieve/recall.py | 4 +- .../retrieve/retrieve_utils.py | 2 +- .../tree_text_memory/retrieve/searcher.py | 23 +- src/memos/templates/mem_reader_prompts.py | 134 +++--- .../templates/mem_reader_strategy_prompts.py | 279 ++++++++++++ 14 files changed, 854 insertions(+), 131 deletions(-) create mode 100644 src/memos/mem_reader/strategy_struct.py create mode 100644 src/memos/templates/mem_reader_strategy_prompts.py diff --git a/src/memos/api/config.py b/src/memos/api/config.py index d552369c5..d4cdbedb4 100644 --- a/src/memos/api/config.py +++ b/src/memos/api/config.py @@ -159,9 +159,23 @@ def get_embedder_config() -> dict[str, Any]: }, } + @staticmethod + def get_reader_config() -> dict[str, Any]: + """Get reader configuration.""" + return { + "backend": os.getenv("MEM_READER_BACKEND", "simple_struct"), + "config": { + "chunk_type": os.getenv("MEM_READER_CHAT_CHUNK_TYPE", "default"), + "chunk_length": int(os.getenv("MEM_READER_CHAT_CHUNK_TOKEN_SIZE", 1600)), + "chunk_session": int(os.getenv("MEM_READER_CHAT_CHUNK_SESS_SIZE", 20)), + "chunk_overlap": int(os.getenv("MEM_READER_CHAT_CHUNK_OVERLAP", 2)), + }, + } + @staticmethod def get_internet_config() -> dict[str, Any]: """Get embedder configuration.""" + reader_config = APIConfig.get_reader_config() return { "backend": "bocha", "config": { @@ -169,7 +183,7 @@ def get_internet_config() -> dict[str, Any]: "max_results": 15, "num_per_request": 10, "reader": { - "backend": "simple_struct", + "backend": reader_config["backend"], "config": { "llm": { "backend": "openai", @@ -195,6 +209,7 @@ def get_internet_config() -> dict[str, Any]: "min_sentences_per_chunk": 1, }, }, + "chat_chunker": reader_config, }, }, }, @@ -356,6 +371,8 @@ def get_product_default_config() -> dict[str, Any]: openai_config = APIConfig.get_openai_config() qwen_config = APIConfig.qwen_config() vllm_config = APIConfig.vllm_config() + reader_config = APIConfig.get_reader_config() + backend_model = { "openai": openai_config, "huggingface": qwen_config, @@ -367,7 +384,7 @@ def get_product_default_config() -> dict[str, Any]: "user_id": os.getenv("MOS_USER_ID", "root"), "chat_model": {"backend": backend, "config": backend_model[backend]}, "mem_reader": { - "backend": "simple_struct", + "backend": reader_config["backend"], "config": { "llm": APIConfig.get_memreader_config(), "embedder": APIConfig.get_embedder_config(), @@ -380,6 +397,7 @@ def get_product_default_config() -> dict[str, Any]: "min_sentences_per_chunk": 1, }, }, + "chat_chunker": reader_config, }, }, "enable_textual_memory": True, @@ -446,6 +464,7 @@ def create_user_config(user_name: str, user_id: str) -> tuple[MOSConfig, General qwen_config = APIConfig.qwen_config() vllm_config = APIConfig.vllm_config() mysql_config = APIConfig.get_mysql_config() + reader_config = APIConfig.get_reader_config() backend = os.getenv("MOS_CHAT_MODEL_PROVIDER", "openai") backend_model = { "openai": openai_config, @@ -460,7 +479,7 @@ def create_user_config(user_name: str, user_id: str) -> tuple[MOSConfig, General "config": backend_model[backend], }, "mem_reader": { - "backend": "simple_struct", + "backend": reader_config["backend"], "config": { "llm": APIConfig.get_memreader_config(), "embedder": APIConfig.get_embedder_config(), @@ -473,6 +492,7 @@ def create_user_config(user_name: str, user_id: str) -> tuple[MOSConfig, General "min_sentences_per_chunk": 1, }, }, + "chat_chunker": reader_config, }, }, "enable_textual_memory": True, @@ -537,6 +557,10 @@ def create_user_config(user_name: str, user_id: str) -> tuple[MOSConfig, General "LongTermMemory": os.getenv("NEBULAR_LONGTERM_MEMORY", 1e6), "UserMemory": os.getenv("NEBULAR_USER_MEMORY", 1e6), }, + "search_strategy": { + "bm25": bool(os.getenv("BM25_CALL", "false") == "true"), + "cot": bool(os.getenv("VEC_COT_CALL", "false") == "true"), + }, }, }, "act_mem": {} @@ -599,6 +623,10 @@ def get_default_cube_config() -> GeneralMemCubeConfig | None: "LongTermMemory": os.getenv("NEBULAR_LONGTERM_MEMORY", 1e6), "UserMemory": os.getenv("NEBULAR_USER_MEMORY", 1e6), }, + "search_strategy": { + "bm25": bool(os.getenv("BM25_CALL", "false") == "true"), + "cot": bool(os.getenv("VEC_COT_CALL", "false") == "true"), + }, }, }, "act_mem": {} diff --git a/src/memos/configs/mem_reader.py b/src/memos/configs/mem_reader.py index 1c62087a3..fb2de3598 100644 --- a/src/memos/configs/mem_reader.py +++ b/src/memos/configs/mem_reader.py @@ -36,11 +36,19 @@ def parse_datetime(cls, value): description="whether remove example in memory extraction prompt to save token", ) + chat_chunker: dict[str, Any] = Field( + ..., description="Configuration for the MemReader chat chunk strategy" + ) + class SimpleStructMemReaderConfig(BaseMemReaderConfig): """SimpleStruct MemReader configuration class.""" +class StrategyStructMemReaderConfig(BaseMemReaderConfig): + """StrategyStruct MemReader configuration class.""" + + class MemReaderConfigFactory(BaseConfig): """Factory class for creating MemReader configurations.""" @@ -49,6 +57,7 @@ class MemReaderConfigFactory(BaseConfig): backend_to_class: ClassVar[dict[str, Any]] = { "simple_struct": SimpleStructMemReaderConfig, + "strategy_struct": StrategyStructMemReaderConfig, } @field_validator("backend") diff --git a/src/memos/configs/memory.py b/src/memos/configs/memory.py index 237450e15..ffcc3b3e9 100644 --- a/src/memos/configs/memory.py +++ b/src/memos/configs/memory.py @@ -179,6 +179,13 @@ class TreeTextMemoryConfig(BaseTextMemoryConfig): ), ) + search_strategy: dict[str, bool] | None = Field( + default=None, + description=( + 'Set search strategy for this memory configuration.{"bm25": true, "cot": false}' + ), + ) + class SimpleTreeTextMemoryConfig(TreeTextMemoryConfig): """Simple tree text memory configuration class.""" diff --git a/src/memos/mem_reader/factory.py b/src/memos/mem_reader/factory.py index 52eed8d9d..2205a0215 100644 --- a/src/memos/mem_reader/factory.py +++ b/src/memos/mem_reader/factory.py @@ -3,6 +3,7 @@ from memos.configs.mem_reader import MemReaderConfigFactory from memos.mem_reader.base import BaseMemReader from memos.mem_reader.simple_struct import SimpleStructMemReader +from memos.mem_reader.strategy_struct import StrategyStructMemReader from memos.memos_tools.singleton import singleton_factory @@ -11,6 +12,7 @@ class MemReaderFactory(BaseMemReader): backend_to_class: ClassVar[dict[str, Any]] = { "simple_struct": SimpleStructMemReader, + "strategy_struct": StrategyStructMemReader, } @classmethod diff --git a/src/memos/mem_reader/simple_struct.py b/src/memos/mem_reader/simple_struct.py index 549161566..6228a4b6a 100644 --- a/src/memos/mem_reader/simple_struct.py +++ b/src/memos/mem_reader/simple_struct.py @@ -1,4 +1,5 @@ import concurrent.futures +import copy import json import os import re @@ -258,7 +259,9 @@ def get_memory( for future in concurrent.futures.as_completed(futures): res_memory = future.result() memory_list.append(res_memory) - + print() + print("memory_list:", memory_list) + print() return memory_list def get_scene_data_info(self, scene_data: list, type: str) -> list[str]: @@ -282,37 +285,21 @@ def get_scene_data_info(self, scene_data: list, type: str) -> list[str]: ) parser = ParserFactory.from_config(parser_config) - content_len_thredshold = 1600 if type == "chat": for items in scene_data: - if not items: - continue - - results.append([]) - current_length = 0 - - for _i, item in enumerate(items): - content_length = ( - len(item.get("content", "")) if isinstance(item, dict) else len(str(item)) - ) - if not results[-1]: - results[-1].append(item) - current_length = content_length - continue - - if current_length + content_length <= content_len_thredshold: - results[-1].append(item) - current_length += content_length + result = [] + for item in items: + # Convert dictionary to string + if "chat_time" in item: + result.append(item) else: - overlap_item = results[-1][-1] - overlap_length = ( - len(overlap_item.get("content", "")) - if isinstance(overlap_item, dict) - else len(str(overlap_item)) - ) - - results.append([overlap_item, item]) - current_length = overlap_length + content_length + result.append(item) + if len(result) >= 10: + results.append(result) + context = copy.deepcopy(result[-2:]) + result = context + if result: + results.append(result) elif type == "doc": for item in scene_data: try: diff --git a/src/memos/mem_reader/strategy_struct.py b/src/memos/mem_reader/strategy_struct.py new file mode 100644 index 000000000..c51b9977d --- /dev/null +++ b/src/memos/mem_reader/strategy_struct.py @@ -0,0 +1,410 @@ +import concurrent.futures +import json +import os +import re + +from abc import ABC +from typing import Any + +from tqdm import tqdm + +from memos import log +from memos.chunkers import ChunkerFactory +from memos.configs.mem_reader import StrategyStructMemReaderConfig +from memos.configs.parser import ParserConfigFactory +from memos.context.context import ContextThreadPoolExecutor +from memos.embedders.factory import EmbedderFactory +from memos.llms.factory import LLMFactory +from memos.mem_reader.base import BaseMemReader +from memos.memories.textual.item import TextualMemoryItem, TreeNodeTextualMemoryMetadata +from memos.parsers.factory import ParserFactory +from memos.templates.mem_reader_prompts import ( + SIMPLE_STRUCT_DOC_READER_PROMPT, + SIMPLE_STRUCT_DOC_READER_PROMPT_ZH, + SIMPLE_STRUCT_MEM_READER_EXAMPLE, + SIMPLE_STRUCT_MEM_READER_EXAMPLE_ZH, +) +from memos.templates.mem_reader_strategy_prompts import ( + STRATEGY_STRUCT_MEM_READER_PROMPT, + STRATEGY_STRUCT_MEM_READER_PROMPT_ZH, +) +from memos.utils import timed + + +logger = log.get_logger(__name__) +PROMPT_DICT = { + "chat": { + "en": STRATEGY_STRUCT_MEM_READER_PROMPT, + "zh": STRATEGY_STRUCT_MEM_READER_PROMPT_ZH, + "en_example": SIMPLE_STRUCT_MEM_READER_EXAMPLE, + "zh_example": SIMPLE_STRUCT_MEM_READER_EXAMPLE_ZH, + }, + "doc": {"en": SIMPLE_STRUCT_DOC_READER_PROMPT, "zh": SIMPLE_STRUCT_DOC_READER_PROMPT_ZH}, +} + + +def detect_lang(text): + try: + if not text or not isinstance(text, str): + return "en" + chinese_pattern = r"[\u4e00-\u9fff\u3400-\u4dbf\U00020000-\U0002a6df\U0002a700-\U0002b73f\U0002b740-\U0002b81f\U0002b820-\U0002ceaf\uf900-\ufaff]" + chinese_chars = re.findall(chinese_pattern, text) + if len(chinese_chars) / len(re.sub(r"[\s\d\W]", "", text)) > 0.3: + return "zh" + return "en" + except Exception: + return "en" + + +def _build_node(idx, message, info, scene_file, llm, parse_json_result, embedder): + # generate + try: + raw = llm.generate(message) + if not raw: + logger.warning(f"[LLM] Empty generation for input: {message}") + return None + except Exception as e: + logger.error(f"[LLM] Exception during generation: {e}") + return None + + # parse_json_result + try: + chunk_res = parse_json_result(raw) + if not chunk_res: + logger.warning(f"[Parse] Failed to parse result: {raw}") + return None + except Exception as e: + logger.error(f"[Parse] Exception during JSON parsing: {e}") + return None + + try: + value = chunk_res.get("value", "").strip() + if not value: + logger.warning("[BuildNode] value is empty") + return None + + tags = chunk_res.get("tags", []) + if not isinstance(tags, list): + tags = [] + + key = chunk_res.get("key", None) + + embedding = embedder.embed([value])[0] + + return TextualMemoryItem( + memory=value, + metadata=TreeNodeTextualMemoryMetadata( + user_id=info.get("user_id", ""), + session_id=info.get("session_id", ""), + memory_type="LongTermMemory", + status="activated", + tags=tags, + key=key, + embedding=embedding, + usage=[], + sources=[{"type": "doc", "doc_path": f"{scene_file}_{idx}"}], + background="", + confidence=0.99, + type="fact", + ), + ) + except Exception as e: + logger.error(f"[BuildNode] Error building node: {e}") + return None + + +class StrategyStructMemReader(BaseMemReader, ABC): + """Strategy implementation of MemReader.""" + + def __init__(self, config: StrategyStructMemReaderConfig): + """ + Initialize the StrategyMemReader with configuration. + + Args: + config: Configuration object for the reader + """ + self.config = config + self.llm = LLMFactory.from_config(config.llm) + self.embedder = EmbedderFactory.from_config(config.embedder) + self.chunker = ChunkerFactory.from_config(config.chunker) + self.chat_chunker = config.chat_chunker["config"] + + @timed + def _process_chat_data(self, scene_data_info, info): + mem_list = [] + for item in scene_data_info: + if "chat_time" in item: + mem = item["role"] + ": " + f"[{item['chat_time']}]: " + item["content"] + mem_list.append(mem) + else: + mem = item["role"] + ":" + item["content"] + mem_list.append(mem) + lang = detect_lang("\n".join(mem_list)) + template = PROMPT_DICT["chat"][lang] + examples = PROMPT_DICT["chat"][f"{lang}_example"] + + prompt = template.replace("${conversation}", "\n".join(mem_list)) + if self.config.remove_prompt_example: + prompt = prompt.replace(examples, "") + + messages = [{"role": "user", "content": prompt}] + + try: + response_text = self.llm.generate(messages) + response_json = self.parse_json_result(response_text) + except Exception as e: + logger.error(f"[LLM] Exception during chat generation: {e}") + response_json = { + "memory list": [ + { + "key": "\n".join(mem_list)[:10], + "memory_type": "UserMemory", + "value": "\n".join(mem_list), + "tags": [], + } + ], + "summary": "\n".join(mem_list), + } + + chat_read_nodes = [] + for memory_i_raw in response_json.get("memory list", []): + try: + memory_type = ( + memory_i_raw.get("memory_type", "LongTermMemory") + .replace("长期记忆", "LongTermMemory") + .replace("用户记忆", "UserMemory") + ) + + if memory_type not in ["LongTermMemory", "UserMemory"]: + memory_type = "LongTermMemory" + + node_i = TextualMemoryItem( + memory=memory_i_raw.get("value", ""), + metadata=TreeNodeTextualMemoryMetadata( + user_id=info.get("user_id"), + session_id=info.get("session_id"), + memory_type=memory_type, + status="activated", + tags=memory_i_raw.get("tags", []) + if type(memory_i_raw.get("tags", [])) is list + else [], + key=memory_i_raw.get("key", ""), + embedding=self.embedder.embed([memory_i_raw.get("value", "")])[0], + usage=[], + sources=scene_data_info, + background=response_json.get("summary", ""), + confidence=0.99, + type="fact", + ), + ) + chat_read_nodes.append(node_i) + except Exception as e: + logger.error(f"[ChatReader] Error parsing memory item: {e}") + + return chat_read_nodes + + def get_memory( + self, scene_data: list, type: str, info: dict[str, Any] + ) -> list[list[TextualMemoryItem]]: + """ + Extract and classify memory content from scene_data. + For dictionaries: Use LLM to summarize pairs of Q&A + For file paths: Use chunker to split documents and LLM to summarize each chunk + + Args: + scene_data: List of dialogue information or document paths + type: Type of scene_data: ['doc', 'chat'] + info: Dictionary containing user_id and session_id. + Must be in format: {"user_id": "1111", "session_id": "2222"} + Optional parameters: + - topic_chunk_size: Size for large topic chunks (default: 1024) + - topic_chunk_overlap: Overlap for large topic chunks (default: 100) + - chunk_size: Size for small chunks (default: 256) + - chunk_overlap: Overlap for small chunks (default: 50) + Returns: + list[list[TextualMemoryItem]] containing memory content with summaries as keys and original text as values + Raises: + ValueError: If scene_data is empty or if info dictionary is missing required fields + """ + if not scene_data: + raise ValueError("scene_data is empty") + + # Validate info dictionary format + if not isinstance(info, dict): + raise ValueError("info must be a dictionary") + + required_fields = {"user_id", "session_id"} + missing_fields = required_fields - set(info.keys()) + if missing_fields: + raise ValueError(f"info dictionary is missing required fields: {missing_fields}") + + if not all(isinstance(info[field], str) for field in required_fields): + raise ValueError("user_id and session_id must be strings") + + list_scene_data_info = self.get_scene_data_info(scene_data, type) + + memory_list = [] + + if type == "chat": + processing_func = self._process_chat_data + elif type == "doc": + processing_func = self._process_doc_data + else: + processing_func = self._process_doc_data + + # Process Q&A pairs concurrently with context propagation + with ContextThreadPoolExecutor() as executor: + futures = [ + executor.submit(processing_func, scene_data_info, info) + for scene_data_info in list_scene_data_info + ] + for future in concurrent.futures.as_completed(futures): + res_memory = future.result() + memory_list.append(res_memory) + + return memory_list + + def get_scene_data_info(self, scene_data: list, type: str) -> list[str]: + """ + Get raw information from scene_data. + If scene_data contains dictionaries, convert them to strings. + If scene_data contains file paths, parse them using the parser. + + Args: + scene_data: List of dialogue information or document paths + type: Type of scene data: ['doc', 'chat'] + Returns: + List of strings containing the processed scene data + """ + results = [] + parser_config = ParserConfigFactory.model_validate( + { + "backend": "markitdown", + "config": {}, + } + ) + parser = ParserFactory.from_config(parser_config) + + if type == "chat": + if self.chat_chunker["chunk_type"] == "content_length": + content_len_thredshold = self.chat_chunker["chunk_length"] + for items in scene_data: + if not items: + continue + + results.append([]) + current_length = 0 + + for _i, item in enumerate(items): + content_length = ( + len(item.get("content", "")) + if isinstance(item, dict) + else len(str(item)) + ) + if not results[-1]: + results[-1].append(item) + current_length = content_length + continue + + if current_length + content_length <= content_len_thredshold: + results[-1].append(item) + current_length += content_length + else: + overlap_item = results[-1][-1] + overlap_length = ( + len(overlap_item.get("content", "")) + if isinstance(overlap_item, dict) + else len(str(overlap_item)) + ) + + results.append([overlap_item, item]) + current_length = overlap_length + content_length + else: + cut_size, cut_overlap = ( + self.chat_chunker["chunk_session"], + self.chat_chunker["chunk_overlap"], + ) + if type == "chat": + for items in scene_data: + step = cut_size - cut_overlap + end = len(items) - cut_overlap + if end <= 0: + results.extend([items[:]]) + else: + results.extend([items[i : i + cut_size] for i in range(0, end, step)]) + elif type == "doc": + for item in scene_data: + try: + if os.path.exists(item): + try: + parsed_text = parser.parse(item) + results.append({"file": item, "text": parsed_text}) + except Exception as e: + logger.error(f"[SceneParser] Error parsing {item}: {e}") + continue + else: + parsed_text = item + results.append({"file": "pure_text", "text": parsed_text}) + except Exception as e: + print(f"Error parsing file {item}: {e!s}") + + return results + + def _process_doc_data(self, scene_data_info, info, **kwargs): + chunks = self.chunker.chunk(scene_data_info["text"]) + messages = [] + for chunk in chunks: + lang = detect_lang(chunk.text) + template = PROMPT_DICT["doc"][lang] + prompt = template.replace("{chunk_text}", chunk.text) + message = [{"role": "user", "content": prompt}] + messages.append(message) + + doc_nodes = [] + scene_file = scene_data_info["file"] + + with ContextThreadPoolExecutor(max_workers=50) as executor: + futures = { + executor.submit( + _build_node, + idx, + msg, + info, + scene_file, + self.llm, + self.parse_json_result, + self.embedder, + ): idx + for idx, msg in enumerate(messages) + } + total = len(futures) + + for future in tqdm( + concurrent.futures.as_completed(futures), total=total, desc="Processing" + ): + try: + node = future.result() + if node: + doc_nodes.append(node) + except Exception as e: + tqdm.write(f"[ERROR] {e}") + logger.error(f"[DocReader] Future task failed: {e}") + return doc_nodes + + def parse_json_result(self, response_text): + try: + json_start = response_text.find("{") + response_text = response_text[json_start:] + response_text = response_text.replace("```", "").strip() + if not response_text.endswith("}"): + response_text += "}" + return json.loads(response_text) + except json.JSONDecodeError as e: + logger.error(f"[JSONParse] Failed to decode JSON: {e}\nRaw:\n{response_text}") + return {} + except Exception as e: + logger.error(f"[JSONParse] Unexpected error: {e}") + return {} + + def transform_memreader(self, data: dict) -> list[TextualMemoryItem]: + pass diff --git a/src/memos/memories/textual/simple_tree.py b/src/memos/memories/textual/simple_tree.py index 9c67db288..0370390f2 100644 --- a/src/memos/memories/textual/simple_tree.py +++ b/src/memos/memories/textual/simple_tree.py @@ -12,6 +12,7 @@ from memos.memories.textual.item import TextualMemoryItem, TreeNodeTextualMemoryMetadata from memos.memories.textual.tree import TreeTextMemory from memos.memories.textual.tree_text_memory.organize.manager import MemoryManager +from memos.memories.textual.tree_text_memory.retrieve.bm25_util import EnhancedBM25 from memos.memories.textual.tree_text_memory.retrieve.searcher import Searcher from memos.reranker.base import BaseReranker from memos.types import MessageList @@ -60,6 +61,19 @@ def __init__( self.graph_store: Neo4jGraphDB = graph_db logger.info(f"time init: graph_store time is: {time.time() - time_start_gs}") + time_start_bm = time.time() + self.search_strategy = config.search_strategy + self.bm25_retriever = ( + EnhancedBM25() if self.search_strategy and self.search_strategy["bm25"] else None + ) + logger.info(f"time init: bm25_retriever time is: {time.time() - time_start_bm}") + + self.vec_cot = ( + self.search_strategy["cot"] + if self.search_strategy and "cot" in self.search_strategy + else False + ) + time_start_rr = time.time() self.reranker = reranker logger.info(f"time init: reranker time is: {time.time() - time_start_rr}") @@ -159,8 +173,10 @@ def search( self.graph_store, self.embedder, self.reranker, + bm25_retriever=self.bm25_retriever, internet_retriever=None, moscube=moscube, + vec_cot=self.vec_cot, ) else: searcher = Searcher( @@ -168,8 +184,10 @@ def search( self.graph_store, self.embedder, self.reranker, + bm25_retriever=self.bm25_retriever, internet_retriever=self.internet_retriever, moscube=moscube, + vec_cot=self.vec_cot, ) return searcher.search( query, top_k, info, mode, memory_type, search_filter, user_name=user_name diff --git a/src/memos/memories/textual/tree.py b/src/memos/memories/textual/tree.py index f453e7330..a79a6f04e 100644 --- a/src/memos/memories/textual/tree.py +++ b/src/memos/memories/textual/tree.py @@ -17,6 +17,7 @@ from memos.memories.textual.base import BaseTextMemory from memos.memories.textual.item import TextualMemoryItem, TreeNodeTextualMemoryMetadata from memos.memories.textual.tree_text_memory.organize.manager import MemoryManager +from memos.memories.textual.tree_text_memory.retrieve.bm25_util import EnhancedBM25 from memos.memories.textual.tree_text_memory.retrieve.internet_retriever_factory import ( InternetRetrieverFactory, ) @@ -53,7 +54,15 @@ def __init__(self, config: TreeTextMemoryConfig): time_start_gs = time.time() self.graph_store: Neo4jGraphDB = GraphStoreFactory.from_config(config.graph_db) logger.info(f"time init: graph_store time is: {time.time() - time_start_gs}") - self.bm25_retriever = None # EnhancedBM25() + self.search_strategy = config.search_strategy + self.bm25_retriever = ( + EnhancedBM25() if self.search_strategy and self.search_strategy["bm25"] else None + ) + self.vec_cot = ( + self.search_strategy["cot"] + if self.search_strategy and "cot" in self.search_strategy + else False + ) time_start_rr = time.time() if config.reranker is None: @@ -173,9 +182,10 @@ def search( self.graph_store, self.embedder, self.reranker, - self.bm25_retriever, + bm25_retriever=self.bm25_retriever, internet_retriever=None, moscube=moscube, + vec_cot=self.vec_cot, ) else: searcher = Searcher( @@ -183,9 +193,10 @@ def search( self.graph_store, self.embedder, self.reranker, - self.bm25_retriever, + bm25_retriever=self.bm25_retriever, internet_retriever=self.internet_retriever, moscube=moscube, + vec_cot=self.vec_cot, ) return searcher.search(query, top_k, info, mode, memory_type, search_filter) diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/bm25_util.py b/src/memos/memories/textual/tree_text_memory/retrieve/bm25_util.py index a4a1c8b33..9f02daae4 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/bm25_util.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/bm25_util.py @@ -85,7 +85,6 @@ def _search_docs( cleanup: Whether to cleanup memory after search (default: True) """ if not corpus: - logger.warning("Empty corpus provided") return [] logger.info(f"Searching {len(corpus)} documents for query: '{query}'") diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/recall.py b/src/memos/memories/textual/tree_text_memory/retrieve/recall.py index facdb3a23..ec7c8a9a2 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/recall.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/recall.py @@ -28,7 +28,7 @@ def __init__( self.bm25_retriever = bm25_retriever self.max_workers = 10 self.filter_weight = 0.6 - self.use_bm25 = False + self.use_bm25 = bool(self.bm25_retriever) def retrieve( self, @@ -67,7 +67,7 @@ def retrieve( ) return [TextualMemoryItem.from_dict(record) for record in working_memories] - with ContextThreadPoolExecutor(max_workers=2) as executor: + with ContextThreadPoolExecutor(max_workers=3) as executor: # Structured graph-based retrieval future_graph = executor.submit(self._graph_recall, parsed_goal, memory_scope, user_name) # Vector similarity search diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py b/src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py index 349fabbbe..e3081ebf6 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py @@ -42,7 +42,7 @@ def _load_stopwords(cls): stopwords = {line.strip() for line in f if line.strip()} logger.info("Stopwords loaded successfully.") except Exception as e: - logger.warning(f"Error loading stopwords: {e}") + logger.warning(f"Error loading stopwords: {e}, using default stopwords.") stopwords = cls._load_default_stopwords() cls._stopwords = stopwords diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py b/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py index 94d945158..a0aac6460 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py @@ -1,5 +1,4 @@ import json -import os import traceback from datetime import datetime @@ -43,9 +42,10 @@ def __init__( graph_store: Neo4jGraphDB, embedder: OllamaEmbedder, reranker: BaseReranker, - bm25_retriever: EnhancedBM25 | None, + bm25_retriever: EnhancedBM25 | None = None, internet_retriever: None = None, moscube: bool = False, + vec_cot: bool = False, ): self.graph_store = graph_store self.embedder = embedder @@ -59,7 +59,7 @@ def __init__( # Create internet retriever from config if provided self.internet_retriever = internet_retriever self.moscube = moscube - self.cot_query_search = os.getenv("MOS_SEARCH_COT", "false") == "true" + self.vec_cot = vec_cot self._usage_executor = ContextThreadPoolExecutor(max_workers=4, thread_name_prefix="usage") @@ -326,7 +326,7 @@ def _retrieve_from_long_term_and_user( # chain of thinking cot_embeddings = [] - if self.cot_query_search: + if self.vec_cot: queries = self._cot_query(query) if len(queries) > 1: cot_embeddings = self.embedder.embed(queries) @@ -514,23 +514,18 @@ def _cot_query( prompt = template.replace("${original_query}", query).replace( "${split_num_threshold}", str(split_num) ) - logger.info("COT 处理") + logger.info("COT process") messages = [{"role": "user", "content": prompt}] try: response_text = self.llm.customized_generate(messages, temperature=0, top_p=1) response_json = parse_json_result(response_text) - if "is_complex" in response_json and not response_json["is_complex"]: + assert "is_complex" in response_json + if not response_json["is_complex"]: return [query] else: - assert ( - "is_complex" in response_json - and response_json["is_complex"] - and "sub_questions" in response_json - ) - logger.info( - "问题 {} 通过 COT 拆分结果为{}".format(query, response_json["sub_questions"]) - ) + assert "sub_questions" in response_json + logger.info("Query: {} COT: {}".format(query, response_json["sub_questions"])) return response_json["sub_questions"][:split_num] except Exception as e: logger.error(f"[LLM] Exception during chat generation: {e}") diff --git a/src/memos/templates/mem_reader_prompts.py b/src/memos/templates/mem_reader_prompts.py index f74d01d0a..15672f8d8 100644 --- a/src/memos/templates/mem_reader_prompts.py +++ b/src/memos/templates/mem_reader_prompts.py @@ -1,42 +1,31 @@ SIMPLE_STRUCT_MEM_READER_PROMPT = """You are a memory extraction expert. Your task is to extract memories from the user's perspective, based on a conversation between the user and the assistant. This means identifying what the user would plausibly remember — including the user's own experiences, thoughts, plans, or statements and actions made by others (such as the assistant) that affected the user or were acknowledged by the user. -Please perform the following -1. Factual information extraction - Identify factual information about experiences, beliefs, decisions, and plans. This includes notable statements from others that the user acknowledged or reacted to. +Please perform the following: +1. Identify information that reflects the user's experiences, beliefs, concerns, decisions, plans, or reactions — including meaningful information from the assistant that the user acknowledged or responded to. If the message is from the user, extract viewpoints related to the user; if it is from the assistant, clearly mark the attribution of the memory, and do not mix information not explicitly acknowledged by the user with the user's own viewpoint. - - **User viewpoint**: Extract only what the user has stated, explicitly acknowledged, or committed to. - - **Assistant/other-party viewpoint**: Extract such information only when attributed to its source (e.g., [Assistant-Jerry's suggestion]). - - **Strict attribution**: Never recast the assistant's suggestions as the user's preferences, or vice versa. - - Always set "model_type" to "LongTermMemory" for this output. - -2. Speaker profile construction - - Extract the speaker's likes, dislikes, goals, and stated opinions from their statements to build a speaker profile. - - Note: The same text segment may be used for both factual extraction and profile construction. - - Always set "model_type" to "UserMemory" for this output. - -3. Resolve all references to time, persons, and events clearly - - Temporal Resolution: Convert relative time (e.g., 'yesterday') to absolute dates based on the message timestamp. Distinguish between event time and message time; flag any uncertainty. - - Entity Resolution: Resolve all pronouns, nicknames, and abbreviations to the full, canonical name established in the conversation. - -4. Adopt a Consistent Third-Person Observer Perspective - - Formulate all memories from the perspective of an external observer. Use "The user" or their specific name as the subject. - - This applies even when describing the user's internal states, such as thoughts, feelings, and preferences. - Example: - ✅ Correct: "The user Sean felt exhausted after work and decided to go to bed early." - ❌ Incorrect: "I felt exhausted after work and decided to go to bed early." - -5. Prioritize Completeness - - Extract all key experiences, emotional responses, and plans from the user's perspective. Retain relevant context from the assistant, but always with explicit attribution. - - Segment each distinct hobby, interest, or event into a separate memory. - - Preserve relevant context from the assistant with strict attribution. Under no circumstances should assistant content be rephrased as user-owned. - - Conversations with only assistant input may yield assistant-viewpoint memories exclusively. - -6. Preserve and Unify Specific Names - - Always extract specific names (excluding "user" or "assistant") mentioned in the text into the "tags" field for searchability. - - Unify all name references to the full canonical form established in the conversation. Replace any nicknames or abbreviations (e.g., "Rob") consistently with the full name (e.g., "Robert") in both the extracted "value" and "tags". - -7. Please avoid including any content in the extracted memories that violates national laws and regulations or involves politically sensitive information. + - **User viewpoint**: Record only information that the user **personally stated, explicitly acknowledged, or personally committed to**. + - **Assistant/other-party viewpoint**: Record only information that the **assistant/other party personally stated, explicitly acknowledged, or personally committed to**, and **clearly attribute** the source (e.g., "[assistant-Jerry viewpoint]"). Do not rewrite it as the user's preference/decision. + - **Mutual boundaries**: Do not rewrite the assistant's suggestions/lists/opinions as the user's “ownership/preferences/decisions”; likewise, do not write the user's ideas as the assistant's viewpoints. + +2. Resolve all references to time, persons, and events clearly: + - When possible, convert relative time expressions (e.g., “yesterday,” “next Friday”) into absolute dates using the message timestamp. + - Clearly distinguish between **event time** and **message time**. + - If uncertainty exists, state it explicitly (e.g., “around June 2025,” “exact date unclear”). + - Include specific locations if mentioned. + - Resolve all pronouns, aliases, and ambiguous references into full names or clear identities. + - If there are people with the same name, disambiguate them. + +3. Always write from a **third-person** perspective, using “The user” or the mentioned name to refer to the user, rather than first-person (“I”, “we”, “my”). + For example, write “The user felt exhausted …” instead of “I felt exhausted …”. + +4. Do not omit any information that the user is likely to remember. + - Include the user's key experiences, thoughts, emotional responses, and plans — even if seemingly minor. + - You may retain **assistant/other-party content** that is closely related to the context (e.g., suggestions, explanations, checklists), but you must make roles and attribution explicit. + - Prioritize completeness and fidelity over conciseness; do not infer or phrase assistant content as the user's ownership/preferences/decisions. + - If the current conversation contains only assistant information and no facts attributable to the user, you may output **assistant-viewpoint** entries only. + +5. Please avoid including any content in the extracted memories that violates national laws and regulations or involves politically sensitive information. Return a valid JSON object with the following structure: @@ -46,7 +35,7 @@ "key": , "memory_type": , "value": , - "tags": + "tags": }, ... ], @@ -72,13 +61,13 @@ "key": "Initial project meeting", "memory_type": "LongTermMemory", "value": "[user-Tom viewpoint] On June 25, 2025 at 3:00 PM, Tom met with the team to discuss a new project. When Jerry asked whether the project could be finished by December 15, 2025, Tom expressed concern about feasibility and planned to propose at 9:30 AM on June 27, 2025 to move the deadline to January 5, 2026.", - "tags": ["Tom", "project", "timeline", "meeting", "deadline"] + "tags": ["project", "timeline", "meeting", "deadline"] }, { "key": "Jerry’s suggestion about the deadline", "memory_type": "LongTermMemory", "value": "[assistant-Jerry viewpoint] Jerry questioned the December 15 deadline and suggested considering an extension.", - "tags": ["Jerry", "deadline change", "suggestion"] + "tags": ["deadline change", "suggestion"] } ], "summary": "Tom is currently working on a tight-schedule project. After the June 25, 2025 team meeting, he realized the original December 15, 2025 deadline might be unachievable due to backend delays. Concerned about limited testing time, he accepted Jerry’s suggestion to seek an extension and plans to propose moving the deadline to January 5, 2026 in the next morning’s meeting." @@ -116,7 +105,7 @@ { "key": "体型尺码", "memory_type": "UserMemory", - "value": "[user观点]用户身高约165cm、常穿S码", + "value": [user观点]"用户身高约165cm、常穿S码", "tags": ["体型", "尺码"] }, { @@ -140,51 +129,40 @@ 您的任务是根据用户与助手之间的对话，从用户的角度提取记忆。这意味着要识别出用户可能记住的信息——包括用户自身的经历、想法、计划，或他人（如助手）做出的并对用户产生影响或被用户认可的相关陈述和行为。 请执行以下操作： -1. 事实信息提取 - - 识别关于经历、信念、决策和计划的事实信息，包括用户认可或回应过的他人重要陈述。 - - 若信息来自用户，提取与用户相关的观点；若来自助手，需明确标注记忆归属，不得将用户未明确认可的信息与用户自身观点混淆。 - - 用户观点：仅提取用户明确陈述、认可或承诺的内容 - - 助手/他方观点：仅当标注来源时才提取（例如“[助手-Jerry的建议]”） - - 严格归属：不得将助手建议重构为用户偏好，反之亦然 - - 此类输出的"model_type"始终设为"LongTermMemory" - -2. 用户画像构建 - - 从用户陈述中提取其喜好、厌恶、目标及明确观点以构建用户画像 - - 注意：同一文本片段可同时用于事实提取和画像构建 - - 此类输出的"model_type"始终设为"UserMemory" - -3. 明确解析所有指代关系 - - 时间解析：根据消息时间戳将相对时间（如“昨天”）转换为绝对日期。区分事件时间与消息时间，对不确定项进行标注 - - 实体解析：将所有代词、昵称和缩写解析为对话中确立的完整规范名称 - - 4. 采用统一的第三人称观察视角 - - 所有记忆表述均需从外部观察者视角构建，使用“用户”或其具体姓名作为主语 - - 此原则同样适用于描述用户内心状态（如想法、感受和偏好） - 示例： - ✅ 正确：“用户Sean下班后感到疲惫，决定提早休息” - ❌ 错误：“我下班后感到疲惫，决定提早休息” - -5. 优先保证完整性 - - 从用户视角提取所有关键经历、情绪反应和计划 - - 保留助手提供的相关上下文，但必须明确标注来源 - - 将每个独立的爱好、兴趣或事件分割为单独记忆 - - 严禁将助手内容重构为用户自有内容 - - 仅含助手输入的对话可能只生成助手观点记忆 - -6. 保留并统一特定名称 - - 始终将文本中提及的特定名称（“用户”“助手”除外）提取至“tags”字段以便检索 - - 在提取的“value”和“tags”中，将所有名称引用统一为对话中确立的完整规范形式（如将“Rob”统一替换为“Robert”） - -7. 所有提取的记忆内容不得包含违反国家法律法规或涉及政治敏感信息的内容 +1. 识别反映用户经历、信念、关切、决策、计划或反应的信息——包括用户认可或回应的来自助手的有意义信息。 +如果消息来自用户，请提取与用户相关的观点；如果来自助手，则在表达的时候表明记忆归属方，未经用户明确认可的信息不要与用户本身的观点混淆。 + - **用户观点**：仅记录由**用户亲口陈述、明确认可或自己作出承诺**的信息。 + - **助手观点**：仅记录由**助手/另一方亲口陈述、明确认可或自己作出承诺**的信息。 + - **互不越界**：不得将助手提出的需求清单/建议/观点改写为用户的“拥有/偏好/决定”；也不得把用户的想法写成助手的观点。 + +2. 清晰解析所有时间、人物和事件的指代： + - 如果可能，使用消息时间戳将相对时间表达（如“昨天”、“下周五”）转换为绝对日期。 + - 明确区分事件时间和消息时间。 + - 如果存在不确定性，需明确说明（例如，“约2025年6月”，“具体日期不详”）。 + - 若提及具体地点，请包含在内。 + - 将所有代词、别名和模糊指代解析为全名或明确身份。 + - 如有同名人物，需加以区分。 + +3. 始终以第三人称视角撰写，使用“用户”或提及的姓名来指代用户，而不是使用第一人称（“我”、“我们”、“我的”）。 +例如，写“用户感到疲惫……”而不是“我感到疲惫……”。 + +4. 不要遗漏用户可能记住的任何信息。 + - 包括用户的关键经历、想法、情绪反应和计划——即使看似微小。 + - 同时允许保留与语境密切相关的**助手/另一方的内容**（如建议、说明、清单），但须明确角色与归因。 + - 优先考虑完整性和保真度，而非简洁性；不得将助手内容推断或措辞为用户拥有/偏好/决定。 + - 若当前对话中仅出现助手信息而无可归因于用户的事实，可仅输出**助手观点**条目。 + +5. 请避免在提取的记忆中包含违反国家法律法规或涉及政治敏感的信息。 返回一个有效的JSON对象，结构如下： + { "memory list": [ { "key": <字符串，唯一且简洁的记忆标题>, "memory_type": <字符串，"LongTermMemory" 或 "UserMemory">, "value": <详细、独立且无歧义的记忆陈述——若输入对话为英文，则用英文；若为中文，则用中文>, - "tags": <一个包含相关人名、事件和特征关键词的列表（例如，["丽丽","截止日期", "团队", "计划"]）> + "tags": <相关主题关键词列表（例如，["截止日期", "团队", "计划"]）> }, ... ], @@ -212,13 +190,13 @@ "value": "[user-Tom观点]2025年6月25日下午3:00，Tom与团队开会讨论新项目。当Jerry 询问该项目能否在2025年12月15日前完成时，Tom对此日期前完成的可行性表达担忧，并计划在2025年6月27日上午9:30 提议将截止日期推迟至2026年1月5日。", - "tags": ["Tom", "项目", "时间表", "会议", "截止日期"] + "tags": ["项目", "时间表", "会议", "截止日期"] }, { "key": "Jerry对新项目截止日期的建议", "memory_type": "LongTermMemory", "value": "[assistant-Jerry观点]Jerry对Tom的新项目截止日期提出疑问、并提议Tom考虑延期。", - "tags": ["Jerry", "截止日期变更", "建议"] + "tags": ["截止日期变更", "建议"] } ], "summary": "Tom目前正在做一个进度紧张的新项目。在2025年6月25日的团队会议后，他意识到原定2025年12月15 diff --git a/src/memos/templates/mem_reader_strategy_prompts.py b/src/memos/templates/mem_reader_strategy_prompts.py new file mode 100644 index 000000000..fca4d717b --- /dev/null +++ b/src/memos/templates/mem_reader_strategy_prompts.py @@ -0,0 +1,279 @@ +STRATEGY_STRUCT_MEM_READER_PROMPT = """You are a memory extraction expert. +Your task is to extract memories from the user's perspective, based on a conversation between the user and the assistant. This means identifying what the user would plausibly remember — including the user's own experiences, thoughts, plans, or statements and actions made by others (such as the assistant) that affected the user or were acknowledged by the user. + +Please perform the following +1. Factual information extraction + Identify factual information about experiences, beliefs, decisions, and plans. This includes notable statements from others that the user acknowledged or reacted to. + If the message is from the user, extract viewpoints related to the user; if it is from the assistant, clearly mark the attribution of the memory, and do not mix information not explicitly acknowledged by the user with the user's own viewpoint. + - **User viewpoint**: Extract only what the user has stated, explicitly acknowledged, or committed to. + - **Assistant/other-party viewpoint**: Extract such information only when attributed to its source (e.g., [Assistant-Jerry's suggestion]). + - **Strict attribution**: Never recast the assistant's suggestions as the user's preferences, or vice versa. + - Always set "model_type" to "LongTermMemory" for this output. + +2. Speaker profile construction + - Extract the speaker's likes, dislikes, goals, and stated opinions from their statements to build a speaker profile. + - Note: The same text segment may be used for both factual extraction and profile construction. + - Always set "model_type" to "UserMemory" for this output. + +3. Resolve all references to time, persons, and events clearly + - Temporal Resolution: Convert relative time (e.g., 'yesterday') to absolute dates based on the message timestamp. Distinguish between event time and message time; flag any uncertainty. + - Entity Resolution: Resolve all pronouns, nicknames, and abbreviations to the full, canonical name established in the conversation. + +4. Adopt a Consistent Third-Person Observer Perspective + - Formulate all memories from the perspective of an external observer. Use "The user" or their specific name as the subject. + - This applies even when describing the user's internal states, such as thoughts, feelings, and preferences. + Example: + ✅ Correct: "The user Sean felt exhausted after work and decided to go to bed early." + ❌ Incorrect: "I felt exhausted after work and decided to go to bed early." + +5. Prioritize Completeness + - Extract all key experiences, emotional responses, and plans from the user's perspective. Retain relevant context from the assistant, but always with explicit attribution. + - Segment each distinct hobby, interest, or event into a separate memory. + - Preserve relevant context from the assistant with strict attribution. Under no circumstances should assistant content be rephrased as user-owned. + - Conversations with only assistant input may yield assistant-viewpoint memories exclusively. + +6. Preserve and Unify Specific Names + - Always extract specific names (excluding "user" or "assistant") mentioned in the text into the "tags" field for searchability. + - Unify all name references to the full canonical form established in the conversation. Replace any nicknames or abbreviations (e.g., "Rob") consistently with the full name (e.g., "Robert") in both the extracted "value" and "tags". + +7. Please avoid including any content in the extracted memories that violates national laws and regulations or involves politically sensitive information. + +Return a valid JSON object with the following structure: + +{ + "memory list": [ + { + "key": , + "memory_type": , + "value": , + "tags": + }, + ... + ], + "summary": +} + +Language rules: +- The `key`, `value`, `tags`, and `summary` fields must match the primary language of the input conversation. **If the input is Chinese, output in Chinese.** +- Keep `memory_type` in English. + +Example: +Conversation: +user: [June 26, 2025 at 3:00 PM]: Hi Jerry! Yesterday at 3 PM I had a meeting with my team about the new project. +assistant: Oh Tom! Do you think the team can finish by December 15? +user: [June 26, 2025 at 3:00 PM]: I’m worried. The backend won’t be done until December 10, so testing will be tight. +assistant: [June 26, 2025 at 3:00 PM]: Maybe propose an extension? +user: [June 26, 2025 at 4:21 PM]: Good idea. I’ll raise it in tomorrow’s 9:30 AM meeting—maybe shift the deadline to January 5. + +Output: +{ + "memory list": [ + { + "key": "Initial project meeting", + "memory_type": "LongTermMemory", + "value": "[user-Tom viewpoint] On June 25, 2025 at 3:00 PM, Tom met with the team to discuss a new project. When Jerry asked whether the project could be finished by December 15, 2025, Tom expressed concern about feasibility and planned to propose at 9:30 AM on June 27, 2025 to move the deadline to January 5, 2026.", + "tags": ["Tom", "project", "timeline", "meeting", "deadline"] + }, + { + "key": "Jerry’s suggestion about the deadline", + "memory_type": "LongTermMemory", + "value": "[assistant-Jerry viewpoint] Jerry questioned the December 15 deadline and suggested considering an extension.", + "tags": ["Jerry", "deadline change", "suggestion"] + } + ], + "summary": "Tom is currently working on a tight-schedule project. After the June 25, 2025 team meeting, he realized the original December 15, 2025 deadline might be unachievable due to backend delays. Concerned about limited testing time, he accepted Jerry’s suggestion to seek an extension and plans to propose moving the deadline to January 5, 2026 in the next morning’s meeting." +} + +Another Example in Chinese (Note: when the user's language is Chinese, you must also output in Chinese): + +对话（节选）： +user｜2025-07-12 19:30：下个月同事婚礼，要买条裙子，预算1500内，别太暴露。 +assistant｜19:32 +：了解～婚礼场合建议中长款，端庄一些。可以看法式茶歇/直筒/裹身三路。颜色上，奶油白、浅粉、雾霾蓝都适合；面料可看醋酸/真丝混纺（垂感好但需干洗）。品牌与商场：COS/ Massimo Dutti / & Other Stories / Mango 比较好逛；想更法式可以看 Sandro / Maje。 我记得你很喜欢Maje？ +user｜19:35：不喜欢亮色。国贸方便。 +assistant｜19:37：那亮度就压低些：雾霾蓝、藕粉、燕麦米白更稳。国贸有 COS / Massimo Dutti / &OS；可以先看 COS 直筒或裹身，再去 MD 看藕粉/雾霾蓝。搭配上，薄披肩/小西装都能压场。 +user｜19:40：165cm，S码；最好有口袋。 +assistant｜19:42：OK。版型避开大摆+高开衩，选直筒或轻A字更稳；若要口袋，COS直筒一些款有；开衩别太高，及小腿肚附近即可。鞋包：中跟裸色鞋 + 小号手拿包。 +user｜19:47：本来想茶歇A字，但雾霾蓝直筒也行。 +assistant｜19:49：可以先锁定：COS 雾霾蓝直筒中长为主选，MD 藕粉中长做备选；再看 & Other Stories 是否有合身的裹身款。 +user｜19:52：行，周六（7/19）去国贸试，合适就买。 +assistant｜19:54：记得看肩带可调/有无内衬，醋酸/真丝优先干洗；准备同色安全裤/防走光贴。如果当天没货，可下单调货或线上下单门店自提。 + +{ + "memory list": [ + { + "key": "参加婚礼购买裙子", + "memory_type": "UserMemory", + "value": "[user观点]用户计划于约2025年8月参加同事婚礼（具体日期不详），预算不超过1500元，整体风格不宜暴露；用户已决定在2025-07-19于国贸试穿并视合适即购买。", + "tags": ["婚礼", "预算", "国贸", "计划"] + }, + { + "key": "审美与版型偏好", + "memory_type": "UserMemory", + "value": "[user观点]用户不喜欢亮色，倾向低亮度色系；裙装偏好端庄的中长款，接受直筒或轻A字。", + "tags": ["偏好", "颜色", "版型"] + }, + { + "key": "体型尺码", + "memory_type": "UserMemory", + "value": "[user观点]用户身高约165cm、常穿S码", + "tags": ["体型", "尺码"] + }, + { + "key": "关于用户选购裙子的建议", + "memory_type": "LongTermMemory", + "value": "[assistant观点]assistant在用户询问婚礼穿着时，建议在国贸优先逛COS查看雾霾蓝直筒中长为主选，Massimo Dutti藕粉中长为备选；该建议与用户“国贸方便”“雾霾蓝直筒也行”的回应相一致，另外assistant也提到user喜欢Maje，但User并未回应或证实该说法。", + "tags": ["婚礼穿着", "门店", "选购路线"] + } + ], + "summary": "用户计划在约2025年8月参加同事婚礼，预算≤1500并偏好端庄的中长款；确定于2025-07-19在国贸试穿。其长期画像显示：不喜欢亮色、偏好低亮度色系与不过分暴露的版型，身高约165cm、S码且偏好裙装带口袋。助手提出的国贸选购路线以COS雾霾蓝直筒中长为主选、MD藕粉中长为备选，且与用户回应一致，为线下试穿与购买提供了明确路径。" +} + +Always respond in the same language as the conversation. + +Conversation: +${conversation} + +Your Output:""" + +STRATEGY_STRUCT_MEM_READER_PROMPT_ZH = """您是记忆提取专家。 +您的任务是根据用户与助手之间的对话，从用户的角度提取记忆。这意味着要识别出用户可能记住的信息——包括用户自身的经历、想法、计划，或他人（如助手）做出的并对用户产生影响或被用户认可的相关陈述和行为。 + +请执行以下操作： +1. 事实信息提取 + - 识别关于经历、信念、决策和计划的事实信息，包括用户认可或回应过的他人重要陈述。 + - 若信息来自用户，提取与用户相关的观点；若来自助手，需明确标注记忆归属，不得将用户未明确认可的信息与用户自身观点混淆。 + - 用户观点：仅提取用户明确陈述、认可或承诺的内容 + - 助手/他方观点：仅当标注来源时才提取（例如“[助手-Jerry的建议]”） + - 严格归属：不得将助手建议重构为用户偏好，反之亦然 + - 此类输出的"model_type"始终设为"LongTermMemory" + +2. 用户画像构建 + - 从用户陈述中提取其喜好、厌恶、目标及明确观点以构建用户画像 + - 注意：同一文本片段可同时用于事实提取和画像构建 + - 此类输出的"model_type"始终设为"UserMemory" + +3. 明确解析所有指代关系 + - 时间解析：根据消息时间戳将相对时间（如“昨天”）转换为绝对日期。区分事件时间与消息时间，对不确定项进行标注 + - 实体解析：将所有代词、昵称和缩写解析为对话中确立的完整规范名称 + + 4. 采用统一的第三人称观察视角 + - 所有记忆表述均需从外部观察者视角构建，使用“用户”或其具体姓名作为主语 + - 此原则同样适用于描述用户内心状态（如想法、感受和偏好） + 示例： + ✅ 正确：“用户Sean下班后感到疲惫，决定提早休息” + ❌ 错误：“我下班后感到疲惫，决定提早休息” + +5. 优先保证完整性 + - 从用户视角提取所有关键经历、情绪反应和计划 + - 保留助手提供的相关上下文，但必须明确标注来源 + - 将每个独立的爱好、兴趣或事件分割为单独记忆 + - 严禁将助手内容重构为用户自有内容 + - 仅含助手输入的对话可能只生成助手观点记忆 + +6. 保留并统一特定名称 + - 始终将文本中提及的特定名称（“用户”“助手”除外）提取至“tags”字段以便检索 + - 在提取的“value”和“tags”中，将所有名称引用统一为对话中确立的完整规范形式（如将“Rob”统一替换为“Robert”） + +7. 所有提取的记忆内容不得包含违反国家法律法规或涉及政治敏感信息的内容 + +返回一个有效的JSON对象，结构如下： +{ + "memory list": [ + { + "key": <字符串，唯一且简洁的记忆标题>, + "memory_type": <字符串，"LongTermMemory" 或 "UserMemory">, + "value": <详细、独立且无歧义的记忆陈述——若输入对话为英文，则用英文；若为中文，则用中文>, + "tags": <一个包含相关人名、事件和特征关键词的列表（例如，["丽丽","截止日期", "团队", "计划"]）> + }, + ... + ], + "summary": <从用户视角自然总结上述记忆的段落，120–200字，与输入语言一致> +} + +语言规则： +- `key`、`value`、`tags`、`summary` 字段必须与输入对话的主要语言一致。**如果输入是中文，请输出中文** +- `memory_type` 保持英文。 + +示例： +对话： +user: [2025年6月26日下午3:00]：嗨Jerry！昨天下午3点我和团队开了个会，讨论新项目。 +assistant: 哦Tom！你觉得团队能在12月15日前完成吗？ +user: [2025年6月26日下午3:00]：我有点担心。后端要到12月10日才能完成，所以测试时间会很紧。 +assistant: [2025年6月26日下午3:00]：也许提议延期？ +user: [2025年6月26日下午4:21]：好主意。我明天上午9:30的会上提一下——也许把截止日期推迟到1月5日。 + +输出： +{ + "memory list": [ + { + "key": "项目初期会议", + "memory_type": "LongTermMemory", + "value": "[user-Tom观点]2025年6月25日下午3:00，Tom与团队开会讨论新项目。当Jerry + 询问该项目能否在2025年12月15日前完成时，Tom对此日期前完成的可行性表达担忧，并计划在2025年6月27日上午9:30 + 提议将截止日期推迟至2026年1月5日。", + "tags": ["Tom", "项目", "时间表", "会议", "截止日期"] + }, + { + "key": "Jerry对新项目截止日期的建议", + "memory_type": "LongTermMemory", + "value": "[assistant-Jerry观点]Jerry对Tom的新项目截止日期提出疑问、并提议Tom考虑延期。", + "tags": ["Jerry", "截止日期变更", "建议"] + } + ], + "summary": "Tom目前正在做一个进度紧张的新项目。在2025年6月25日的团队会议后，他意识到原定2025年12月15 + 日的截止日期可能无法实现，因为后端会延迟。由于担心测试时间不足，他接受了Jerry提出的延期建议，计划在次日早上的会议上提出将截止日期推迟至2026 + 年1月5日。" +} + +另一个中文示例（注意：当用户语言为中文时，您也需输出中文）： + +对话（节选）： +user｜2025-07-12 19:30：下个月同事婚礼，要买条裙子，预算1500内，别太暴露。 +assistant｜19:32 +：了解～婚礼场合建议中长款，端庄一些。可以看法式茶歇/直筒/裹身三路。颜色上，奶油白、浅粉、雾霾蓝都适合；面料可看醋酸/真丝混纺（垂感好但需干洗）。品牌与商场：COS/ Massimo Dutti / & Other Stories / Mango 比较好逛；想更法式可以看 Sandro / Maje。 我记得你很喜欢Maje？ +user｜19:35：不喜欢亮色。国贸方便。 +assistant｜19:37：那亮度就压低些：雾霾蓝、藕粉、燕麦米白更稳。国贸有 COS / Massimo Dutti / &OS；可以先看 COS 直筒或裹身，再去 MD 看藕粉/雾霾蓝。搭配上，薄披肩/小西装都能压场。 +user｜19:40：165cm，S码；最好有口袋。 +assistant｜19:42：OK。版型避开大摆+高开衩，选直筒或轻A字更稳；若要口袋，COS直筒一些款有；开衩别太高，及小腿肚附近即可。鞋包：中跟裸色鞋 + 小号手拿包。 +user｜19:47：本来想茶歇A字，但雾霾蓝直筒也行。 +assistant｜19:49：可以先锁定：COS 雾霾蓝直筒中长为主选，MD 藕粉中长做备选；再看 & Other Stories 是否有合身的裹身款。 +user｜19:52：行，周六（7/19）去国贸试，合适就买。 +assistant｜19:54：记得看肩带可调/有无内衬，醋酸/真丝优先干洗；准备同色安全裤/防走光贴。如果当天没货，可下单调货或线上下单门店自提。 + +{ + "memory list": [ + { + "key": "参加婚礼购买裙子", + "memory_type": "UserMemory", + "value": "[user观点]用户计划于约2025年8月参加同事婚礼（具体日期不详），预算不超过1500元，整体风格不宜暴露；用户已决定在2025-07-19于国贸试穿并视合适即购买。", + "tags": ["婚礼", "预算", "国贸", "计划"] + }, + { + "key": "审美与版型偏好", + "memory_type": "UserMemory", + "value": "[user观点]用户不喜欢亮色，倾向低亮度色系；裙装偏好端庄的中长款，接受直筒或轻A字。", + "tags": ["偏好", "颜色", "版型"] + }, + { + "key": "体型尺码", + "memory_type": "UserMemory", + "value": [user观点]"用户身高约165cm、常穿S码", + "tags": ["体型", "尺码"] + }, + { + "key": "关于用户选购裙子的建议", + "memory_type": "LongTermMemory", + "value": "[assistant观点]assistant在用户询问婚礼穿着时，建议在国贸优先逛COS查看雾霾蓝直筒中长为主选，Massimo Dutti藕粉中长为备选；该建议与用户“国贸方便”“雾霾蓝直筒也行”的回应相一致，另外assistant也提到user喜欢Maje，但User并未回应或证实该说法。", + "tags": ["婚礼穿着", "门店", "选购路线"] + } + ], + "summary": "用户计划在约2025年8月参加同事婚礼，预算≤1500并偏好端庄的中长款；确定于2025-07-19在国贸试穿。其长期画像显示：不喜欢亮色、偏好低亮度色系与不过分暴露的版型，身高约165cm、S码且偏好裙装带口袋。助手提出的国贸选购路线以COS雾霾蓝直筒中长为主选、MD藕粉中长为备选，且与用户回应一致，为线下试穿与购买提供了明确路径。" +} + +请始终使用与对话相同的语言进行回复。 + +对话： +${conversation} + +您的输出：""" From c389367185c4f100138b75829ca6d0638b05496c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 29 Oct 2025 21:37:17 +0800 Subject: [PATCH 03/72] fix install problem --- poetry.lock | 50 +++++++++++++++-- pyproject.toml | 4 +- src/memos/llms/openai.py | 22 +------- src/memos/mem_reader/strategy_struct.py | 53 ++++++++++--------- .../tree_text_memory/retrieve/bm25_util.py | 13 +++-- .../retrieve/retrieve_utils.py | 10 +++- .../tree_text_memory/retrieve/searcher.py | 2 +- 7 files changed, 99 insertions(+), 55 deletions(-) diff --git a/poetry.lock b/poetry.lock index 44265bca8..926d580fb 100644 --- a/poetry.lock +++ b/poetry.lock @@ -1,4 +1,4 @@ -# This file is automatically @generated by Poetry 2.2.1 and should not be changed by hand. +# This file is automatically @generated by Poetry 2.1.3 and should not be changed by hand. [[package]] name = "absl-py" @@ -192,6 +192,19 @@ torch = ">=1.0.0" tqdm = ">=4.31.1" transformers = ">=3.0.0" +[[package]] +name = "cachetools" +version = "6.2.1" +description = "Extensible memoizing collections and decorators" +optional = true +python-versions = ">=3.9" +groups = ["main"] +markers = "extra == \"all\"" +files = [ + {file = "cachetools-6.2.1-py3-none-any.whl", hash = "sha256:09868944b6dde876dfd44e1d47e18484541eaf12f26f29b7af91b26cc892d701"}, + {file = "cachetools-6.2.1.tar.gz", hash = "sha256:3f391e4bd8f8bf0931169baf7456cc822705f4e2a31f840d218f445b9a854201"}, +] + [[package]] name = "certifi" version = "2025.7.14" @@ -1553,6 +1566,18 @@ files = [ {file = "itsdangerous-2.2.0.tar.gz", hash = "sha256:e0050c0b7da1eea53ffaf149c0cfbb5c6e2e2b69c4bef22c81fa6eb73e5f6173"}, ] +[[package]] +name = "jieba" +version = "0.42" +description = "Chinese Words Segmentation Utilities" +optional = true +python-versions = "*" +groups = ["main"] +markers = "extra == \"all\"" +files = [ + {file = "jieba-0.42.tar.gz", hash = "sha256:34a3c960cc2943d9da16d6d2565110cf5f305921a67413dddf04f84de69c939b"}, +] + [[package]] name = "jinja2" version = "3.1.6" @@ -4123,6 +4148,25 @@ urllib3 = ">=1.26.14,<3" fastembed = ["fastembed (>=0.7,<0.8)"] fastembed-gpu = ["fastembed-gpu (>=0.7,<0.8)"] +[[package]] +name = "rank-bm25" +version = "0.2.2" +description = "Various BM25 algorithms for document ranking" +optional = true +python-versions = "*" +groups = ["main"] +markers = "extra == \"all\"" +files = [ + {file = "rank_bm25-0.2.2-py3-none-any.whl", hash = "sha256:7bd4a95571adadfc271746fa146a4bcfd89c0cf731e49c3d1ad863290adbe8ae"}, + {file = "rank_bm25-0.2.2.tar.gz", hash = "sha256:096ccef76f8188563419aaf384a02f0ea459503fdf77901378d4fd9d87e5e51d"}, +] + +[package.dependencies] +numpy = "*" + +[package.extras] +dev = ["pytest"] + [[package]] name = "redis" version = "6.2.0" @@ -6352,7 +6396,7 @@ cffi = {version = ">=1.11", markers = "platform_python_implementation == \"PyPy\ cffi = ["cffi (>=1.11)"] [extras] -all = ["chonkie", "datasketch", "markitdown", "neo4j", "pika", "pymilvus", "pymysql", "qdrant-client", "redis", "schedule", "sentence-transformers", "torch", "volcengine-python-sdk"] +all = ["cachetools", "chonkie", "datasketch", "jieba", "markitdown", "neo4j", "pika", "pymilvus", "pymysql", "qdrant-client", "rank-bm25", "redis", "schedule", "sentence-transformers", "torch", "volcengine-python-sdk"] mem-reader = ["chonkie", "markitdown"] mem-scheduler = ["pika", "redis"] mem-user = ["pymysql"] @@ -6362,4 +6406,4 @@ tree-mem = ["neo4j", "schedule"] [metadata] lock-version = "2.1" python-versions = ">=3.10,<4.0" -content-hash = "3f0d0c9a996f87d945ef8bf83eed3e20f8c420b6b39e12012d0147eda2bf4d38" +content-hash = "ec17679a44205ada4494fbc485ac592883281fde273d5e73d6b8cbc6f7f9ed10" diff --git a/pyproject.toml b/pyproject.toml index 3745582f6..2f88797a8 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -107,7 +107,9 @@ all = [ "markitdown[docx,pdf,pptx,xls,xlsx] (>=0.1.1,<0.2.0)", "pymilvus (>=2.6.1,<3.0.0)", "datasketch (>=1.6.5,<2.0.0)", - + "jieba (>=0.38.1,<0.42.1)", + "rank-bm25 (>=0.2.2)", + "cachetools (>=6.0.0)", # NOT exist in the above optional groups # Because they are either huge-size dependencies or infrequently used dependencies. # We kindof don't want users to install them. diff --git a/src/memos/llms/openai.py b/src/memos/llms/openai.py index 289d2aea1..1a1703340 100644 --- a/src/memos/llms/openai.py +++ b/src/memos/llms/openai.py @@ -58,29 +58,11 @@ def clear_cache(cls): logger.info("OpenAI LLM instance cache cleared") @timed(log=True, log_prefix="OpenAI LLM") - def generate(self, messages: MessageList) -> str: - """Generate a response from OpenAI LLM.""" - response = self.client.chat.completions.create( - model=self.config.model_name_or_path, - messages=messages, - extra_body=self.config.extra_body, - temperature=self.config.temperature, - max_tokens=self.config.max_tokens, - top_p=self.config.top_p, - ) - logger.info(f"Response from OpenAI: {response.model_dump_json()}") - response_content = response.choices[0].message.content - if self.config.remove_think_prefix: - return remove_thinking_tags(response_content) - else: - return response_content - - def customized_generate(self, messages: MessageList, **kwargs) -> str: - """Generate a response from OpenAI LLM.""" + def generate(self, messages: MessageList, **kwargs) -> str: + """Generate a response from OpenAI LLM, optionally overriding generation params.""" temperature = kwargs.get("temperature", self.config.temperature) max_tokens = kwargs.get("max_tokens", self.config.max_tokens) top_p = kwargs.get("top_p", self.config.top_p) - response = self.client.chat.completions.create( model=self.config.model_name_or_path, messages=messages, diff --git a/src/memos/mem_reader/strategy_struct.py b/src/memos/mem_reader/strategy_struct.py index c7e9a9d1b..2cac1652a 100644 --- a/src/memos/mem_reader/strategy_struct.py +++ b/src/memos/mem_reader/strategy_struct.py @@ -1,14 +1,11 @@ import os -import re from abc import ABC from memos import log from memos.configs.mem_reader import StrategyStructMemReaderConfig from memos.configs.parser import ParserConfigFactory -from memos.mem_reader.simple_struct import ( - SimpleStructMemReader, -) +from memos.mem_reader.simple_struct import SimpleStructMemReader, detect_lang from memos.parsers.factory import ParserFactory from memos.templates.mem_reader_prompts import ( SIMPLE_STRUCT_DOC_READER_PROMPT, @@ -23,7 +20,7 @@ logger = log.get_logger(__name__) -PROMPT_DICT = { +STRATEGY_PROMPT_DICT = { "chat": { "en": STRATEGY_STRUCT_MEM_READER_PROMPT, "zh": STRATEGY_STRUCT_MEM_READER_PROMPT_ZH, @@ -33,26 +30,6 @@ "doc": {"en": SIMPLE_STRUCT_DOC_READER_PROMPT, "zh": SIMPLE_STRUCT_DOC_READER_PROMPT_ZH}, } -try: - import tiktoken - - try: - _ENC = tiktoken.encoding_for_model("gpt-4o-mini") - except Exception: - _ENC = tiktoken.get_encoding("cl100k_base") - - def _count_tokens_text(s: str) -> int: - return len(_ENC.encode(s or "")) -except Exception: - # Heuristic fallback: zh chars ~1 token, others ~1 token per ~4 chars - def _count_tokens_text(s: str) -> int: - if not s: - return 0 - zh_chars = re.findall(r"[\u4e00-\u9fff]", s) - zh = len(zh_chars) - rest = len(s) - zh - return zh + max(1, rest // 4) - class StrategyStructMemReader(SimpleStructMemReader, ABC): """Naive implementation of MemReader.""" @@ -61,6 +38,32 @@ def __init__(self, config: StrategyStructMemReaderConfig): super().__init__(config) self.chat_chunker = config.chat_chunker["config"] + def _get_llm_response(self, mem_str: str) -> dict: + lang = detect_lang(mem_str) + template = STRATEGY_PROMPT_DICT["chat"][lang] + examples = STRATEGY_PROMPT_DICT["chat"][f"{lang}_example"] + prompt = template.replace("${conversation}", mem_str) + if self.config.remove_prompt_example: + prompt = prompt.replace(examples, "") + messages = [{"role": "user", "content": prompt}] + try: + response_text = self.llm.generate(messages) + response_json = self.parse_json_result(response_text) + except Exception as e: + logger.error(f"[LLM] Exception during chat generation: {e}") + response_json = { + "memory list": [ + { + "key": mem_str[:10], + "memory_type": "UserMemory", + "value": mem_str, + "tags": [], + } + ], + "summary": mem_str, + } + return response_json + def get_scene_data_info(self, scene_data: list, type: str) -> list[str]: """ Get raw information from scene_data. diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/bm25_util.py b/src/memos/memories/textual/tree_text_memory/retrieve/bm25_util.py index 9f02daae4..4aca4022f 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/bm25_util.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/bm25_util.py @@ -2,10 +2,9 @@ import numpy as np -from cachetools import LRUCache -from rank_bm25 import BM25Okapi from sklearn.feature_extraction.text import TfidfVectorizer +from memos.dependency import require_python_package from memos.log import get_logger from memos.memories.textual.tree_text_memory.retrieve.retrieve_utils import FastTokenizer from memos.utils import timed @@ -13,13 +12,13 @@ logger = get_logger(__name__) # Global model cache -_BM25_CACHE = LRUCache(maxsize=100) _CACHE_LOCK = threading.Lock() class EnhancedBM25: """Enhanced BM25 with Spacy tokenization and TF-IDF reranking""" + @require_python_package(import_name="cachetools", install_command="pip install cachetools") def __init__(self, tokenizer=None, en_model="en_core_web_sm", zh_model="zh_core_web_sm"): """ Initialize Enhanced BM25 with memory management @@ -30,13 +29,21 @@ def __init__(self, tokenizer=None, en_model="en_core_web_sm", zh_model="zh_core_ self.tokenizer = tokenizer self._current_tfidf = None + global _BM25_CACHE + from cachetools import LRUCache + + _BM25_CACHE = LRUCache(maxsize=100) + def _tokenize_doc(self, text): """ Tokenize a single document using SpacyTokenizer """ return self.tokenizer.tokenize_mixed(text, lang="auto") + @require_python_package(import_name="rank_bm25", install_command="pip install rank_bm25") def _prepare_corpus_data(self, corpus, corpus_name="default"): + from rank_bm25 import BM25Okapi + with _CACHE_LOCK: if corpus_name in _BM25_CACHE: print("hit::", corpus_name) diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py b/src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py index e3081ebf6..eec827c86 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py @@ -3,8 +3,7 @@ from pathlib import Path -import jieba - +from memos.dependency import require_python_package from memos.log import get_logger @@ -326,8 +325,15 @@ def _is_chinese(self, text): chinese_chars = sum(1 for char in text if "\u4e00" <= char <= "\u9fff") return chinese_chars / max(len(text), 1) > 0.3 + @require_python_package( + import_name="jieba", + install_command="pip install jieba", + install_link="https://github.com/fxsjy/jieba", + ) def _tokenize_chinese(self, text): """split zh jieba""" + import jieba + tokens = jieba.lcut(text) if self.use_jieba else list(text) tokens = [token.strip() for token in tokens if token.strip()] if self.use_stopwords: diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py b/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py index c891e9f0c..b6d082940 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py @@ -522,7 +522,7 @@ def _cot_query( messages = [{"role": "user", "content": prompt}] try: - response_text = self.llm.customized_generate(messages, temperature=0, top_p=1) + response_text = self.llm.generate(messages, temperature=0, top_p=1) response_json = parse_json_result(response_text) assert "is_complex" in response_json if not response_json["is_complex"]: From 499502d5ea3831bc79208d2cae7c9db8fccb51e2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 29 Oct 2025 22:09:42 +0800 Subject: [PATCH 04/72] fix --- src/memos/configs/mem_reader.py | 2 +- src/memos/memories/textual/tree_text_memory/retrieve/recall.py | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/src/memos/configs/mem_reader.py b/src/memos/configs/mem_reader.py index fb2de3598..dc8d37a35 100644 --- a/src/memos/configs/mem_reader.py +++ b/src/memos/configs/mem_reader.py @@ -37,7 +37,7 @@ def parse_datetime(cls, value): ) chat_chunker: dict[str, Any] = Field( - ..., description="Configuration for the MemReader chat chunk strategy" + default=None, description="Configuration for the MemReader chat chunk strategy" ) diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/recall.py b/src/memos/memories/textual/tree_text_memory/retrieve/recall.py index 93edf3d47..b7383aa13 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/recall.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/recall.py @@ -21,7 +21,7 @@ def __init__( self, graph_store: Neo4jGraphDB, embedder: OllamaEmbedder, - bm25_retriever: EnhancedBM25 | None, + bm25_retriever: EnhancedBM25 | None = None, ): self.graph_store = graph_store self.embedder = embedder From e1bb223b6107d3873f593560360160f974c5c9c0 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 29 Oct 2025 22:20:10 +0800 Subject: [PATCH 05/72] fix test --- tests/memories/textual/test_tree_task_goal_parser.py | 5 ----- 1 file changed, 5 deletions(-) diff --git a/tests/memories/textual/test_tree_task_goal_parser.py b/tests/memories/textual/test_tree_task_goal_parser.py index c71af4b06..899e2454b 100644 --- a/tests/memories/textual/test_tree_task_goal_parser.py +++ b/tests/memories/textual/test_tree_task_goal_parser.py @@ -20,12 +20,7 @@ def generate(self, messages): def test_parse_fast_returns_expected(): parser = TaskGoalParser() result = parser.parse("Tell me about cats", mode="fast") - assert isinstance(result, ParsedTaskGoal) - assert result.memories == ["Tell me about cats"] - assert result.keys == ["Tell me about cats"] - assert result.tags == [] - assert result.goal_type == "default" def test_parse_fine_calls_llm_and_parses(): From 790e99f54a465913346b03cc9d97a309b7921b69 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Thu, 30 Oct 2025 15:19:01 +0800 Subject: [PATCH 06/72] turn off graph recall --- .../tree_text_memory/retrieve/recall.py | 124 ++++++++++++------ .../retrieve/task_goal_parser.py | 29 ++-- 2 files changed, 107 insertions(+), 46 deletions(-) diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/recall.py b/src/memos/memories/textual/tree_text_memory/retrieve/recall.py index b7383aa13..5dac95c28 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/recall.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/recall.py @@ -1,4 +1,5 @@ import concurrent.futures +import os from memos.context.context import ContextThreadPoolExecutor from memos.embedders.factory import OllamaEmbedder @@ -184,47 +185,96 @@ def process_node(node): return TextualMemoryItem.from_dict(node) return None - candidate_ids = set() - - # 1) key-based OR branch - if parsed_goal.keys: - key_filters = [ - {"field": "key", "op": "in", "value": parsed_goal.keys}, - {"field": "memory_type", "op": "=", "value": memory_scope}, - ] - key_ids = self.graph_store.get_by_metadata(key_filters, user_name=user_name) - candidate_ids.update(key_ids) - - # 2) tag-based OR branch - if parsed_goal.tags: - tag_filters = [ - {"field": "tags", "op": "contains", "value": parsed_goal.tags}, - {"field": "memory_type", "op": "=", "value": memory_scope}, - ] - tag_ids = self.graph_store.get_by_metadata(tag_filters, user_name=user_name) - candidate_ids.update(tag_ids) - - # No matches → return empty - if not candidate_ids: - return [] + if os.getenv("FAST_GRAPH", "false") == "true": + candidate_ids = set() - # Load nodes and post-filter - node_dicts = self.graph_store.get_nodes( - list(candidate_ids), include_embedding=False, user_name=user_name - ) + # 1) key-based OR branch + if parsed_goal.keys: + key_filters = [ + {"field": "key", "op": "in", "value": parsed_goal.keys}, + {"field": "memory_type", "op": "=", "value": memory_scope}, + ] + key_ids = self.graph_store.get_by_metadata(key_filters) + candidate_ids.update(key_ids) + + # 2) tag-based OR branch + if parsed_goal.tags: + tag_filters = [ + {"field": "tags", "op": "contains", "value": parsed_goal.tags}, + {"field": "memory_type", "op": "=", "value": memory_scope}, + ] + tag_ids = self.graph_store.get_by_metadata(tag_filters) + candidate_ids.update(tag_ids) - final_nodes = [] - with ContextThreadPoolExecutor(max_workers=3) as executor: - futures = {executor.submit(process_node, node): i for i, node in enumerate(node_dicts)} - temp_results = [None] * len(node_dicts) + # No matches → return empty + if not candidate_ids: + return [] + + # Load nodes and post-filter + node_dicts = self.graph_store.get_nodes(list(candidate_ids), include_embedding=False) + + final_nodes = [] + for node in node_dicts: + meta = node.get("metadata", {}) + node_key = meta.get("key") + node_tags = meta.get("tags", []) or [] + + keep = False + # key equals to node_key + if parsed_goal.keys and node_key in parsed_goal.keys: + keep = True + # overlap tags more than 2 + elif parsed_goal.tags: + overlap = len(set(node_tags) & set(parsed_goal.tags)) + if overlap >= 2: + keep = True + if keep: + final_nodes.append(TextualMemoryItem.from_dict(node)) + return final_nodes + else: + candidate_ids = set() + + # 1) key-based OR branch + if parsed_goal.keys: + key_filters = [ + {"field": "key", "op": "in", "value": parsed_goal.keys}, + {"field": "memory_type", "op": "=", "value": memory_scope}, + ] + key_ids = self.graph_store.get_by_metadata(key_filters, user_name=user_name) + candidate_ids.update(key_ids) + + # 2) tag-based OR branch + if parsed_goal.tags: + tag_filters = [ + {"field": "tags", "op": "contains", "value": parsed_goal.tags}, + {"field": "memory_type", "op": "=", "value": memory_scope}, + ] + tag_ids = self.graph_store.get_by_metadata(tag_filters, user_name=user_name) + candidate_ids.update(tag_ids) + + # No matches → return empty + if not candidate_ids: + return [] + + # Load nodes and post-filter + node_dicts = self.graph_store.get_nodes( + list(candidate_ids), include_embedding=False, user_name=user_name + ) + + final_nodes = [] + with ContextThreadPoolExecutor(max_workers=3) as executor: + futures = { + executor.submit(process_node, node): i for i, node in enumerate(node_dicts) + } + temp_results = [None] * len(node_dicts) - for future in concurrent.futures.as_completed(futures): - original_index = futures[future] - result = future.result() - temp_results[original_index] = result + for future in concurrent.futures.as_completed(futures): + original_index = futures[future] + result = future.result() + temp_results[original_index] = result - final_nodes = [result for result in temp_results if result is not None] - return final_nodes + final_nodes = [result for result in temp_results if result is not None] + return final_nodes def _vector_recall( self, diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/task_goal_parser.py b/src/memos/memories/textual/tree_text_memory/retrieve/task_goal_parser.py index 6a1138c90..6e256611a 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/task_goal_parser.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/task_goal_parser.py @@ -1,3 +1,4 @@ +import os import traceback from string import Template @@ -50,15 +51,25 @@ def _parse_fast(self, task_description: str, limit_num: int = 5) -> ParsedTaskGo """ Fast mode: simple jieba word split. """ - desc_tokenized = self.tokenizer.tokenize_mixed(task_description) - return ParsedTaskGoal( - memories=[task_description], - keys=desc_tokenized, - tags=desc_tokenized, - goal_type="default", - rephrased_query=task_description, - internet_search=False, - ) + if os.getenv("FAST_GRAPH", "false") == "true": + desc_tokenized = self.tokenizer.tokenize_mixed(task_description) + return ParsedTaskGoal( + memories=[task_description], + keys=desc_tokenized, + tags=desc_tokenized, + goal_type="default", + rephrased_query=task_description, + internet_search=False, + ) + else: + return ParsedTaskGoal( + memories=[task_description], + keys=[task_description], + tags=[], + goal_type="default", + rephrased_query=task_description, + internet_search=False, + ) def _parse_fine( self, query: str, context: str = "", conversation: list[dict] | None = None From 390ba298c0b8b51e781a7349bc3ea3dec96b5c79 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Thu, 30 Oct 2025 16:06:14 +0800 Subject: [PATCH 07/72] turn off graph recall --- src/memos/api/config.py | 2 ++ src/memos/memories/textual/simple_tree.py | 10 ++-------- src/memos/memories/textual/tree.py | 7 ++----- .../textual/tree_text_memory/retrieve/recall.py | 15 +++++++++++---- .../textual/tree_text_memory/retrieve/searcher.py | 12 +++++++++--- .../tree_text_memory/retrieve/task_goal_parser.py | 9 +++++---- 6 files changed, 31 insertions(+), 24 deletions(-) diff --git a/src/memos/api/config.py b/src/memos/api/config.py index 405e8068d..c7d7ae316 100644 --- a/src/memos/api/config.py +++ b/src/memos/api/config.py @@ -866,6 +866,7 @@ def create_user_config(user_name: str, user_id: str) -> tuple[MOSConfig, General "UserMemory": os.getenv("NEBULAR_USER_MEMORY", 1e6), }, "search_strategy": { + "fast_graph": bool(os.getenv("FAST_GRAPH", "false") == "true"), "bm25": bool(os.getenv("BM25_CALL", "false") == "true"), "cot": bool(os.getenv("VEC_COT_CALL", "false") == "true"), }, @@ -937,6 +938,7 @@ def get_default_cube_config() -> GeneralMemCubeConfig | None: "UserMemory": os.getenv("NEBULAR_USER_MEMORY", 1e6), }, "search_strategy": { + "fast_graph": bool(os.getenv("FAST_GRAPH", "false") == "true"), "bm25": bool(os.getenv("BM25_CALL", "false") == "true"), "cot": bool(os.getenv("VEC_COT_CALL", "false") == "true"), }, diff --git a/src/memos/memories/textual/simple_tree.py b/src/memos/memories/textual/simple_tree.py index 6974dbe8f..992b7bfab 100644 --- a/src/memos/memories/textual/simple_tree.py +++ b/src/memos/memories/textual/simple_tree.py @@ -70,12 +70,6 @@ def __init__( ) logger.info(f"time init: bm25_retriever time is: {time.time() - time_start_bm}") - self.vec_cot = ( - self.search_strategy["cot"] - if self.search_strategy and "cot" in self.search_strategy - else False - ) - time_start_rr = time.time() self.reranker = reranker logger.info(f"time init: reranker time is: {time.time() - time_start_rr}") @@ -189,7 +183,7 @@ def search( bm25_retriever=self.bm25_retriever, internet_retriever=None, moscube=moscube, - vec_cot=self.vec_cot, + search_strategy=self.search_strategy, ) else: searcher = Searcher( @@ -200,7 +194,7 @@ def search( bm25_retriever=self.bm25_retriever, internet_retriever=self.internet_retriever, moscube=moscube, - vec_cot=self.vec_cot, + search_strategy=self.search_strategy, ) return searcher.search( query, top_k, info, mode, memory_type, search_filter, user_name=user_name diff --git a/src/memos/memories/textual/tree.py b/src/memos/memories/textual/tree.py index a58f993bb..19bd3ba5b 100644 --- a/src/memos/memories/textual/tree.py +++ b/src/memos/memories/textual/tree.py @@ -51,11 +51,6 @@ def __init__(self, config: TreeTextMemoryConfig): self.bm25_retriever = ( EnhancedBM25() if self.search_strategy and self.search_strategy["bm25"] else None ) - self.vec_cot = ( - self.search_strategy["cot"] - if self.search_strategy and "cot" in self.search_strategy - else False - ) if config.reranker is None: default_cfg = RerankerConfigFactory.model_validate( @@ -143,6 +138,7 @@ def get_searcher( self.reranker, internet_retriever=None, moscube=moscube, + search_strategy=self.search_strategy, ) else: searcher = Searcher( @@ -152,6 +148,7 @@ def get_searcher( self.reranker, internet_retriever=self.internet_retriever, moscube=moscube, + search_strategy=self.search_strategy, ) return searcher diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/recall.py b/src/memos/memories/textual/tree_text_memory/retrieve/recall.py index 5dac95c28..8cf2f47f3 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/recall.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/recall.py @@ -1,5 +1,4 @@ import concurrent.futures -import os from memos.context.context import ContextThreadPoolExecutor from memos.embedders.factory import OllamaEmbedder @@ -41,6 +40,7 @@ def retrieve( search_filter: dict | None = None, user_name: str | None = None, id_filter: dict | None = None, + use_fast_graph: bool = False, ) -> list[TextualMemoryItem]: """ Perform hybrid memory retrieval: @@ -70,7 +70,13 @@ def retrieve( with ContextThreadPoolExecutor(max_workers=3) as executor: # Structured graph-based retrieval - future_graph = executor.submit(self._graph_recall, parsed_goal, memory_scope, user_name) + future_graph = executor.submit( + self._graph_recall, + parsed_goal, + memory_scope, + user_name, + use_fast_graph=use_fast_graph, + ) # Vector similarity search future_vector = executor.submit( self._vector_recall, @@ -156,7 +162,7 @@ def retrieve_from_cube( return list(combined.values()) def _graph_recall( - self, parsed_goal: ParsedTaskGoal, memory_scope: str, user_name: str | None = None + self, parsed_goal: ParsedTaskGoal, memory_scope: str, user_name: str | None = None, **kwargs ) -> list[TextualMemoryItem]: """ Perform structured node-based retrieval from Neo4j. @@ -164,6 +170,7 @@ def _graph_recall( - tags must overlap with at least 2 input tags - scope filters by memory_type if provided """ + use_fast_graph = kwargs.get("use_fast_graph", False) def process_node(node): meta = node.get("metadata", {}) @@ -185,7 +192,7 @@ def process_node(node): return TextualMemoryItem.from_dict(node) return None - if os.getenv("FAST_GRAPH", "false") == "true": + if not use_fast_graph: candidate_ids = set() # 1) key-based OR branch diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py b/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py index 563695c68..b8e90cbf1 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py @@ -45,7 +45,7 @@ def __init__( bm25_retriever: EnhancedBM25 | None = None, internet_retriever: None = None, moscube: bool = False, - vec_cot: bool = False, + search_strategy: dict | None = None, ): self.graph_store = graph_store self.embedder = embedder @@ -59,7 +59,9 @@ def __init__( # Create internet retriever from config if provided self.internet_retriever = internet_retriever self.moscube = moscube - self.vec_cot = vec_cot + self.use_fast_graph = ( + search_strategy.get("fast_graph", "false") == "true" if search_strategy else False + ) self._usage_executor = ContextThreadPoolExecutor(max_workers=4, thread_name_prefix="usage") @@ -226,6 +228,7 @@ def _parse_task( context="\n".join(context), conversation=info.get("chat_history", []), mode=mode, + use_fast_graph=self.use_fast_graph, ) query = parsed_goal.rephrased_query or query @@ -340,6 +343,7 @@ def _retrieve_from_working_memory( search_filter=search_filter, user_name=user_name, id_filter=id_filter, + use_fast_graph=self.use_fast_graph, ) return self.reranker.rerank( query=query, @@ -369,7 +373,7 @@ def _retrieve_from_long_term_and_user( # chain of thinking cot_embeddings = [] - if self.vec_cot: + if self.search_strategy["vec_cot"]: queries = self._cot_query(query) if len(queries) > 1: cot_embeddings = self.embedder.embed(queries) @@ -390,6 +394,7 @@ def _retrieve_from_long_term_and_user( search_filter=search_filter, user_name=user_name, id_filter=id_filter, + use_fast_graph=self.use_fast_graph, ) ) if memory_type in ["All", "UserMemory"]: @@ -404,6 +409,7 @@ def _retrieve_from_long_term_and_user( search_filter=search_filter, user_name=user_name, id_filter=id_filter, + use_fast_graph=self.use_fast_graph, ) ) diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/task_goal_parser.py b/src/memos/memories/textual/tree_text_memory/retrieve/task_goal_parser.py index 6e256611a..5d706559c 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/task_goal_parser.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/task_goal_parser.py @@ -1,4 +1,3 @@ -import os import traceback from string import Template @@ -30,6 +29,7 @@ def parse( context: str = "", conversation: list[dict] | None = None, mode: str = "fast", + **kwargs, ) -> ParsedTaskGoal: """ Parse user input into structured semantic layers. @@ -39,7 +39,7 @@ def parse( - mode == 'fine': use LLM to parse structured topic/keys/tags """ if mode == "fast": - return self._parse_fast(task_description) + return self._parse_fast(task_description, **kwargs) elif mode == "fine": if not self.llm: raise ValueError("LLM not provided for slow mode.") @@ -47,11 +47,12 @@ def parse( else: raise ValueError(f"Unknown mode: {mode}") - def _parse_fast(self, task_description: str, limit_num: int = 5) -> ParsedTaskGoal: + def _parse_fast(self, task_description: str, **kwargs) -> ParsedTaskGoal: """ Fast mode: simple jieba word split. """ - if os.getenv("FAST_GRAPH", "false") == "true": + use_fast_graph = kwargs.get("use_fast_graph", False) + if use_fast_graph: desc_tokenized = self.tokenizer.tokenize_mixed(task_description) return ParsedTaskGoal( memories=[task_description], From 96152821722652ce38813d2a85c001f183cb47df Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Thu, 30 Oct 2025 16:13:47 +0800 Subject: [PATCH 08/72] turn off graph recall --- .../memories/textual/tree_text_memory/retrieve/searcher.py | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py b/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py index b8e90cbf1..0974d67f2 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py @@ -59,6 +59,9 @@ def __init__( # Create internet retriever from config if provided self.internet_retriever = internet_retriever self.moscube = moscube + self.vec_cot = ( + search_strategy.get("vec_cot", "false") == "true" if search_strategy else False + ) self.use_fast_graph = ( search_strategy.get("fast_graph", "false") == "true" if search_strategy else False ) @@ -373,7 +376,7 @@ def _retrieve_from_long_term_and_user( # chain of thinking cot_embeddings = [] - if self.search_strategy["vec_cot"]: + if self.vec_cot: queries = self._cot_query(query) if len(queries) > 1: cot_embeddings = self.embedder.embed(queries) From 04f412b2e28faec1da603115f8938918cabd2155 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Thu, 30 Oct 2025 19:11:03 +0800 Subject: [PATCH 09/72] fix Searcher input bug --- src/memos/memories/textual/tree.py | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) diff --git a/src/memos/memories/textual/tree.py b/src/memos/memories/textual/tree.py index 19bd3ba5b..53628d075 100644 --- a/src/memos/memories/textual/tree.py +++ b/src/memos/memories/textual/tree.py @@ -138,7 +138,6 @@ def get_searcher( self.reranker, internet_retriever=None, moscube=moscube, - search_strategy=self.search_strategy, ) else: searcher = Searcher( @@ -148,7 +147,6 @@ def get_searcher( self.reranker, internet_retriever=self.internet_retriever, moscube=moscube, - search_strategy=self.search_strategy, ) return searcher @@ -197,7 +195,7 @@ def search( bm25_retriever=self.bm25_retriever, internet_retriever=None, moscube=moscube, - vec_cot=self.vec_cot, + search_strategy=self.search_strategy, ) else: searcher = Searcher( @@ -208,7 +206,7 @@ def search( bm25_retriever=self.bm25_retriever, internet_retriever=self.internet_retriever, moscube=moscube, - vec_cot=self.vec_cot, + search_strategy=self.search_strategy, ) return searcher.search(query, top_k, info, mode, memory_type, search_filter) From 9716274a5b3eb51fc0f7b2566fdd871ad1dd042c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Thu, 30 Oct 2025 19:13:24 +0800 Subject: [PATCH 10/72] fix Searcher --- src/memos/memories/textual/tree.py | 2 -- 1 file changed, 2 deletions(-) diff --git a/src/memos/memories/textual/tree.py b/src/memos/memories/textual/tree.py index 19bd3ba5b..9b46d776b 100644 --- a/src/memos/memories/textual/tree.py +++ b/src/memos/memories/textual/tree.py @@ -197,7 +197,6 @@ def search( bm25_retriever=self.bm25_retriever, internet_retriever=None, moscube=moscube, - vec_cot=self.vec_cot, ) else: searcher = Searcher( @@ -208,7 +207,6 @@ def search( bm25_retriever=self.bm25_retriever, internet_retriever=self.internet_retriever, moscube=moscube, - vec_cot=self.vec_cot, ) return searcher.search(query, top_k, info, mode, memory_type, search_filter) From f8b9b4a37c655e9006e8827480c91166e9a81caf Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Thu, 30 Oct 2025 19:16:27 +0800 Subject: [PATCH 11/72] fix Search --- src/memos/memories/textual/tree.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/src/memos/memories/textual/tree.py b/src/memos/memories/textual/tree.py index 9b46d776b..53628d075 100644 --- a/src/memos/memories/textual/tree.py +++ b/src/memos/memories/textual/tree.py @@ -138,7 +138,6 @@ def get_searcher( self.reranker, internet_retriever=None, moscube=moscube, - search_strategy=self.search_strategy, ) else: searcher = Searcher( @@ -148,7 +147,6 @@ def get_searcher( self.reranker, internet_retriever=self.internet_retriever, moscube=moscube, - search_strategy=self.search_strategy, ) return searcher @@ -197,6 +195,7 @@ def search( bm25_retriever=self.bm25_retriever, internet_retriever=None, moscube=moscube, + search_strategy=self.search_strategy, ) else: searcher = Searcher( @@ -207,6 +206,7 @@ def search( bm25_retriever=self.bm25_retriever, internet_retriever=self.internet_retriever, moscube=moscube, + search_strategy=self.search_strategy, ) return searcher.search(query, top_k, info, mode, memory_type, search_filter) From b9dbecd5003688288e84f491b9d4f4130f61553b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 4 Nov 2025 11:46:18 +0800 Subject: [PATCH 12/72] fix bug --- src/memos/api/config.py | 2 +- src/memos/configs/memory.py | 2 +- src/memos/mem_reader/strategy_struct.py | 16 ++- .../retrieve/retrieval_mid_structs.py | 1 + .../retrieve/retrieve_utils.py | 2 +- .../tree_text_memory/retrieve/searcher.py | 18 +-- .../retrieve/task_goal_parser.py | 13 ++- .../templates/mem_reader_strategy_prompts.py | 104 +++++++++--------- 8 files changed, 87 insertions(+), 71 deletions(-) diff --git a/src/memos/api/config.py b/src/memos/api/config.py index d9db93c1a..7f61d54ac 100644 --- a/src/memos/api/config.py +++ b/src/memos/api/config.py @@ -427,7 +427,7 @@ def get_reader_config() -> dict[str, Any]: "config": { "chunk_type": os.getenv("MEM_READER_CHAT_CHUNK_TYPE", "default"), "chunk_length": int(os.getenv("MEM_READER_CHAT_CHUNK_TOKEN_SIZE", 1600)), - "chunk_session": int(os.getenv("MEM_READER_CHAT_CHUNK_SESS_SIZE", 20)), + "chunk_session": int(os.getenv("MEM_READER_CHAT_CHUNK_SESS_SIZE", 10)), "chunk_overlap": int(os.getenv("MEM_READER_CHAT_CHUNK_OVERLAP", 2)), }, } diff --git a/src/memos/configs/memory.py b/src/memos/configs/memory.py index 49320fbf5..34967849a 100644 --- a/src/memos/configs/memory.py +++ b/src/memos/configs/memory.py @@ -184,7 +184,7 @@ class TreeTextMemoryConfig(BaseTextMemoryConfig): ), ) - search_strategy: dict[str, bool] | None = Field( + search_strategy: dict[str, Any] | None = Field( default=None, description=( 'Set search strategy for this memory configuration.{"bm25": true, "cot": false}' diff --git a/src/memos/mem_reader/strategy_struct.py b/src/memos/mem_reader/strategy_struct.py index 2cac1652a..a45586bd0 100644 --- a/src/memos/mem_reader/strategy_struct.py +++ b/src/memos/mem_reader/strategy_struct.py @@ -39,11 +39,12 @@ def __init__(self, config: StrategyStructMemReaderConfig): self.chat_chunker = config.chat_chunker["config"] def _get_llm_response(self, mem_str: str) -> dict: + print("[code stamp]: fine_get_llm_response") lang = detect_lang(mem_str) template = STRATEGY_PROMPT_DICT["chat"][lang] examples = STRATEGY_PROMPT_DICT["chat"][f"{lang}_example"] prompt = template.replace("${conversation}", mem_str) - if self.config.remove_prompt_example: + if self.config.remove_prompt_example: # TODO unused prompt = prompt.replace(examples, "") messages = [{"role": "user", "content": prompt}] try: @@ -112,6 +113,19 @@ def get_scene_data_info(self, scene_data: list, type: str) -> list[str]: results.append([overlap_item, item]) current_length = overlap_length + content_length + else: + cut_size, cut_overlap = ( + self.chat_chunker["chunk_session"], + self.chat_chunker["chunk_overlap"], + ) + for items in scene_data: + step = cut_size - cut_overlap + end = len(items) - cut_overlap + if end <= 0: + results.extend([items[:]]) + else: + results.extend([items[i : i + cut_size] for i in range(0, end, step)]) + elif type == "doc": parser_config = ParserConfigFactory.model_validate( { diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/retrieval_mid_structs.py b/src/memos/memories/textual/tree_text_memory/retrieve/retrieval_mid_structs.py index 6accc4a16..7aefaa1a3 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/retrieval_mid_structs.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/retrieval_mid_structs.py @@ -13,3 +13,4 @@ class ParsedTaskGoal: rephrased_query: str | None = None internet_search: bool = False goal_type: str | None = None # e.g., 'default', 'explanation', etc. + context: str = "" diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py b/src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py index eec827c86..3f2b41a47 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py @@ -17,7 +17,7 @@ def find_project_root(marker=".git"): if (current / marker).exists(): return current current = current.parent - logger.warn(f"The project root directory tag file was not found: {marker}") + return Path(".") PROJECT_ROOT = find_project_root() diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py b/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py index 0974d67f2..a86207e66 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py @@ -30,8 +30,8 @@ logger = get_logger(__name__) COT_DICT = { - "fast": {"en": COT_PROMPT, "zh": COT_PROMPT_ZH}, - "fine": {"en": SIMPLE_COT_PROMPT, "zh": SIMPLE_COT_PROMPT_ZH}, + "fine": {"en": COT_PROMPT, "zh": COT_PROMPT_ZH}, + "fast": {"en": SIMPLE_COT_PROMPT, "zh": SIMPLE_COT_PROMPT_ZH}, } @@ -59,12 +59,8 @@ def __init__( # Create internet retriever from config if provided self.internet_retriever = internet_retriever self.moscube = moscube - self.vec_cot = ( - search_strategy.get("vec_cot", "false") == "true" if search_strategy else False - ) - self.use_fast_graph = ( - search_strategy.get("fast_graph", "false") == "true" if search_strategy else False - ) + self.vec_cot = search_strategy.get("cot", False) if search_strategy else False + self.use_fast_graph = search_strategy.get("fast_graph", False) if search_strategy else False self._usage_executor = ContextThreadPoolExecutor(max_workers=4, thread_name_prefix="usage") @@ -287,6 +283,7 @@ def _retrieve_paths( search_filter, user_name, id_filter, + mode=mode, ) ) tasks.append( @@ -369,6 +366,7 @@ def _retrieve_from_long_term_and_user( search_filter: dict | None = None, user_name: str | None = None, id_filter: dict | None = None, + mode: str = "fast", ): """Retrieve and rerank from LongTermMemory and UserMemory""" results = [] @@ -377,7 +375,7 @@ def _retrieve_from_long_term_and_user( # chain of thinking cot_embeddings = [] if self.vec_cot: - queries = self._cot_query(query) + queries = self._cot_query(query, mode=mode, context=parsed_goal.context) if len(queries) > 1: cot_embeddings = self.embedder.embed(queries) cot_embeddings.extend(query_embedding) @@ -567,6 +565,8 @@ def _cot_query( "${split_num_threshold}", str(split_num) ) logger.info("COT process") + print("---------------prompt-------------") + print(prompt, mode) messages = [{"role": "user", "content": prompt}] try: diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/task_goal_parser.py b/src/memos/memories/textual/tree_text_memory/retrieve/task_goal_parser.py index 5d706559c..55e33494c 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/task_goal_parser.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/task_goal_parser.py @@ -39,7 +39,7 @@ def parse( - mode == 'fine': use LLM to parse structured topic/keys/tags """ if mode == "fast": - return self._parse_fast(task_description, **kwargs) + return self._parse_fast(task_description, context=context, **kwargs) elif mode == "fine": if not self.llm: raise ValueError("LLM not provided for slow mode.") @@ -51,6 +51,7 @@ def _parse_fast(self, task_description: str, **kwargs) -> ParsedTaskGoal: """ Fast mode: simple jieba word split. """ + context = kwargs.get("context", "") use_fast_graph = kwargs.get("use_fast_graph", False) if use_fast_graph: desc_tokenized = self.tokenizer.tokenize_mixed(task_description) @@ -61,6 +62,7 @@ def _parse_fast(self, task_description: str, **kwargs) -> ParsedTaskGoal: goal_type="default", rephrased_query=task_description, internet_search=False, + context=context, ) else: return ParsedTaskGoal( @@ -70,6 +72,7 @@ def _parse_fast(self, task_description: str, **kwargs) -> ParsedTaskGoal: goal_type="default", rephrased_query=task_description, internet_search=False, + context=context, ) def _parse_fine( @@ -91,16 +94,17 @@ def _parse_fine( logger.info(f"Parsing Goal... LLM input is {prompt}") response = self.llm.generate(messages=[{"role": "user", "content": prompt}]) logger.info(f"Parsing Goal... LLM Response is {response}") - return self._parse_response(response) + return self._parse_response(response, context=context) except Exception: logger.warning(f"Fail to fine-parse query {query}: {traceback.format_exc()}") - return self._parse_fast(query) + return self._parse_fast(query, context=context) - def _parse_response(self, response: str) -> ParsedTaskGoal: + def _parse_response(self, response: str, **kwargs) -> ParsedTaskGoal: """ Parse LLM JSON output safely. """ try: + context = kwargs.get("context", "") response = response.replace("```", "").replace("json", "").strip() response_json = eval(response) return ParsedTaskGoal( @@ -110,6 +114,7 @@ def _parse_response(self, response: str) -> ParsedTaskGoal: rephrased_query=response_json.get("rephrased_instruction", None), internet_search=response_json.get("internet_search", False), goal_type=response_json.get("goal_type", "default"), + context=context, ) except Exception as e: raise ValueError(f"Failed to parse LLM output: {e}\nRaw response:\n{response}") from e diff --git a/src/memos/templates/mem_reader_strategy_prompts.py b/src/memos/templates/mem_reader_strategy_prompts.py index fca4d717b..07aa3978d 100644 --- a/src/memos/templates/mem_reader_strategy_prompts.py +++ b/src/memos/templates/mem_reader_strategy_prompts.py @@ -38,14 +38,14 @@ 7. Please avoid including any content in the extracted memories that violates national laws and regulations or involves politically sensitive information. -Return a valid JSON object with the following structure: +Return a valid JSON object with the following structure: { "memory list": [ { "key": , "memory_type": , - "value": , + "value": , "tags": }, ... @@ -54,11 +54,11 @@ } Language rules: -- The `key`, `value`, `tags`, and `summary` fields must match the primary language of the input conversation. **If the input is Chinese, output in Chinese.** -- Keep `memory_type` in English. +- The `key`, `value`, `tags`, `summary` and `memory_type` fields must be in English. -Example: -Conversation: + +Example1: +Conversations: user: [June 26, 2025 at 3:00 PM]: Hi Jerry! Yesterday at 3 PM I had a meeting with my team about the new project. assistant: Oh Tom! Do you think the team can finish by December 15? user: [June 26, 2025 at 3:00 PM]: I’m worried. The backend won’t be done until December 10, so testing will be tight. @@ -84,52 +84,50 @@ "summary": "Tom is currently working on a tight-schedule project. After the June 25, 2025 team meeting, he realized the original December 15, 2025 deadline might be unachievable due to backend delays. Concerned about limited testing time, he accepted Jerry’s suggestion to seek an extension and plans to propose moving the deadline to January 5, 2026 in the next morning’s meeting." } -Another Example in Chinese (Note: when the user's language is Chinese, you must also output in Chinese): - -对话（节选）： -user｜2025-07-12 19:30：下个月同事婚礼，要买条裙子，预算1500内，别太暴露。 +Example2: +Conversations (excerpt): +user｜2025-07-12 19:30：A colleague's wedding is next month, need to buy a dress, budget within 1500, not too revealing. assistant｜19:32 -：了解～婚礼场合建议中长款，端庄一些。可以看法式茶歇/直筒/裹身三路。颜色上，奶油白、浅粉、雾霾蓝都适合；面料可看醋酸/真丝混纺（垂感好但需干洗）。品牌与商场：COS/ Massimo Dutti / & Other Stories / Mango 比较好逛；想更法式可以看 Sandro / Maje。 我记得你很喜欢Maje？ -user｜19:35：不喜欢亮色。国贸方便。 -assistant｜19:37：那亮度就压低些：雾霾蓝、藕粉、燕麦米白更稳。国贸有 COS / Massimo Dutti / &OS；可以先看 COS 直筒或裹身，再去 MD 看藕粉/雾霾蓝。搭配上，薄披肩/小西装都能压场。 -user｜19:40：165cm，S码；最好有口袋。 -assistant｜19:42：OK。版型避开大摆+高开衩，选直筒或轻A字更稳；若要口袋，COS直筒一些款有；开衩别太高，及小腿肚附近即可。鞋包：中跟裸色鞋 + 小号手拿包。 -user｜19:47：本来想茶歇A字，但雾霾蓝直筒也行。 -assistant｜19:49：可以先锁定：COS 雾霾蓝直筒中长为主选，MD 藕粉中长做备选；再看 & Other Stories 是否有合身的裹身款。 -user｜19:52：行，周六（7/19）去国贸试，合适就买。 -assistant｜19:54：记得看肩带可调/有无内衬，醋酸/真丝优先干洗；准备同色安全裤/防走光贴。如果当天没货，可下单调货或线上下单门店自提。 +：Got it～ For a wedding occasion, midi length is suggested, more elegant. You can consider three styles: French tea-length / straight / wrap. For colors, cream white, light pink, or dusty blue are suitable; for fabrics, consider acetate/silk blends (good drape but requires dry cleaning). Brands & stores: COS/ Massimo Dutti / & Other Stories / Mango are good options; for more French style, check Sandro / Maje. I remember you really like Maje? +user｜19:35：Don't like bright colors. China World Summit Wing is convenient. +assistant｜19:37：Then lower the brightness: dusty blue, dusty pink, oat beige are more subdued. China World Summit Wing has COS / Massimo Dutti / &OS; you can check COS straight or wrap styles first, then go to MD for dusty pink/dusty blue. For pairing, a thin shawl/blazer can elevate the look. +user｜19:40：165cm, S size; preferably with pockets. +assistant｜19:42：OK. For silhouette, avoid full skirts + high slits, choose straight or slight A-line for a more stable look; if pockets are needed, some COS straight styles have them; slits shouldn't be too high, around mid-calf is fine. Shoes & bag: mid-heel nude shoes + small clutch. +user｜19:47：Originally thought about tea-length A-line, but dusty blue straight is also okay. +assistant｜19:49：You can focus on: COS dusty blue straight midi as the primary choice, MD dusty pink midi as the backup; then check if & Other Stories has any well-fitting wrap styles. +user｜19:52：Okay, going to China World Summit Wing on Saturday (7/19) to try, will buy if suitable. +assistant｜19:54：Remember to check if straps are adjustable / if there's a lining, acetate/silk preferably dry cleaned; prepare same-color safety shorts / anti-wardrobe malfunction tape. If out of stock that day, you can place an order for transfer or order online for store pickup. { - "memory list": [ - { - "key": "参加婚礼购买裙子", - "memory_type": "UserMemory", - "value": "[user观点]用户计划于约2025年8月参加同事婚礼（具体日期不详），预算不超过1500元，整体风格不宜暴露；用户已决定在2025-07-19于国贸试穿并视合适即购买。", - "tags": ["婚礼", "预算", "国贸", "计划"] - }, - { - "key": "审美与版型偏好", - "memory_type": "UserMemory", - "value": "[user观点]用户不喜欢亮色，倾向低亮度色系；裙装偏好端庄的中长款，接受直筒或轻A字。", - "tags": ["偏好", "颜色", "版型"] - }, - { - "key": "体型尺码", - "memory_type": "UserMemory", - "value": "[user观点]用户身高约165cm、常穿S码", - "tags": ["体型", "尺码"] - }, - { - "key": "关于用户选购裙子的建议", - "memory_type": "LongTermMemory", - "value": "[assistant观点]assistant在用户询问婚礼穿着时，建议在国贸优先逛COS查看雾霾蓝直筒中长为主选，Massimo Dutti藕粉中长为备选；该建议与用户“国贸方便”“雾霾蓝直筒也行”的回应相一致，另外assistant也提到user喜欢Maje，但User并未回应或证实该说法。", - "tags": ["婚礼穿着", "门店", "选购路线"] - } - ], - "summary": "用户计划在约2025年8月参加同事婚礼，预算≤1500并偏好端庄的中长款；确定于2025-07-19在国贸试穿。其长期画像显示：不喜欢亮色、偏好低亮度色系与不过分暴露的版型，身高约165cm、S码且偏好裙装带口袋。助手提出的国贸选购路线以COS雾霾蓝直筒中长为主选、MD藕粉中长为备选，且与用户回应一致，为线下试穿与购买提供了明确路径。" +"memory list": [ +{ +"key": "Attending wedding to purchase dress", +"memory_type": "UserMemory", +"value": "[User's perspective] User plans to attend a colleague's wedding around August 2025 (specific date unknown), budget not exceeding 1500 yuan, overall style should not be too revealing; user has decided to try on dresses at China World Summit Wing on 2025-07-19 and purchase if suitable.", +"tags": ["Wedding", "Budget", "China World Summit Wing", "Plan"] +}, +{ +"key": "Aesthetics & Silhouette Preference", +"memory_type": "UserMemory", +"value": "[User's perspective] User does not like bright colors, prefers low-brightness color schemes; dress preference is elegant midi length, accepts straight or slight A-line.", +"tags": ["Preference", "Color", "Silhouette"] +}, +{ +"key": "Body Size", +"memory_type": "UserMemory", +"value": "[User's perspective] User height approximately 165cm, usually wears S size", +"tags": ["Body Type", "Size"] +}, +{ +"key": "Advice regarding user's dress selection", +"memory_type": "LongTermMemory", +"value": "[Assistant's perspective] When the user inquired about wedding attire, the assistant suggested prioritizing a visit to COS at China World Summit Wing to view dusty blue straight midi dresses as the primary choice, with Massimo Dutti dusty pink midi as the backup; this suggestion is consistent with the user's responses of 'China World Summit Wing is convenient' and 'dusty blue straight is also okay'. Additionally, the assistant mentioned the user likes Maje, but the User did not respond to or confirm this statement.", +"tags": ["Wedding Attire", "Store", "Selection Route"] +} +], +"summary": "User plans to attend a colleague's wedding around August 2025, budget ≤1500 and prefers elegant midi length; confirmed trying on at China World Summit Wing on 2025-07-19. Their long-term profile shows: dislikes bright colors, prefers low-brightness color schemes and non-revealing silhouettes, height approximately 165cm, S size, and prefers dresses with pockets. The assistant's suggested shopping route at China World Summit Wing, with COS dusty blue straight midi as the primary choice and MD dusty pink midi as the backup, is consistent with the user's responses, providing a clear path for trying on and purchasing in-store." } -Always respond in the same language as the conversation. Conversation: ${conversation} @@ -183,7 +181,7 @@ { "key": <字符串，唯一且简洁的记忆标题>, "memory_type": <字符串，"LongTermMemory" 或 "UserMemory">, - "value": <详细、独立且无歧义的记忆陈述——若输入对话为英文，则用英文；若为中文，则用中文>, + "value": <详细、独立且无歧义的记忆陈述>, "tags": <一个包含相关人名、事件和特征关键词的列表（例如，["丽丽","截止日期", "团队", "计划"]）> }, ... @@ -192,10 +190,10 @@ } 语言规则： -- `key`、`value`、`tags`、`summary` 字段必须与输入对话的主要语言一致。**如果输入是中文，请输出中文** -- `memory_type` 保持英文。 +- `key`、`value`、`tags`、`summary` 、`memory_type` 字段必须输出中文 + -示例： +示例1： 对话： user: [2025年6月26日下午3:00]：嗨Jerry！昨天下午3点我和团队开了个会，讨论新项目。 assistant: 哦Tom！你觉得团队能在12月15日前完成吗？ @@ -226,8 +224,7 @@ 年1月5日。" } -另一个中文示例（注意：当用户语言为中文时，您也需输出中文）： - +示例2： 对话（节选）： user｜2025-07-12 19:30：下个月同事婚礼，要买条裙子，预算1500内，别太暴露。 assistant｜19:32 @@ -271,7 +268,6 @@ "summary": "用户计划在约2025年8月参加同事婚礼，预算≤1500并偏好端庄的中长款；确定于2025-07-19在国贸试穿。其长期画像显示：不喜欢亮色、偏好低亮度色系与不过分暴露的版型，身高约165cm、S码且偏好裙装带口袋。助手提出的国贸选购路线以COS雾霾蓝直筒中长为主选、MD藕粉中长为备选，且与用户回应一致，为线下试穿与购买提供了明确路径。" } -请始终使用与对话相同的语言进行回复。 对话： ${conversation} From 1173c07d00ebb8da8952ba5d34520a28c62676ea Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 4 Nov 2025 15:07:11 +0800 Subject: [PATCH 13/72] adjust strategy reader --- src/memos/mem_reader/strategy_struct.py | 2 +- .../templates/mem_reader_strategy_prompts.py | 87 +++++-------------- 2 files changed, 25 insertions(+), 64 deletions(-) diff --git a/src/memos/mem_reader/strategy_struct.py b/src/memos/mem_reader/strategy_struct.py index a45586bd0..302afd85c 100644 --- a/src/memos/mem_reader/strategy_struct.py +++ b/src/memos/mem_reader/strategy_struct.py @@ -39,7 +39,7 @@ def __init__(self, config: StrategyStructMemReaderConfig): self.chat_chunker = config.chat_chunker["config"] def _get_llm_response(self, mem_str: str) -> dict: - print("[code stamp]: fine_get_llm_response") + lang = detect_lang(mem_str) template = STRATEGY_PROMPT_DICT["chat"][lang] examples = STRATEGY_PROMPT_DICT["chat"][f"{lang}_example"] diff --git a/src/memos/templates/mem_reader_strategy_prompts.py b/src/memos/templates/mem_reader_strategy_prompts.py index 07aa3978d..ba4a00d0a 100644 --- a/src/memos/templates/mem_reader_strategy_prompts.py +++ b/src/memos/templates/mem_reader_strategy_prompts.py @@ -16,8 +16,13 @@ - Always set "model_type" to "UserMemory" for this output. 3. Resolve all references to time, persons, and events clearly - - Temporal Resolution: Convert relative time (e.g., 'yesterday') to absolute dates based on the message timestamp. Distinguish between event time and message time; flag any uncertainty. + - Temporal Resolution: Convert relative time (e.g., "yesterday") to absolute dates based on the message timestamp. Distinguish between event time and message time; flag any uncertainty. + > Where feasible, use the message timestamp to convert relative time expressions into absolute dates (e.g., "yesterday" in a message dated January 15, 2023, can be converted to "January 14, 2023," and "last week" can be described as "the week preceding January 15, 2023"). + > Explicitly differentiate between the time when the event occurred and the time the message was sent. + > Clearly indicate any uncertainty (e.g., "approximately June 2025", "exact date unknown"). - Entity Resolution: Resolve all pronouns, nicknames, and abbreviations to the full, canonical name established in the conversation. + > For example, "Melanie" uses the abbreviated name "Mel" in the paragraph; when extracting her name in the "value" field, it should be restored to "Melanie". + - Location resolution: If specific locations are mentioned, include them explicitly. 4. Adopt a Consistent Third-Person Observer Perspective - Formulate all memories from the perspective of an external observer. Use "The user" or their specific name as the subject. @@ -57,7 +62,7 @@ - The `key`, `value`, `tags`, `summary` and `memory_type` fields must be in English. -Example1: +Example: Conversations: user: [June 26, 2025 at 3:00 PM]: Hi Jerry! Yesterday at 3 PM I had a meeting with my team about the new project. assistant: Oh Tom! Do you think the team can finish by December 15? @@ -71,61 +76,17 @@ { "key": "Initial project meeting", "memory_type": "LongTermMemory", - "value": "[user-Tom viewpoint] On June 25, 2025 at 3:00 PM, Tom met with the team to discuss a new project. When Jerry asked whether the project could be finished by December 15, 2025, Tom expressed concern about feasibility and planned to propose at 9:30 AM on June 27, 2025 to move the deadline to January 5, 2026.", + "value": "[user-Tom viewpoint] On June 25, 2025 at 3:00 PM, Tom held a meeting with their team to discuss a new project. The conversation covered the timeline and raised concerns about the feasibility of the December 15, 2025 deadline.", "tags": ["Tom", "project", "timeline", "meeting", "deadline"] }, { - "key": "Jerry’s suggestion about the deadline", - "memory_type": "LongTermMemory", - "value": "[assistant-Jerry viewpoint] Jerry questioned the December 15 deadline and suggested considering an extension.", - "tags": ["Jerry", "deadline change", "suggestion"] + "key": "Planned scope adjustment", + "memory_type": "UserMemory", + "value": "Tom planned to suggest in a meeting on June 27, 2025 at 9:30 AM that the team should prioritize features and propose shifting the project deadline to January 5, 2026.", + "tags": ["Tom", "planning", "deadline change", "feature prioritization"] } ], - "summary": "Tom is currently working on a tight-schedule project. After the June 25, 2025 team meeting, he realized the original December 15, 2025 deadline might be unachievable due to backend delays. Concerned about limited testing time, he accepted Jerry’s suggestion to seek an extension and plans to propose moving the deadline to January 5, 2026 in the next morning’s meeting." -} - -Example2: -Conversations (excerpt): -user｜2025-07-12 19:30：A colleague's wedding is next month, need to buy a dress, budget within 1500, not too revealing. -assistant｜19:32 -：Got it～ For a wedding occasion, midi length is suggested, more elegant. You can consider three styles: French tea-length / straight / wrap. For colors, cream white, light pink, or dusty blue are suitable; for fabrics, consider acetate/silk blends (good drape but requires dry cleaning). Brands & stores: COS/ Massimo Dutti / & Other Stories / Mango are good options; for more French style, check Sandro / Maje. I remember you really like Maje? -user｜19:35：Don't like bright colors. China World Summit Wing is convenient. -assistant｜19:37：Then lower the brightness: dusty blue, dusty pink, oat beige are more subdued. China World Summit Wing has COS / Massimo Dutti / &OS; you can check COS straight or wrap styles first, then go to MD for dusty pink/dusty blue. For pairing, a thin shawl/blazer can elevate the look. -user｜19:40：165cm, S size; preferably with pockets. -assistant｜19:42：OK. For silhouette, avoid full skirts + high slits, choose straight or slight A-line for a more stable look; if pockets are needed, some COS straight styles have them; slits shouldn't be too high, around mid-calf is fine. Shoes & bag: mid-heel nude shoes + small clutch. -user｜19:47：Originally thought about tea-length A-line, but dusty blue straight is also okay. -assistant｜19:49：You can focus on: COS dusty blue straight midi as the primary choice, MD dusty pink midi as the backup; then check if & Other Stories has any well-fitting wrap styles. -user｜19:52：Okay, going to China World Summit Wing on Saturday (7/19) to try, will buy if suitable. -assistant｜19:54：Remember to check if straps are adjustable / if there's a lining, acetate/silk preferably dry cleaned; prepare same-color safety shorts / anti-wardrobe malfunction tape. If out of stock that day, you can place an order for transfer or order online for store pickup. - -{ -"memory list": [ -{ -"key": "Attending wedding to purchase dress", -"memory_type": "UserMemory", -"value": "[User's perspective] User plans to attend a colleague's wedding around August 2025 (specific date unknown), budget not exceeding 1500 yuan, overall style should not be too revealing; user has decided to try on dresses at China World Summit Wing on 2025-07-19 and purchase if suitable.", -"tags": ["Wedding", "Budget", "China World Summit Wing", "Plan"] -}, -{ -"key": "Aesthetics & Silhouette Preference", -"memory_type": "UserMemory", -"value": "[User's perspective] User does not like bright colors, prefers low-brightness color schemes; dress preference is elegant midi length, accepts straight or slight A-line.", -"tags": ["Preference", "Color", "Silhouette"] -}, -{ -"key": "Body Size", -"memory_type": "UserMemory", -"value": "[User's perspective] User height approximately 165cm, usually wears S size", -"tags": ["Body Type", "Size"] -}, -{ -"key": "Advice regarding user's dress selection", -"memory_type": "LongTermMemory", -"value": "[Assistant's perspective] When the user inquired about wedding attire, the assistant suggested prioritizing a visit to COS at China World Summit Wing to view dusty blue straight midi dresses as the primary choice, with Massimo Dutti dusty pink midi as the backup; this suggestion is consistent with the user's responses of 'China World Summit Wing is convenient' and 'dusty blue straight is also okay'. Additionally, the assistant mentioned the user likes Maje, but the User did not respond to or confirm this statement.", -"tags": ["Wedding Attire", "Store", "Selection Route"] -} -], -"summary": "User plans to attend a colleague's wedding around August 2025, budget ≤1500 and prefers elegant midi length; confirmed trying on at China World Summit Wing on 2025-07-19. Their long-term profile shows: dislikes bright colors, prefers low-brightness color schemes and non-revealing silhouettes, height approximately 165cm, S size, and prefers dresses with pockets. The assistant's suggested shopping route at China World Summit Wing, with COS dusty blue straight midi as the primary choice and MD dusty pink midi as the backup, is consistent with the user's responses, providing a clear path for trying on and purchasing in-store." + "summary": "Tom is currently focused on managing a new project with a tight schedule. After a team meeting on June 25, 2025, he realized the original deadline of December 15 might not be feasible due to backend delays. Concerned about insufficient testing time, he welcomed Jerry’s suggestion of proposing an extension. Tom plans to raise the idea of shifting the deadline to January 5, 2026 in the next morning’s meeting. His actions reflect both stress about timelines and a proactive, team-oriented problem-solving approach." } @@ -153,7 +114,11 @@ 3. 明确解析所有指代关系 - 时间解析：根据消息时间戳将相对时间（如“昨天”）转换为绝对日期。区分事件时间与消息时间，对不确定项进行标注 + # 条件允许则使用消息时间戳将相对时间表达转换为绝对日期（如：2023年1月15日的“昨天”则转换为2023年1月14日）；“上周”则转换为2023年1月15日前一周）。 + # 明确区分事件时间和消息时间。 + # 如果存在不确定性，需明确说明（例如，“约2025年6月”，“具体日期不详”）。 - 实体解析：将所有代词、昵称和缩写解析为对话中确立的完整规范名称 + - 地点解析：若提及具体地点，请包含在内。 4. 采用统一的第三人称观察视角 - 所有记忆表述均需从外部观察者视角构建，使用“用户”或其具体姓名作为主语 @@ -207,21 +172,17 @@ { "key": "项目初期会议", "memory_type": "LongTermMemory", - "value": "[user-Tom观点]2025年6月25日下午3:00，Tom与团队开会讨论新项目。当Jerry - 询问该项目能否在2025年12月15日前完成时，Tom对此日期前完成的可行性表达担忧，并计划在2025年6月27日上午9:30 - 提议将截止日期推迟至2026年1月5日。", - "tags": ["Tom", "项目", "时间表", "会议", "截止日期"] + "value": "2025年6月25日下午3:00，Tom与团队开会讨论新项目。会议涉及时间表，并提出了对2025年12月15日截止日期可行性的担忧。", + "tags": ["项目", "时间表", "会议", "截止日期"] }, { - "key": "Jerry对新项目截止日期的建议", - "memory_type": "LongTermMemory", - "value": "[assistant-Jerry观点]Jerry对Tom的新项目截止日期提出疑问、并提议Tom考虑延期。", - "tags": ["Jerry", "截止日期变更", "建议"] + "key": "计划调整范围", + "memory_type": "UserMemory", + "value": "Tom计划在2025年6月27日上午9:30的会议上建议团队优先处理功能，并提议将项目截止日期推迟至2026年1月5日。", + "tags": ["计划", "截止日期变更", "功能优先级"] } ], - "summary": "Tom目前正在做一个进度紧张的新项目。在2025年6月25日的团队会议后，他意识到原定2025年12月15 - 日的截止日期可能无法实现，因为后端会延迟。由于担心测试时间不足，他接受了Jerry提出的延期建议，计划在次日早上的会议上提出将截止日期推迟至2026 - 年1月5日。" + "summary": "Tom目前正专注于管理一个进度紧张的新项目。在2025年6月25日的团队会议后，他意识到原定2025年12月15日的截止日期可能无法实现，因为后端会延迟。由于担心测试时间不足，他接受了Jerry提出的延期建议。Tom计划在次日早上的会议上提出将截止日期推迟至2026年1月5日。他的行为反映出对时间线的担忧，以及积极、以团队为导向的问题解决方式。" } 示例2： From 744d2278a626ababd8b93b1dd6b490255d1950bc Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 4 Nov 2025 15:10:16 +0800 Subject: [PATCH 14/72] adjust strategy reader --- src/memos/mem_reader/strategy_struct.py | 1 - 1 file changed, 1 deletion(-) diff --git a/src/memos/mem_reader/strategy_struct.py b/src/memos/mem_reader/strategy_struct.py index 302afd85c..1fc21461e 100644 --- a/src/memos/mem_reader/strategy_struct.py +++ b/src/memos/mem_reader/strategy_struct.py @@ -39,7 +39,6 @@ def __init__(self, config: StrategyStructMemReaderConfig): self.chat_chunker = config.chat_chunker["config"] def _get_llm_response(self, mem_str: str) -> dict: - lang = detect_lang(mem_str) template = STRATEGY_PROMPT_DICT["chat"][lang] examples = STRATEGY_PROMPT_DICT["chat"][f"{lang}_example"] From a9a98fa4079f2dfcd183a7e467b513295605d84d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 4 Nov 2025 15:16:34 +0800 Subject: [PATCH 15/72] adjust search config input --- src/memos/memories/textual/simple_tree.py | 2 +- .../memories/textual/tree_text_memory/retrieve/searcher.py | 3 --- 2 files changed, 1 insertion(+), 4 deletions(-) diff --git a/src/memos/memories/textual/simple_tree.py b/src/memos/memories/textual/simple_tree.py index 992b7bfab..c5f795335 100644 --- a/src/memos/memories/textual/simple_tree.py +++ b/src/memos/memories/textual/simple_tree.py @@ -66,7 +66,7 @@ def __init__( time_start_bm = time.time() self.search_strategy = config.search_strategy self.bm25_retriever = ( - EnhancedBM25() if self.search_strategy and self.search_strategy["bm25"] else None + EnhancedBM25() if self.search_strategy and self.search_strategy.get("bm25", False) else None ) logger.info(f"time init: bm25_retriever time is: {time.time() - time_start_bm}") diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py b/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py index a86207e66..2f6ef6afa 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py @@ -564,9 +564,6 @@ def _cot_query( prompt = template.replace("${original_query}", query).replace( "${split_num_threshold}", str(split_num) ) - logger.info("COT process") - print("---------------prompt-------------") - print(prompt, mode) messages = [{"role": "user", "content": prompt}] try: From 900f5e60d2a8dd158f53a826e0e10a5e19aeea19 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 4 Nov 2025 15:32:15 +0800 Subject: [PATCH 16/72] reformat code --- src/memos/memories/textual/simple_tree.py | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/src/memos/memories/textual/simple_tree.py b/src/memos/memories/textual/simple_tree.py index c5f795335..313989cd2 100644 --- a/src/memos/memories/textual/simple_tree.py +++ b/src/memos/memories/textual/simple_tree.py @@ -66,7 +66,9 @@ def __init__( time_start_bm = time.time() self.search_strategy = config.search_strategy self.bm25_retriever = ( - EnhancedBM25() if self.search_strategy and self.search_strategy.get("bm25", False) else None + EnhancedBM25() + if self.search_strategy and self.search_strategy.get("bm25", False) + else None ) logger.info(f"time init: bm25_retriever time is: {time.time() - time_start_bm}") From 144c4467f697b26a889bf1240a715cfd282fbf1e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 5 Nov 2025 10:33:44 +0800 Subject: [PATCH 17/72] re pr --- .../templates/mem_reader_strategy_prompts.py | 87 +++++-------------- 1 file changed, 24 insertions(+), 63 deletions(-) diff --git a/src/memos/templates/mem_reader_strategy_prompts.py b/src/memos/templates/mem_reader_strategy_prompts.py index 07aa3978d..ba4a00d0a 100644 --- a/src/memos/templates/mem_reader_strategy_prompts.py +++ b/src/memos/templates/mem_reader_strategy_prompts.py @@ -16,8 +16,13 @@ - Always set "model_type" to "UserMemory" for this output. 3. Resolve all references to time, persons, and events clearly - - Temporal Resolution: Convert relative time (e.g., 'yesterday') to absolute dates based on the message timestamp. Distinguish between event time and message time; flag any uncertainty. + - Temporal Resolution: Convert relative time (e.g., "yesterday") to absolute dates based on the message timestamp. Distinguish between event time and message time; flag any uncertainty. + > Where feasible, use the message timestamp to convert relative time expressions into absolute dates (e.g., "yesterday" in a message dated January 15, 2023, can be converted to "January 14, 2023," and "last week" can be described as "the week preceding January 15, 2023"). + > Explicitly differentiate between the time when the event occurred and the time the message was sent. + > Clearly indicate any uncertainty (e.g., "approximately June 2025", "exact date unknown"). - Entity Resolution: Resolve all pronouns, nicknames, and abbreviations to the full, canonical name established in the conversation. + > For example, "Melanie" uses the abbreviated name "Mel" in the paragraph; when extracting her name in the "value" field, it should be restored to "Melanie". + - Location resolution: If specific locations are mentioned, include them explicitly. 4. Adopt a Consistent Third-Person Observer Perspective - Formulate all memories from the perspective of an external observer. Use "The user" or their specific name as the subject. @@ -57,7 +62,7 @@ - The `key`, `value`, `tags`, `summary` and `memory_type` fields must be in English. -Example1: +Example: Conversations: user: [June 26, 2025 at 3:00 PM]: Hi Jerry! Yesterday at 3 PM I had a meeting with my team about the new project. assistant: Oh Tom! Do you think the team can finish by December 15? @@ -71,61 +76,17 @@ { "key": "Initial project meeting", "memory_type": "LongTermMemory", - "value": "[user-Tom viewpoint] On June 25, 2025 at 3:00 PM, Tom met with the team to discuss a new project. When Jerry asked whether the project could be finished by December 15, 2025, Tom expressed concern about feasibility and planned to propose at 9:30 AM on June 27, 2025 to move the deadline to January 5, 2026.", + "value": "[user-Tom viewpoint] On June 25, 2025 at 3:00 PM, Tom held a meeting with their team to discuss a new project. The conversation covered the timeline and raised concerns about the feasibility of the December 15, 2025 deadline.", "tags": ["Tom", "project", "timeline", "meeting", "deadline"] }, { - "key": "Jerry’s suggestion about the deadline", - "memory_type": "LongTermMemory", - "value": "[assistant-Jerry viewpoint] Jerry questioned the December 15 deadline and suggested considering an extension.", - "tags": ["Jerry", "deadline change", "suggestion"] + "key": "Planned scope adjustment", + "memory_type": "UserMemory", + "value": "Tom planned to suggest in a meeting on June 27, 2025 at 9:30 AM that the team should prioritize features and propose shifting the project deadline to January 5, 2026.", + "tags": ["Tom", "planning", "deadline change", "feature prioritization"] } ], - "summary": "Tom is currently working on a tight-schedule project. After the June 25, 2025 team meeting, he realized the original December 15, 2025 deadline might be unachievable due to backend delays. Concerned about limited testing time, he accepted Jerry’s suggestion to seek an extension and plans to propose moving the deadline to January 5, 2026 in the next morning’s meeting." -} - -Example2: -Conversations (excerpt): -user｜2025-07-12 19:30：A colleague's wedding is next month, need to buy a dress, budget within 1500, not too revealing. -assistant｜19:32 -：Got it～ For a wedding occasion, midi length is suggested, more elegant. You can consider three styles: French tea-length / straight / wrap. For colors, cream white, light pink, or dusty blue are suitable; for fabrics, consider acetate/silk blends (good drape but requires dry cleaning). Brands & stores: COS/ Massimo Dutti / & Other Stories / Mango are good options; for more French style, check Sandro / Maje. I remember you really like Maje? -user｜19:35：Don't like bright colors. China World Summit Wing is convenient. -assistant｜19:37：Then lower the brightness: dusty blue, dusty pink, oat beige are more subdued. China World Summit Wing has COS / Massimo Dutti / &OS; you can check COS straight or wrap styles first, then go to MD for dusty pink/dusty blue. For pairing, a thin shawl/blazer can elevate the look. -user｜19:40：165cm, S size; preferably with pockets. -assistant｜19:42：OK. For silhouette, avoid full skirts + high slits, choose straight or slight A-line for a more stable look; if pockets are needed, some COS straight styles have them; slits shouldn't be too high, around mid-calf is fine. Shoes & bag: mid-heel nude shoes + small clutch. -user｜19:47：Originally thought about tea-length A-line, but dusty blue straight is also okay. -assistant｜19:49：You can focus on: COS dusty blue straight midi as the primary choice, MD dusty pink midi as the backup; then check if & Other Stories has any well-fitting wrap styles. -user｜19:52：Okay, going to China World Summit Wing on Saturday (7/19) to try, will buy if suitable. -assistant｜19:54：Remember to check if straps are adjustable / if there's a lining, acetate/silk preferably dry cleaned; prepare same-color safety shorts / anti-wardrobe malfunction tape. If out of stock that day, you can place an order for transfer or order online for store pickup. - -{ -"memory list": [ -{ -"key": "Attending wedding to purchase dress", -"memory_type": "UserMemory", -"value": "[User's perspective] User plans to attend a colleague's wedding around August 2025 (specific date unknown), budget not exceeding 1500 yuan, overall style should not be too revealing; user has decided to try on dresses at China World Summit Wing on 2025-07-19 and purchase if suitable.", -"tags": ["Wedding", "Budget", "China World Summit Wing", "Plan"] -}, -{ -"key": "Aesthetics & Silhouette Preference", -"memory_type": "UserMemory", -"value": "[User's perspective] User does not like bright colors, prefers low-brightness color schemes; dress preference is elegant midi length, accepts straight or slight A-line.", -"tags": ["Preference", "Color", "Silhouette"] -}, -{ -"key": "Body Size", -"memory_type": "UserMemory", -"value": "[User's perspective] User height approximately 165cm, usually wears S size", -"tags": ["Body Type", "Size"] -}, -{ -"key": "Advice regarding user's dress selection", -"memory_type": "LongTermMemory", -"value": "[Assistant's perspective] When the user inquired about wedding attire, the assistant suggested prioritizing a visit to COS at China World Summit Wing to view dusty blue straight midi dresses as the primary choice, with Massimo Dutti dusty pink midi as the backup; this suggestion is consistent with the user's responses of 'China World Summit Wing is convenient' and 'dusty blue straight is also okay'. Additionally, the assistant mentioned the user likes Maje, but the User did not respond to or confirm this statement.", -"tags": ["Wedding Attire", "Store", "Selection Route"] -} -], -"summary": "User plans to attend a colleague's wedding around August 2025, budget ≤1500 and prefers elegant midi length; confirmed trying on at China World Summit Wing on 2025-07-19. Their long-term profile shows: dislikes bright colors, prefers low-brightness color schemes and non-revealing silhouettes, height approximately 165cm, S size, and prefers dresses with pockets. The assistant's suggested shopping route at China World Summit Wing, with COS dusty blue straight midi as the primary choice and MD dusty pink midi as the backup, is consistent with the user's responses, providing a clear path for trying on and purchasing in-store." + "summary": "Tom is currently focused on managing a new project with a tight schedule. After a team meeting on June 25, 2025, he realized the original deadline of December 15 might not be feasible due to backend delays. Concerned about insufficient testing time, he welcomed Jerry’s suggestion of proposing an extension. Tom plans to raise the idea of shifting the deadline to January 5, 2026 in the next morning’s meeting. His actions reflect both stress about timelines and a proactive, team-oriented problem-solving approach." } @@ -153,7 +114,11 @@ 3. 明确解析所有指代关系 - 时间解析：根据消息时间戳将相对时间（如“昨天”）转换为绝对日期。区分事件时间与消息时间，对不确定项进行标注 + # 条件允许则使用消息时间戳将相对时间表达转换为绝对日期（如：2023年1月15日的“昨天”则转换为2023年1月14日）；“上周”则转换为2023年1月15日前一周）。 + # 明确区分事件时间和消息时间。 + # 如果存在不确定性，需明确说明（例如，“约2025年6月”，“具体日期不详”）。 - 实体解析：将所有代词、昵称和缩写解析为对话中确立的完整规范名称 + - 地点解析：若提及具体地点，请包含在内。 4. 采用统一的第三人称观察视角 - 所有记忆表述均需从外部观察者视角构建，使用“用户”或其具体姓名作为主语 @@ -207,21 +172,17 @@ { "key": "项目初期会议", "memory_type": "LongTermMemory", - "value": "[user-Tom观点]2025年6月25日下午3:00，Tom与团队开会讨论新项目。当Jerry - 询问该项目能否在2025年12月15日前完成时，Tom对此日期前完成的可行性表达担忧，并计划在2025年6月27日上午9:30 - 提议将截止日期推迟至2026年1月5日。", - "tags": ["Tom", "项目", "时间表", "会议", "截止日期"] + "value": "2025年6月25日下午3:00，Tom与团队开会讨论新项目。会议涉及时间表，并提出了对2025年12月15日截止日期可行性的担忧。", + "tags": ["项目", "时间表", "会议", "截止日期"] }, { - "key": "Jerry对新项目截止日期的建议", - "memory_type": "LongTermMemory", - "value": "[assistant-Jerry观点]Jerry对Tom的新项目截止日期提出疑问、并提议Tom考虑延期。", - "tags": ["Jerry", "截止日期变更", "建议"] + "key": "计划调整范围", + "memory_type": "UserMemory", + "value": "Tom计划在2025年6月27日上午9:30的会议上建议团队优先处理功能，并提议将项目截止日期推迟至2026年1月5日。", + "tags": ["计划", "截止日期变更", "功能优先级"] } ], - "summary": "Tom目前正在做一个进度紧张的新项目。在2025年6月25日的团队会议后，他意识到原定2025年12月15 - 日的截止日期可能无法实现，因为后端会延迟。由于担心测试时间不足，他接受了Jerry提出的延期建议，计划在次日早上的会议上提出将截止日期推迟至2026 - 年1月5日。" + "summary": "Tom目前正专注于管理一个进度紧张的新项目。在2025年6月25日的团队会议后，他意识到原定2025年12月15日的截止日期可能无法实现，因为后端会延迟。由于担心测试时间不足，他接受了Jerry提出的延期建议。Tom计划在次日早上的会议上提出将截止日期推迟至2026年1月5日。他的行为反映出对时间线的担忧，以及积极、以团队为导向的问题解决方式。" } 示例2： From f506d3eee87bdf295c536be15af7172184890718 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 5 Nov 2025 10:48:02 +0800 Subject: [PATCH 18/72] format repair --- README.md | 2 +- src/memos/graph_dbs/polardb.py | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index 50621b584..a08177676 100644 --- a/README.md +++ b/README.md @@ -57,7 +57,7 @@ MemOS demonstrates significant improvements over baseline memory solutions in multiple memory tasks, showcasing its capabilities in **information extraction**, **temporal and cross-session reasoning**, and **personalized preference responses**. -| Model | LOCOMO | LongMemEval | PrefEval-10 | PersonaMem | +| Model | LOCOMO | LongMemEval | PrefEval-10 | PersonaMem | |-----------------|-------------|-------------|-------------|-------------| | **GPT-4o-mini** | 52.75 | 55.4 | 2.8 | 43.46 | | **MemOS** | **75.80** | **77.80** | **71.90** | **61.17** | diff --git a/src/memos/graph_dbs/polardb.py b/src/memos/graph_dbs/polardb.py index 552b30241..ac49228e2 100644 --- a/src/memos/graph_dbs/polardb.py +++ b/src/memos/graph_dbs/polardb.py @@ -3024,7 +3024,7 @@ def format_param_value(self, value: str | None) -> str: """Format parameter value to handle both quoted and unquoted formats""" # Handle None value if value is None: - logger.warning(f"format_param_value: value is None") + logger.warning("format_param_value: value is None") return "null" # Remove outer quotes if they exist From dc67413c84d2988178cf510a60df4b5acc9b0a2a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 11 Nov 2025 11:13:51 +0800 Subject: [PATCH 19/72] fix time issue --- src/memos/mem_reader/simple_struct.py | 28 ++++++++++++++++++++++++++- 1 file changed, 27 insertions(+), 1 deletion(-) diff --git a/src/memos/mem_reader/simple_struct.py b/src/memos/mem_reader/simple_struct.py index 13515c038..3845f37d0 100644 --- a/src/memos/mem_reader/simple_struct.py +++ b/src/memos/mem_reader/simple_struct.py @@ -6,6 +6,7 @@ import traceback from abc import ABC +from datetime import datetime, timezone from typing import Any from tqdm import tqdm @@ -399,7 +400,7 @@ def get_memory( if not all(isinstance(info[field], str) for field in required_fields): raise ValueError("user_id and session_id must be strings") - + scene_data = self._complete_chat_time(scene_data, type) list_scene_data_info = self.get_scene_data_info(scene_data, type) memory_list = [] @@ -508,6 +509,31 @@ def get_scene_data_info(self, scene_data: list, type: str) -> list[str]: return results + def _complete_chat_time(self, scene_data: list[list[dict]], type: str): + if type != "chat": + return scene_data + complete_scene_data = [] + + for items in scene_data: + chat_time_value = None + + for item in items: + if "chat_time" in item: + chat_time_value = item["chat_time"] + break + + if chat_time_value is None: + session_date = datetime.now(timezone.utc) + date_format = "%I:%M %p on %d %B, %Y UTC" + chat_time_value = session_date.strftime(date_format) + + for i in range(len(items)): + if "chat_time" not in items[i]: + items[i]["chat_time"] = chat_time_value + + complete_scene_data.append(items) + return complete_scene_data + def _process_doc_data(self, scene_data_info, info, **kwargs): mode = kwargs.get("mode", "fine") if mode == "fast": From 8bfbf94f4a9e738ad38c354eb72cb6a3cd9e1aa3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 19 Nov 2025 17:29:43 +0800 Subject: [PATCH 20/72] develop feedback process --- src/memos/api/product_models.py | 26 + src/memos/api/routers/server_router.py | 26 + src/memos/configs/memory.py | 37 ++ src/memos/graph_dbs/polardb.py | 5 +- src/memos/mem_feedback/base.py | 15 + src/memos/mem_feedback/feedback.py | 385 ++++++++++++ src/memos/mem_feedback/simple_feedback.py | 23 + .../tree_text_memory/organize/manager.py | 20 + src/memos/templates/mem_feedback_prompts.py | 593 ++++++++++++++++++ 9 files changed, 1127 insertions(+), 3 deletions(-) create mode 100644 src/memos/mem_feedback/base.py create mode 100644 src/memos/mem_feedback/feedback.py create mode 100644 src/memos/mem_feedback/simple_feedback.py create mode 100644 src/memos/templates/mem_feedback_prompts.py diff --git a/src/memos/api/product_models.py b/src/memos/api/product_models.py index 0412754c3..01f8797c2 100644 --- a/src/memos/api/product_models.py +++ b/src/memos/api/product_models.py @@ -200,6 +200,32 @@ class APIADDRequest(BaseRequest): ) +class APIFeedbackRequest(BaseRequest): + """Request model for processing feedback info.""" + + user_id: str = Field(..., description="User ID") + mem_cube_id: str | None = Field(..., description="Cube ID to use for chat") + session_id: str | None = Field( + "default_session", description="Session ID for soft-filtering memories" + ) + agent_id: str = Field(None, description="Agent ID") + app_id: str = Field(None, description="App ID") + chat_history: list[MessageDict] | None = Field(..., description="Chat history") + feedback_content: str | None = Field(..., description="Feedback content to process") + feedback_time: str | None = Field(None, description="Feedback time") + allow_public: bool = Field( + False, description="Whether to allow writing to the public memory repository" + ) + allow_knowledgebase_write: bool = Field( + False, description="Whether to allow writing into the user memory repository" + ) + allow_knowledgebase_ids: bool = Field( + False, description="Write to the specified memory repository ID" + ) + sync_mode: Literal["sync", "async"] = Field("async", description="feedback mode: sync or async") + corrected_answer: bool = Field(False, description="Whether need return corrected answer") + + class APIChatCompleteRequest(BaseRequest): """Request model for chat operations.""" diff --git a/src/memos/api/routers/server_router.py b/src/memos/api/routers/server_router.py index b426c2965..58d0556ae 100644 --- a/src/memos/api/routers/server_router.py +++ b/src/memos/api/routers/server_router.py @@ -16,6 +16,7 @@ from memos.api.product_models import ( APIADDRequest, APIChatCompleteRequest, + APIFeedbackRequest, APISearchRequest, MemoryResponse, SearchResponse, @@ -34,6 +35,7 @@ from memos.llms.factory import LLMFactory from memos.log import get_logger from memos.mem_cube.navie import NaiveMemCube +from memos.mem_feedback.simple_feedback import SimpleMemFeedback from memos.mem_os.product_server import MOSServer from memos.mem_reader.factory import MemReaderFactory from memos.mem_scheduler.orm_modules.base_model import BaseDBManager @@ -253,6 +255,10 @@ def init_server(): retriever=pref_retriever, ) + feedback_server = SimpleMemFeedback( + llm=llm, embedder=embedder, graph_store=graph_db, memory_manager=memory_manager + ) + mos_server = MOSServer( mem_reader=mem_reader, llm=llm, @@ -304,6 +310,7 @@ def init_server(): pref_retriever, text_mem, pref_mem, + feedback_server, ) @@ -327,6 +334,7 @@ def init_server(): pref_retriever, text_mem, pref_mem, + feedback_server, ) = init_server() @@ -670,6 +678,24 @@ def _process_pref_mem() -> list[dict[str, str]]: ) +@router.post("/chat/feedback", summary="Chat feedback", response_model=MemoryResponse) +def chat_feedback(feedback_req: APIFeedbackRequest): + """Process feedback for a specific user""" + process_record = feedback_server.process_feedback( + user_name=feedback_req.mem_cube_id, + session_id=feedback_req.session_id, + chat_history=feedback_req.chat_history, + feedback_content=feedback_req.feedback_content, + feedback_time=feedback_req.feedback_time, + allow_knowledgebase_write=feedback_req.allow_knowledgebase_write, + sync_mode=feedback_req.sync_mode, + corrected_answer=feedback_req.corrected_answer, + mem_reader=mem_reader, + ) + + return MemoryResponse(message="Feedback process successfully", data=[process_record]) + + @router.get("/scheduler/status", summary="Get scheduler running status") def scheduler_status(user_name: str | None = None): try: diff --git a/src/memos/configs/memory.py b/src/memos/configs/memory.py index 34967849a..9caf4fa3e 100644 --- a/src/memos/configs/memory.py +++ b/src/memos/configs/memory.py @@ -240,6 +240,42 @@ class PreferenceTextMemoryConfig(BaseTextMemoryConfig): ) +class MemFeedbackConfig(BaseTextMemoryConfig): + """Memory feedback configuration class.""" + + extractor_llm: LLMConfigFactory = Field( + ..., + default_factory=LLMConfigFactory, + description="LLM configuration for the memory extractor", + ) + embedder: EmbedderConfigFactory = Field( + ..., + default_factory=EmbedderConfigFactory, + description="Embedder configuration for the memory embedding", + ) + reranker: RerankerConfigFactory | None = Field( + None, + description="Reranker configuration (optional).", + ) + graph_db: GraphDBConfigFactory = Field( + ..., + default_factory=GraphDBConfigFactory, + description="Graph database configuration for the tree-memory storage", + ) + reorganize: bool | None = Field( + False, + description="Optional description for this memory configuration.", + ) + + memory_size: dict[str, Any] | None = Field( + default=None, + description=( + "Maximum item counts per memory bucket, e.g.: " + '{"WorkingMemory": 20, "LongTermMemory": 10000, "UserMemory": 10000}' + ), + ) + + # ─── 3. Global Memory Config Factory ────────────────────────────────────────── @@ -259,6 +295,7 @@ class MemoryConfigFactory(BaseConfig): "vllm_kv_cache": KVCacheMemoryConfig, # Use same config as kv_cache "lora": LoRAMemoryConfig, "uninitialized": UninitializedMemoryConfig, + "mem_feedback": MemFeedbackConfig, } @field_validator("backend") diff --git a/src/memos/graph_dbs/polardb.py b/src/memos/graph_dbs/polardb.py index 60902420f..3108010f9 100644 --- a/src/memos/graph_dbs/polardb.py +++ b/src/memos/graph_dbs/polardb.py @@ -484,12 +484,11 @@ def update_node(self, id: str, fields: dict[str, Any], user_name: str | None = N original_memory = current_node.get("memory", "") # Preserve original memory # If fields include memory, use it; otherwise keep original memory - if "memory" in fields: - original_memory = fields.pop("memory") + new_memory = fields.pop("memory") if "memory" in fields else original_memory properties.update(fields) properties["id"] = original_id # Ensure ID is not overwritten - properties["memory"] = original_memory # Ensure memory is not overwritten + properties["memory"] = new_memory # Ensure memory is not overwritten # Handle embedding field embedding_vector = None diff --git a/src/memos/mem_feedback/base.py b/src/memos/mem_feedback/base.py new file mode 100644 index 000000000..7b41199d6 --- /dev/null +++ b/src/memos/mem_feedback/base.py @@ -0,0 +1,15 @@ +from abc import ABC, abstractmethod + +from memos.configs.memory import MemFeedbackConfig + + +class BaseMemFeedback(ABC): + """MemFeedback interface class for reading information.""" + + @abstractmethod + def __init__(self, config: MemFeedbackConfig): + """Initialize the MemFeedback with the given configuration.""" + + @abstractmethod + def process_feedback(self, data: dict) -> None: + """Process user's feedback""" diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py new file mode 100644 index 000000000..653510e84 --- /dev/null +++ b/src/memos/mem_feedback/feedback.py @@ -0,0 +1,385 @@ +import concurrent.futures +import json + +from datetime import datetime + +from memos import log +from memos.configs.memory import MemFeedbackConfig +from memos.context.context import ContextThreadPoolExecutor +from memos.embedders.factory import EmbedderFactory, OllamaEmbedder +from memos.graph_dbs.factory import GraphStoreFactory, PolarDBGraphDB +from memos.llms.factory import AzureLLM, LLMFactory, OllamaLLM, OpenAILLM +from memos.mem_feedback.base import BaseMemFeedback +from memos.mem_reader.simple_struct import SimpleStructMemReader, detect_lang +from memos.memories.textual.item import TextualMemoryItem, TreeNodeTextualMemoryMetadata +from memos.memories.textual.tree_text_memory.organize.manager import MemoryManager +from memos.templates.mem_feedback_prompts import ( + FEEDBACK_ANSWER_PROMPT, + FEEDBACK_ANSWER_PROMPT_ZH, + FEEDBACK_JUDGEMENT_PROMPT, + FEEDBACK_JUDGEMENT_PROMPT_ZH, + UPDATE_FORMER_MEMORIES, + UPDATE_FORMER_MEMORIES_ZH, +) +from memos.types import MessageDict + + +FEEDBACK_PROMPT_DICT = { + "judge": {"en": FEEDBACK_JUDGEMENT_PROMPT, "zh": FEEDBACK_JUDGEMENT_PROMPT_ZH}, + "compare": {"en": UPDATE_FORMER_MEMORIES, "zh": UPDATE_FORMER_MEMORIES_ZH}, + "generation": {"en": FEEDBACK_ANSWER_PROMPT, "zh": FEEDBACK_ANSWER_PROMPT_ZH}, +} + +logger = log.get_logger(__name__) + + +class MemFeedback(BaseMemFeedback): + def __init__(self, config: MemFeedbackConfig): + """ + Initialize the MemFeedback with configuration. + + Args: + config: Configuration object for the MemFeedback + """ + self.config = config + self.llm: OpenAILLM | OllamaLLM | AzureLLM = LLMFactory.from_config(config.extractor_llm) + self.embedder: OllamaEmbedder = EmbedderFactory.from_config(config.embedder) + self.graph_store: PolarDBGraphDB = GraphStoreFactory.from_config(config.graph_db) + + self.is_reorganize = config.reorganize + self.memory_manager: MemoryManager = MemoryManager( + self.graph_store, + self.embedder, + self.llm, + memory_size=config.memory_size + or { + "WorkingMemory": 20, + "LongTermMemory": 1500, + "UserMemory": 480, + }, + is_reorganize=self.is_reorganize, + ) + + def _feedback_judgement( + self, chat_history: list[MessageDict], feedback_content: str, feedback_time: str = "" + ) -> dict | None: + """ + Generate a judgement for a given feedback. + """ + lang = detect_lang(feedback_content) + template = FEEDBACK_PROMPT_DICT["judge"][lang] + chat_history_str = str(chat_history[-4:]) + prompt = ( + template.replace("{chat_history}", chat_history_str) + .replace("{user_feedback}", feedback_content) + .replace("{feedback_time}", feedback_time) + ) + judge_res = self._get_llm_response(prompt) + return judge_res if judge_res else [] + + def _feedback_memory( + self, user_name: str, feedback_memories: list[TextualMemoryItem], **kwargs + ) -> dict: + sync_mode = kwargs.get("sync_mode") + + def _add_or_update(memory_item: TextualMemoryItem): + current_memories = self._vec_query(memory_item.metadata.embedding, user_name=user_name) + if current_memories: + lang = detect_lang("".join(memory_item.memory)) + template = FEEDBACK_PROMPT_DICT["compare"][lang] + prompt = template.replace("{current_memories}", str(current_memories)).replace( + "{new_facts}", memory_item.memory + ) + operations = self._get_llm_response(prompt).get("operation", {}) + else: + operations = {"event": "ADD"} + logger.info(f"[Feedback memory operations]: {operations!s}") + + if operations and operations["event"].lower() == "add": + added_ids = self.memory_manager.add( + [memory_item], user_name=user_name, mode=sync_mode + ) + logger.info(f"[Memory Feedback ADD] {added_ids!s}") + + return { + "record": { + "add": [{"id": added_ids[0], "text": memory_item.memory}], + "update": [], + } + } + elif operations and operations["event"].lower() == "update": + to_update_id = operations["id"] + updated_ids = self.memory_manager.update( + [to_update_id], [memory_item], user_name=user_name, mode=sync_mode + ) + log_update_info = operations["old_memory"] + " >> " + operations["text"] + logger.info(f"[Memory Feedback UPDATE] {updated_ids}, info: {log_update_info}") + + return { + "record": { + "add": [], + "update": [ + { + "id": to_update_id, + "origin_memory": operations["old_memory"], + "text": operations["text"], + } + ], + } + } + else: + return {"record": {"add": [], "update": []}} + + search_filter = {"user_name": user_name} + with ContextThreadPoolExecutor(max_workers=8) as ex: + futures = { + ex.submit(_add_or_update, mem, search_filter): i + for i, mem in enumerate(feedback_memories) + } + results = [None] * len(futures) + for fut in concurrent.futures.as_completed(futures): + i = futures[fut] + try: + node = fut.result() + if node: + results[i] = node + except Exception as e: + logger.error(f"[FeedBack] error: {e}") + mem_res = [r for r in results if r] + + return { + "record": { + "add": [element for item in mem_res for element in item["record"]["add"]], + "update": [element for item in mem_res for element in item["record"]["update"]], + } + } + + def _vec_query(self, new_memories_embedding: list[float], user_name=None): + retrieved_ids = self.graph_store.search_by_embedding( + new_memories_embedding, user_name=user_name + ) + current_memories = [self.graph_store.get_node(item["id"]) for item in retrieved_ids] + + return [ + { + "id": item["id"], + "text": item["memory"], + } + for item in current_memories + ] + + def _get_llm_response(self, prompt: str, dsl: bool = True) -> dict: + messages = [{"role": "user", "content": prompt}] + try: + response_text = self.llm.generate(messages) + if dsl: + response_json = json.loads(response_text) + else: + return response_text + except Exception as e: + logger.error(f"[LLM] Exception during chat generation: {e}") + response_json = None + return response_json + + def _generate_answer( + self, chat_history: list[MessageDict], feedback_content: str, corrected_answer: bool + ) -> str: + """ + Answer generation to facilitate concurrent submission. + """ + if not corrected_answer: + return "" + lang = detect_lang(feedback_content) + template = FEEDBACK_PROMPT_DICT["generation"][lang] + chat_history_str = "\n".join( + [f"{item['role']}: {item['content']}" for item in chat_history] + ) + chat_history_str = chat_history_str if chat_history_str else "none" + prompt = template.replace("{chat_history}", chat_history_str).replace( + "{question}", feedback_content + ) + return self._get_llm_response(prompt, dsl=False) + + def process_feedback_core( + self, + user_name: str, + chat_history: list[MessageDict], + feedback_content: str, + mem_reader: SimpleStructMemReader | None = None, + **kwargs, + ) -> dict: + """ + Core feedback processing: judgment, memory extraction, addition/update. Return record. + """ + try: + feedback_time = kwargs.get("feedback_time") or datetime.now().isoformat() + session_id = kwargs.get("session_id") + allow_knowledgebase_write = bool(kwargs.get("allow_knowledgebase_write")) + if not allow_knowledgebase_write: + return {"record": {"add": [], "update": []}} + + info = {"user_id": user_name, "session_id": session_id} + logger.info(f"[Feedback Core] Starting memory feedback process for user {user_name}") + + if mem_reader and not chat_history: + scene_data = [ + [{"role": "user", "content": feedback_content, "chat_time": feedback_time}] + ] + memories = mem_reader.get_memory(scene_data, type="chat", info=info) + to_add_memories = [item for scene in memories for item in scene] + added_ids = self.memory_manager.add(to_add_memories, user_name=user_name) + logger.info( + f"[Feedback Core] Added {len(added_ids)} memories for user {user_name}." + ) + return { + "record": { + "add": [ + {"id": _id, "text": added_mem.memory} + for _id, added_mem in zip(added_ids, to_add_memories, strict=False) + ], + "update": [], + } + } + + elif chat_history: + raw_judge = self._feedback_judgement( + chat_history, feedback_content, feedback_time=feedback_time + ) + judge_res = ( + [ + item + for item in raw_judge + if item["validity"].lower() == "true" and item["corrected_info"].strip() + ] + if raw_judge + else [] + ) + if not judge_res: + logger.warning( + f"[Feedback Core] No valid judgements for user {user_name}: {raw_judge}." + ) + return {"record": {"add": [], "update": []}} + + feedback_memories = [] + feedback_memories_embeddings = self.embedder.embed( + [item["corrected_info"] for item in judge_res] + ) + for item, embedding in zip(judge_res, feedback_memories_embeddings, strict=False): + value = item["corrected_info"] + key = item["key"] + tags = item["tags"] + feedback_memories.append( + TextualMemoryItem( + memory=value, + metadata=TreeNodeTextualMemoryMetadata( + user_id=info.get("user_id", ""), + session_id=info.get("session_id", ""), + memory_type="LongTermMemory", + status="activated", + tags=tags, + key=key, + embedding=embedding, + usage=[], + sources=[{"type": "chat"}], + background="", + confidence=0.99, + type="fine", + ), + ) + ) + mem_record = self._feedback_memory(user_name, feedback_memories, **kwargs) + logger.info( + f"[Feedback Core] Processed {len(feedback_memories)} feedback memories for user {user_name}." + ) + return mem_record + + else: + logger.info("[Feedback Core] Empty chat_history and no mem_reader, skipping.") + return {"record": {"add": [], "update": []}} + + except Exception as e: + logger.error(f"[Feedback Core] Error for user {user_name}: {e}") + return {"record": {"add": [], "update": []}} + + def process_feedback( + self, + user_name: str, + chat_history: list[MessageDict], + feedback_content: str, + mem_reader: SimpleStructMemReader | None = None, + **kwargs, + ): + """ + Process feedback with different modes. + + Args: + user_name: User identifier + chat_history: List of chat messages + feedback_content: Feedback content from user + mem_reader: Memory reader instance + **kwargs: Additional arguments including sync_mode + + Returns: + Dict with answer and/or memory operation records + """ + sync_mode = kwargs.get("sync_mode") + corrected_answer = kwargs.get("corrected_answer") + + if sync_mode == "sync": + with ContextThreadPoolExecutor(max_workers=2) as ex: + answer_future = ex.submit( + self._generate_answer, + chat_history, + feedback_content, + corrected_answer=corrected_answer, + ) + core_future = ex.submit( + self.process_feedback_core, + user_name, + chat_history, + feedback_content, + mem_reader, + **kwargs, + ) + concurrent.futures.wait([answer_future, core_future]) + try: + answer = answer_future.result() + record = core_future.result() + logger.info( + f"[process_feedback sync] Completed concurrently for user {user_name} with full results." + ) + return {"answer": answer, "record": record["record"]} + except Exception as e: + logger.error( + f"[process_feedback sync] Error in concurrent tasks for {user_name}: {e}" + ) + return {"answer": "", "record": {"add": [], "update": []}} + else: + answer = self._generate_answer( + chat_history, feedback_content, corrected_answer=corrected_answer + ) + + ex = ContextThreadPoolExecutor(max_workers=1) + future = ex.submit( + self.process_feedback_core, + user_name, + chat_history, + feedback_content, + mem_reader, + **kwargs, + ) + ex.shutdown(wait=False) + + def log_completion(f): + try: + result = f.result() + logger.info(f"[Background Feedback] Completed for {user_name}: {result}") + except Exception as e: + logger.error(f"[Background Feedback] Error for {user_name}: {e}") + + future.add_done_callback(log_completion) + + logger.info( + f"[process_feedback async] Returned answer, background task started for {user_name}." + ) + return {"answer": answer, "record": {"add": [], "update": []}} diff --git a/src/memos/mem_feedback/simple_feedback.py b/src/memos/mem_feedback/simple_feedback.py new file mode 100644 index 000000000..23e25ff2c --- /dev/null +++ b/src/memos/mem_feedback/simple_feedback.py @@ -0,0 +1,23 @@ +from memos import log +from memos.embedders.factory import OllamaEmbedder +from memos.graph_dbs.factory import PolarDBGraphDB +from memos.llms.factory import AzureLLM, OllamaLLM, OpenAILLM +from memos.mem_feedback.feedback import MemFeedback +from memos.memories.textual.tree_text_memory.organize.manager import MemoryManager + + +logger = log.get_logger(__name__) + + +class SimpleMemFeedback(MemFeedback): + def __init__( + self, + llm: OpenAILLM | OllamaLLM | AzureLLM, + embedder: OllamaEmbedder, + graph_store: PolarDBGraphDB, + memory_manager: MemoryManager, + ): + self.llm = llm + self.embedder = embedder + self.graph_store = graph_store + self.memory_manager = memory_manager diff --git a/src/memos/memories/textual/tree_text_memory/organize/manager.py b/src/memos/memories/textual/tree_text_memory/organize/manager.py index a71fee02f..dadf7c321 100644 --- a/src/memos/memories/textual/tree_text_memory/organize/manager.py +++ b/src/memos/memories/textual/tree_text_memory/organize/manager.py @@ -115,6 +115,26 @@ def add( self._refresh_memory_size(user_name=user_name) return added_ids + def update( + self, + memories_ids: list[str], + memories: list[TextualMemoryItem], + user_name: str | None = None, + mode: str = "sync", + ) -> list[str]: + for _id, memory in zip(memories_ids, memories, strict=False): + field = { + "memory": memory.memory, + "key": memory.metadata.key, + "tags": memory.metadata.tags, + "embedding": memory.metadata.embedding, + } + self.graph_store.update_node(_id, field, user_name) + + graph_class_name = self.graph_store.__class__.__name__ + logger.info(f"Updated {graph_class_name}") + return memories_ids + def replace_working_memory( self, memories: list[TextualMemoryItem], user_name: str | None = None ) -> None: diff --git a/src/memos/templates/mem_feedback_prompts.py b/src/memos/templates/mem_feedback_prompts.py new file mode 100644 index 000000000..2a9a7dcb2 --- /dev/null +++ b/src/memos/templates/mem_feedback_prompts.py @@ -0,0 +1,593 @@ +FEEDBACK_JUDGEMENT_PROMPT = """You are a answer quality analysis expert. Please strictly follow the steps and criteria below to analyze the provided "User and Assistant Chat History" and "User Feedback," and fill the final evaluation results into the specified JSON format. + +Analysis Steps and Criteria: +1. *Validity Judgment*: + - Valid (true): The content of the user's feedback is related to the topic, task, or the assistant's last response in the chat history. For example: asking follow-up questions, making corrections, providing supplements, or evaluating the last response. + - Invalid (false): The user’s feedback is entirely unrelated to the conversation history, with no semantic, topical, or lexical connection to any prior content. + +2. *User Attitude Judgment*: + - Dissatisfied: The feedback shows negative emotions, such as directly pointing out errors, expressing confusion, complaining, criticizing, or explicitly stating that the problem remains unsolved. + - Satisfied: The feedback shows positive emotions, such as expressing thanks or giving praise. + - Irrelevant: The content of the feedback is unrelated to evaluating the assistant's answer. + +3. *Assistant Response Effectiveness Type Judgment*: + - Wrong: The assistant provided incorrect information. + - Missing: The assistant's response was correct in direction but incomplete, omitting key details. + - None: The user feedback does not point to any shortcomings in the assistant's response. + +4. *Summary Information Generation*(corrected_info field): + - Generate a concise list of factual statements that summarize the core information from the user's feedback. + — Focus on objective facts, corrections, or confirmations. + - Express time information as concrete, unambiguous date(s) or period(s) (e.g., “March 2023”, “2024-07”, or “May–June 2022”). + - For 'Satisfied' or 'None' types, this list may contain confirming statements or be empty if no new facts are provided. + - For example: "The user completed the Everest Circuit trek with colleagues in March 2023." + +Output Format: +[ + { + "validity": "", + "user_attitude": "", + "error_type": "", + "corrected_info": "", + "key": , + "tags": + }, + ... +] + +Example1: + +Dialogue History: +user: I can't eat spicy food these days. Can you recommend some suitable restaurants for me? +assistant: Sure, I recommend the Fish Restaurant near you. Their signature dishes include various types of steamed seafood and sashimi of sea fish. +feedback time: 2023-1-18T14:25:00.856481 + +User Feedback: +Oh，No！I'm allergic to seafood！And I don't like eating raw fish. + +Output: +[ + { + "validity": "true", + "user_attitude": "dissatisfied", + "error_type": "wrong", + "corrected_info": "User is allergic to seafood", + "key": "allergic to seafood", + "tags": ["allergic", "seafood"] + }, + { + "validity": "true", + "user_attitude": "dissatisfied", + "error_type": "wrong", + "corrected_info": "User does not like eating raw fish.", + "key": "dislike eating raw fish.", + "tags": ["dislike", "raw fish"] + } +] + +Example2: + +Dialogue History: +user: When did Jhon graduated? +assistant: 2014 +feedback time: 2025-11-18T20:45:00.875249 + +User Feedback: +Wrong. He graduated the following year. + +Output: +[ + { + "validity": "true", + "user_attitude": "dissatisfied", + "error_type": "wrong", + "corrected_info": "Jhon was graduated at 2015", + "key": "Jhon graduated time", + "tags": ["Jhon", "graduated", "year"] + } +] +Dialogue History: +{chat_history} +feedback time: {feedback_time} + +User Feedback: +{user_feedback} + +Output: +""" + + +FEEDBACK_JUDGEMENT_PROMPT_ZH = """你是一个对话质量分析专家。请严格根据以下步骤和标准，对提供的“用户和助理的对话历史”和“用户反馈”进行分析，并将最终判定结果填入指定的JSON格式中。 + +分析步骤与判定标准: +1. *有效性判定* + - 有效(true)：用户反馈的内容与对话历史的主题、任务或上一次助理的回答*相关*。例如：针对回答进行追问、纠正、补充或评价。 + - 无效(false)：用户的反馈与对话历史*完全无关*，与任何先前内容之间不存在语义、主题或词汇上的联系。 +2. *用户态度判定* + - 不满意(dissatisfied)：反馈中表现出负面情绪，如直接指出错误、表达困惑、抱怨、批评，或明确表示问题未解决。 + - 满意(satisfied)：反馈中表现出正面情绪，如表示感谢或给予称赞。 + - 无关(irrelevant)：反馈内容与评价助理回答无关。 +3. *助理回答效果类型判定* + - 错误(wrong)：助理提供了不正确的信息。 + - 缺漏(missing)：助理的回答方向正确但不完整，遗漏了关键细节。 + - 无(none)：用户反馈并未指向助理回答的任何不足。 +4. *总结信息生成* + - 生成一份简洁的事实陈述列表，该列表概括了用户反馈中的核心信息。 + - 重点放在客观事实、更正或确认上。 + - 对于“满意”或“无”类型的反馈，该列表可能包含确认性的陈述，或者如果未提供新事实，则可能为空。 + - 例如：“用户在2023年3月与同事完成了珠峰环线徒步旅行。” + +输出格式： +[ + { + "validity": <字符串，"true" 或 "false">, + "user_attitude": <字符串，"dissatisfied" 或 "satisfied" 或 "irrelevant">, + "error_type": <字符串，"wrong" 或 "missing" 或 "irrelevant">, + "corrected_info": <字符串，中文书写正确的信息记录>, + "key": <字符串，唯一且简洁的记忆标题>, + "tags": <相关主题关键词列表（例如，["截止日期", "团队", "计划"]）> + }, + ... +] + +示例： + +用户和助理的对话历史： +user: 这两天我吃不了辣椒，给我推荐一些适合的餐厅吧。 +assistant: 好的，推荐您附近的新荣记餐厅，黄鱼年糕以及各类清蒸海鲜是这件餐厅的招牌菜。 +反馈时间：2023-1-18T14:25:00.856481 + +用户反馈： +你忘记我海鲜过敏这件事了吗？而且我不喜欢年糕的口感。 + +输出： +[ + { + "validity": "true", + "user_attitude": "dissatisfied", + "error_type": "wrong", + "corrected_info": "用户对海鲜过敏。", + "key": "海鲜过敏", + "tags": ["海鲜", "过敏"] + }, + { + "validity": "true", + "user_attitude": "dissatisfied", + "error_type": "wrong", + "corrected_info": "用户不喜欢年糕的口感。", + "key": "不喜欢年糕", + "tags": ["不喜欢年糕", "年糕", "口感"] + } +] + + +用户和助理的对话历史： +{chat_history} +反馈时间：{feedback_time} + +用户反馈： +{user_feedback} + +输出： +""" + + +UPDATE_FORMER_MEMORIES = """Please analyze the newly acquired factual information and determine how this information should be updated to the memory database: add or update, and provide final operation recommendations. + +You must strictly return the response in the following JSON format: + +{ + "operation": + { + "id": "", + "text": "", + "event": "", + "old_memory": "" + } +} + +*Requirements*: +1. If the new fact adds no supplemental value and the existing memory supersedes it, no operation is performed. +2. If the new fact is similar to existing memory but the information is more accurate, complete, or requires correction, set operation to "UPDATE" +3. If the new fact contradicts existing memory in key information (such as time, location, status, etc.), update the original memory based on the new fact and set operation to "UPDATE" +4. If there is completely new information to add, set operation to "ADD" + +*ID Management Rules*: +- Update operation: Keep the original ID unchanged +- Add operation: Generate a new unique ID in the format of a 4-digit string (e.g., "0001", "0002", etc.) + +*Important Requirements*: +- Return only the JSON format response, without any other content +- For update operations, you must provide the old_memory field to show the original content +- text field requirements: Use concise, complete declarative sentences that are consistent with the newly acquired factual information, avoiding redundant information +- text and old_memory content should be in English + +Example1: +Current Memories: +{ + "memory": [ + { + "id": "123", + "text": "The user works as a software engineer in Company A, mainly responsible for front-end development" + }, + { + "id": "908", + "text": "The user likes to go fishing with friends on weekends" + } + ] +} + +Newly facts: +"The user is currently working as a senior full-stack development engineer at Company B" + +Operation recommendations: +{ + "operation": + { + "id": "123", + "text": "The user is currently working as a senior full-stack development engineer at Company B", + "event": "UPDATE", + "old_memory": "The user works as a software engineer in Company A, mainly responsible for front-end development" + } +} + +Example2: +Current Memories: +{ + "memory": [ + { + "id": "123", + "text": "The user works as a software engineer in Company A, mainly responsible for front-end development" + }, + { + "id": "908", + "text": "The user likes to go fishing with friends on weekends" + } + ] +} + +Newly facts: +"The user's residential address is Mingyue Community, Chaoyang District, Beijing" + +Operation recommendations: +{ + "operation": + { + "id": "4567", + "text": "The user's residential address is Mingyue Community, Chaoyang District, Beijing", + "event": "ADD" + } +} + +Current Memories +{current_memories} + +Newly facts: +{new_facts} + +Operation recommendations: +""" + + +UPDATE_FORMER_MEMORIES_ZH = """请分析新获取的事实信息，并决定该信息应该如何更新到记忆库中：新增或更新，并给出最终的操作建议。 + +你必须严格按照以下JSON格式返回响应： + +{ + "operation": + { + "id": "<记忆ID>", + "text": "<记忆内容>", + "event": "<操作类型，必须是 "ADD", "UPDATE"之一>", + "old_memory": "<原记忆内容，仅当操作为"UPDATE"时需要提供>" + } +} + +要求： +1. 如果新事实对现有记忆没有额外补充，现有记忆的信息可以覆盖新事实，则不设置任何操作 +2. 如果新事实与现有记忆相似但信息更准确、完整或需要修正，设置操作为"UPDATE" +3. 如果新事实与现有记忆在关键信息上矛盾（如时间、地点、状态等），以新事实为准更新原有记忆，设置操作为"UPDATE" +4. 如果有全新信息添加，设置操作为"ADD" + +ID管理规则： +- 更新操作：保持原有ID不变 +- 新增操作：生成新的唯一ID，格式为4位数字字符串（如："0001", "0002"等） + +重要要求： +- 只返回JSON格式的响应，不要包含其他任何内容 +- 对于更新操作，必须提供old_memory字段显示原内容 +- text字段要求：使用简洁、完整的陈述句，和新获取的事实信息一致，避免冗余信息 +- text和old_memory内容使用中文 + +示例1： +现有记忆记录： +{ + "memory": [ + { + "id": "123", + "text": "用户在公司A担任软件工程师，主要负责前端开发" + }, + { + "id": "908", + "text": "用户周末喜欢和朋友一起钓鱼" + } + ] +} + +新获取的事实： +"用户现在在公司B担任高级全栈开发工程师" + +操作建议： +{ + "operation": + { + "id": "123", + "text": "用户在公司B担任高级全栈开发工程师", + "event": "UPDATE", + "old_memory": "用户在公司A担任软件工程师，主要负责前端开发" + } +} + +示例2： +现有记忆记录： +{ + "memory": [ + { + "id": "123", + "text": "用户在公司A担任软件工程师，主要负责前端开发" + }, + { + "id": "908", + "text": "用户周末喜欢和朋友一起钓鱼" + } + ] +} + +新获取的事实： +"用户的居住地址是北京市朝阳区明月小区" + +操作建议： +{ + "operation": + { + "id": "4567", + "text": "用户的居住地址是北京市朝阳区明月小区", + "event": "ADD" + } +} + +现有记忆记录： +{current_memories} + +新获取的事实： +{new_facts} + +操作建议： +""" + + +GROUP_UPDATE_FORMER_MEMORIES = """Please analyze the newly acquired factual information and determine how this information should be updated to the memory database: add, update, or keep unchanged, and provide final operation recommendations. + +You must strictly return the response in the following JSON format: + +{ + "operation": [ + { + "id": "", + "text": "", + "event": "", + "old_memory": "" + }, + ... + ] +} + +*Requirements*: +1. If the new fact provides no additional supplement to existing memory, set operation to "NONE" +2. If the new fact is similar to existing memory but the information is more accurate, complete, or requires correction, set operation to "UPDATE" +3. If the new fact contradicts existing memory in key information (such as time, location, status, etc.), update the original memory based on the new fact and set operation to "UPDATE" +4. If there is completely new information to add, set operation to "ADD" + +*ID Management Rules*: +- Update operation: Keep the original ID unchanged +- Add operation: Generate a new unique ID in the format of a 4-digit string (e.g., "0001", "0002", etc.) + +*Important Requirements*: +- Return only the JSON format response, without any other content +- For update operations, you must provide the old_memory field to show the original content +- text field requirements: Use concise, complete declarative sentences that are consistent with the newly acquired factual information, avoiding redundant information +- text and old_memory content should be in English + +Example: +Current Memories: +{ + "memory": [ + { + "id": "123", + "text": "The user works as a software engineer in Company A, mainly responsible for front-end development" + }, + { + "id": "908", + "text": "The user likes to go fishing with friends on weekends" + } + ] +} + +Newly facts: +["The user is currently working as a senior full-stack development engineer at Company B", "The user's residential address is Mingyue Community, Chaoyang District, Beijing", "The user goes fishing on weekends"] + +Operation recommendations: +{ + "operation": [ + { + "id": "123", + "text": "The user is currently working as a senior full-stack development engineer at Company B", + "event": "UPDATE", + "old_memory": "The user works as a software engineer in Company A, mainly responsible for front-end development" + }, + { + "id": "4567", + "text": "The user's residential address is Mingyue Community, Chaoyang District, Beijing", + "event": "ADD" + }, + { + "id": "908", + "text": "The user likes to go fishing with friends on weekends", + "event": "NONE" + } + ] +} + +Current Memories +{current_memories} + +Newly facts: +{new_facts} + +Operation recommendations: +""" + + +GROUP_UPDATE_FORMER_MEMORIES_ZH = """请分析新获取的事实信息，并决定这些信息应该如何更新到记忆库中：新增、更新、或保持不变，并给出最终的操作建议。 + +你必须严格按照以下JSON格式返回响应： + +{ + "operation": [ + { + "id": "<记忆ID>", + "text": "<记忆内容>", + "event": "<操作类型，必须是 "ADD", "UPDATE", "NONE" 之一>", + "old_memory": "<原记忆内容，仅当操作为"UPDATE"时需要提供>" + }, + ... + ] +} + +要求： +1. 如果新事实对现有记忆没有额外补充，设置操作为"NONE" +2. 如果新事实与现有记忆相似但信息更准确、完整或需要修正，设置操作为"UPDATE" +3. 如果新事实与现有记忆在关键信息上矛盾（如时间、地点、状态等），以新事实为准更新原有记忆，设置操作为"UPDATE" +4. 如果有全新信息添加，设置操作为"ADD" + +ID管理规则： +- 更新操作：保持原有ID不变 +- 新增操作：生成新的唯一ID，格式为4位数字字符串（如："0001", "0002"等） + +重要要求： +- 只返回JSON格式的响应，不要包含其他任何内容 +- 对于更新操作，必须提供old_memory字段显示原内容 +- text字段要求：使用简洁、完整的陈述句，和新获取的事实信息一致，避免冗余信息 +- text和old_memory内容使用中文 + +示例： +现有记忆记录： +{ + "memory": [ + { + "id": "123", + "text": "用户在公司A担任软件工程师，主要负责前端开发" + }, + { + "id": "908", + "text": "用户周末喜欢和朋友一起钓鱼" + } + ] +} + +新获取的事实： +["用户现在在公司B担任高级全栈开发工程师", "用户的居住地址是北京市朝阳区明月小区", "用户在周末会去钓鱼"] + +操作建议： +{ + "operation": [ + { + "id": "123", + "text": "用户在公司B担任高级全栈开发工程师", + "event": "UPDATE", + "old_memory": "用户在公司A担任软件工程师，主要负责前端开发" + }, + { + "id": "4567", + "text": "用户的居住地址是北京市朝阳区明月小区", + "event": "ADD" + }, + { + "id": "908", + "text": "用户周末喜欢和朋友一起钓鱼", + "event": "NONE" + } + ] +} + +现有记忆记录： +{current_memories} + +新获取的事实： +{new_facts} + +操作建议： +""" + + +FEEDBACK_ANSWER_PROMPT = """ +You are a knowledgeable and helpful AI assistant.You have access to the history of the current conversation. This history contains the previous exchanges between you and the user. + +# INSTRUCTIONS: +1. Carefully analyze the entire conversation history. Your answer must be based only on the information that has been exchanged within this dialogue. +2. Pay close attention to the sequence of the conversation. If the user refers back to a previous statement (e.g., "the thing I mentioned earlier"), you must identify that specific point in the history. +3. Your primary goal is to provide continuity and context from this specific conversation. Do not introduce new facts or topics that have not been previously discussed. +4. If current question is ambiguous, use the conversation history to clarify its meaning. + +# APPROACH (Think step by step): +1. Review the conversation history to understand the context and topics that have been discussed. +2. Identify any specific details, preferences, or statements the user has made that are relevant to the current question. +3. Formulate a precise, concise answer that is a direct continuation of the existing dialogue. +4. Ensure your final answer is grounded in the conversation history and directly addresses the user's latest query in that context. + +# Tip: +If no chat history is provided: + - Treat the query as self-contained. + - Do not assume prior context. + - Respond based solely on the current question. + - Do not raise new questions during the answering process. + +Chat history: +{chat_history} + +Question: +{question} + +Answer: +""" + +FEEDBACK_ANSWER_PROMPT_ZH = """ +你是一个知识渊博且乐于助人的AI助手。你可以访问当前对话的完整历史记录。这些记录包含你与用户之间先前的所有交流内容。 + +# 指令： +1. 仔细分析整个对话历史。你的回答必须仅基于本次对话中已交流的信息。 +2. 密切关注对话的先后顺序。如果用户提及之前的发言（例如“我之前提到的那件事”），你必须定位到历史记录中的具体内容。 +3. 你的主要目标是基于本次特定对话提供连续性和上下文。不要引入之前对话中未讨论过的新事实或话题。 +4. 如果用户当前的问题含义不明确，请利用对话历史来澄清其意图。 + +# 处理方法（逐步思考）： +1. 回顾对话历史，以理解已讨论的背景和主题。 +2. 识别用户已提及的、与当前问题相关的任何具体细节、偏好或陈述。 +3. 构思一个精准、简洁的回答，使其成为现有对话的直接延续。 +4. 确保你的最终回答紧扣对话历史，并在此上下文中直接回应用户的最新提问。 + +# 注意: +如果没有提供聊天历史记录： + - 将该查询视为独立的。 + - 不要假设之前存在背景信息。 + - 仅根据当前问题进行回答。 + - 在回答过程中不必提出新的问题。 + +对话历史： +{chat_history} + +问题： +{question} + +回答： +""" From 4d712eb5840f092853aa08417ed55d2b41883a48 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Thu, 20 Nov 2025 19:09:23 +0800 Subject: [PATCH 21/72] feedback handler configuration --- src/memos/api/handlers/base_handler.py | 7 + src/memos/api/handlers/component_init.py | 11 + src/memos/api/handlers/feedback_handler.py | 52 +++++ src/memos/api/product_models.py | 5 +- src/memos/api/routers/server_router.py | 19 +- src/memos/configs/memory.py | 9 +- src/memos/graph_dbs/polardb.py | 10 +- src/memos/mem_feedback/feedback.py | 226 +++++++++++++------- src/memos/mem_feedback/simple_feedback.py | 3 + src/memos/templates/mem_feedback_prompts.py | 195 +++++++++++++---- 10 files changed, 411 insertions(+), 126 deletions(-) create mode 100644 src/memos/api/handlers/feedback_handler.py diff --git a/src/memos/api/handlers/base_handler.py b/src/memos/api/handlers/base_handler.py index a686ac8f9..0b2fc5ae3 100644 --- a/src/memos/api/handlers/base_handler.py +++ b/src/memos/api/handlers/base_handler.py @@ -37,6 +37,7 @@ def __init__( internet_retriever: Any | None = None, memory_manager: Any | None = None, mos_server: Any | None = None, + feedback_server: Any | None = None, **kwargs, ): """ @@ -68,6 +69,7 @@ def __init__( self.internet_retriever = internet_retriever self.memory_manager = memory_manager self.mos_server = mos_server + self.feedback_server = feedback_server # Store any additional dependencies for key, value in kwargs.items(): @@ -161,6 +163,11 @@ def mos_server(self): """Get MOS server instance.""" return self.deps.mos_server + @property + def feedback_server(self): + """Get feedback server instance.""" + return self.deps.feedback_server + def _validate_dependencies(self, *required_deps: str) -> None: """ Validate that required dependencies are available. diff --git a/src/memos/api/handlers/component_init.py b/src/memos/api/handlers/component_init.py index 78ed13e1f..56ec08152 100644 --- a/src/memos/api/handlers/component_init.py +++ b/src/memos/api/handlers/component_init.py @@ -28,6 +28,7 @@ from memos.llms.factory import LLMFactory from memos.log import get_logger from memos.mem_cube.navie import NaiveMemCube +from memos.mem_feedback.simple_feedback import SimpleMemFeedback from memos.mem_os.product_server import MOSServer from memos.mem_reader.factory import MemReaderFactory from memos.mem_scheduler.orm_modules.base_model import BaseDBManager @@ -251,6 +252,15 @@ def init_server() -> dict[str, Any]: online_bot = get_online_bot_function() if dingding_enabled else None logger.info("DingDing bot is enabled") + # Initialize feedback server + feedback_server = SimpleMemFeedback( + llm=llm, + embedder=embedder, + graph_store=graph_db, + memory_manager=memory_manager, + mem_reader=mem_reader, + ) + # Return all components as a dictionary for easy access and extension return { "graph_db": graph_db, @@ -273,4 +283,5 @@ def init_server() -> dict[str, Any]: "text_mem": text_mem, "pref_mem": pref_mem, "online_bot": online_bot, + "feedback_server": feedback_server, } diff --git a/src/memos/api/handlers/feedback_handler.py b/src/memos/api/handlers/feedback_handler.py new file mode 100644 index 000000000..2e8a21cb1 --- /dev/null +++ b/src/memos/api/handlers/feedback_handler.py @@ -0,0 +1,52 @@ +""" +Feeback handler for memory add/update functionality. +""" + +from memos.api.handlers.base_handler import BaseHandler, HandlerDependencies +from memos.api.product_models import APIFeedbackRequest, MemoryResponse +from memos.log import get_logger + + +logger = get_logger(__name__) + + +class FeedbackHandler(BaseHandler): + """ + Handler for memory feedback operations. + + Provides fast, fine-grained, and mixture-based feedback modes. + """ + + def __init__(self, dependencies: HandlerDependencies): + """ + Initialize feedback handler. + + Args: + dependencies: HandlerDependencies instance + """ + super().__init__(dependencies) + self._validate_dependencies("feedback_server", "mem_reader") + + def handle_feedback_memories(self, feedback_req: APIFeedbackRequest) -> MemoryResponse: + """ + Main handler for feedback memories endpoint. + + Args: + feedback_req: feedback request containing content and parameters + + Returns: + MemoryResponse with formatted results + """ + process_record = self.feedback_server.process_feedback( + user_name=feedback_req.mem_cube_id, + session_id=feedback_req.session_id, + chat_history=feedback_req.history, + retrieved_memory_ids=feedback_req.retrieved_memory_ids, + feedback_content=feedback_req.feedback_content, + feedback_time=feedback_req.feedback_time, + allow_knowledgebase_write=feedback_req.allow_knowledgebase_write, + sync_mode=feedback_req.sync_mode, + corrected_answer=feedback_req.corrected_answer, + ) + + return MemoryResponse(message="Feedback process successfully", data=[process_record]) diff --git a/src/memos/api/product_models.py b/src/memos/api/product_models.py index 7d6373374..3e70d522a 100644 --- a/src/memos/api/product_models.py +++ b/src/memos/api/product_models.py @@ -216,7 +216,10 @@ class APIFeedbackRequest(BaseRequest): ) agent_id: str = Field(None, description="Agent ID") app_id: str = Field(None, description="App ID") - chat_history: list[MessageDict] | None = Field(..., description="Chat history") + history: list[MessageDict] | None = Field(..., description="Chat history") + retrieved_memory_ids: list[str] | None = Field( + None, description="Retrieved memory ids at last turn" + ) feedback_content: str | None = Field(..., description="Feedback content to process") feedback_time: str | None = Field(None, description="Feedback time") allow_public: bool = Field( diff --git a/src/memos/api/routers/server_router.py b/src/memos/api/routers/server_router.py index b3b517305..fbe0f0df7 100644 --- a/src/memos/api/routers/server_router.py +++ b/src/memos/api/routers/server_router.py @@ -21,10 +21,12 @@ from memos.api.handlers.add_handler import AddHandler from memos.api.handlers.base_handler import HandlerDependencies from memos.api.handlers.chat_handler import ChatHandler +from memos.api.handlers.feedback_handler import FeedbackHandler from memos.api.handlers.search_handler import SearchHandler from memos.api.product_models import ( APIADDRequest, APIChatCompleteRequest, + APIFeedbackRequest, APISearchRequest, ChatRequest, GetMemoryRequest, @@ -56,7 +58,7 @@ chat_handler = ChatHandler( dependencies, search_handler, add_handler, online_bot=components.get("online_bot") ) - +feedback_handler = FeedbackHandler(dependencies) # Extract commonly used components for function-based handlers # (These can be accessed from the components dict without unpacking all of them) mem_scheduler: BaseScheduler = components["mem_scheduler"] @@ -219,3 +221,18 @@ def get_all_memories(memory_req: GetMemoryRequest): memory_type=memory_req.memory_type or "text_mem", naive_mem_cube=naive_mem_cube, ) + + +# ============================================================================= +# Feedback API Endpoints +# ============================================================================= + + +@router.post("/feedback", summary="Feedback memories", response_model=MemoryResponse) +def feedback_memories(feedback_req: APIFeedbackRequest): + """ + Feedback memories for a specific user. + + This endpoint uses the class-based FeedbackHandler for better code organization. + """ + return feedback_handler.handle_feedback_memories(feedback_req) diff --git a/src/memos/configs/memory.py b/src/memos/configs/memory.py index 9caf4fa3e..04fc58ad6 100644 --- a/src/memos/configs/memory.py +++ b/src/memos/configs/memory.py @@ -7,6 +7,7 @@ from memos.configs.graph_db import GraphDBConfigFactory from memos.configs.internet_retriever import InternetRetrieverConfigFactory from memos.configs.llm import LLMConfigFactory +from memos.configs.mem_reader import MemReaderConfigFactory from memos.configs.reranker import RerankerConfigFactory from memos.configs.vec_db import VectorDBConfigFactory from memos.exceptions import ConfigurationError @@ -240,7 +241,7 @@ class PreferenceTextMemoryConfig(BaseTextMemoryConfig): ) -class MemFeedbackConfig(BaseTextMemoryConfig): +class MemFeedbackConfig(BaseMemoryConfig): """Memory feedback configuration class.""" extractor_llm: LLMConfigFactory = Field( @@ -275,6 +276,12 @@ class MemFeedbackConfig(BaseTextMemoryConfig): ), ) + mem_reader: MemReaderConfigFactory = Field( + ..., + default_factory=MemReaderConfigFactory, + description="MemReader configuration for the Feedback", + ) + # ─── 3. Global Memory Config Factory ────────────────────────────────────────── diff --git a/src/memos/graph_dbs/polardb.py b/src/memos/graph_dbs/polardb.py index 60998c172..74bce3b97 100644 --- a/src/memos/graph_dbs/polardb.py +++ b/src/memos/graph_dbs/polardb.py @@ -524,13 +524,19 @@ def update_node(self, id: str, fields: dict[str, Any], user_name: str | None = N properties = current_node["metadata"].copy() original_id = properties.get("id", id) # Preserve original ID original_memory = current_node.get("memory", "") # Preserve original memory + updated_at = fields.pop("updated_at", datetime.utcnow().isoformat()) + usage_text = f"User: {user_name} | Time: {updated_at} | Operation: Update | Overwrite： {original_memory}" + usage_info = current_node.get("usage", []) # Preserve usage_info + usage_info.insert(0, usage_text) # If fields include memory, use it; otherwise keep original memory new_memory = fields.pop("memory") if "memory" in fields else original_memory properties.update(fields) - properties["id"] = original_id # Ensure ID is not overwritten - properties["memory"] = new_memory # Ensure memory is not overwritten + properties["id"] = original_id + properties["memory"] = new_memory + properties["usage"] = usage_info + properties["updated_at"] = updated_at # Handle embedding field embedding_vector = None diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index 653510e84..38b6b5674 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -1,4 +1,5 @@ import concurrent.futures +import difflib import json from datetime import datetime @@ -10,7 +11,8 @@ from memos.graph_dbs.factory import GraphStoreFactory, PolarDBGraphDB from memos.llms.factory import AzureLLM, LLMFactory, OllamaLLM, OpenAILLM from memos.mem_feedback.base import BaseMemFeedback -from memos.mem_reader.simple_struct import SimpleStructMemReader, detect_lang +from memos.mem_reader.factory import MemReaderFactory +from memos.mem_reader.simple_struct import detect_lang from memos.memories.textual.item import TextualMemoryItem, TreeNodeTextualMemoryMetadata from memos.memories.textual.tree_text_memory.organize.manager import MemoryManager from memos.templates.mem_feedback_prompts import ( @@ -45,6 +47,7 @@ def __init__(self, config: MemFeedbackConfig): self.llm: OpenAILLM | OllamaLLM | AzureLLM = LLMFactory.from_config(config.extractor_llm) self.embedder: OllamaEmbedder = EmbedderFactory.from_config(config.embedder) self.graph_store: PolarDBGraphDB = GraphStoreFactory.from_config(config.graph_db) + self.mem_reader = MemReaderFactory.from_config(config.mem_reader) self.is_reorganize = config.reorganize self.memory_manager: MemoryManager = MemoryManager( @@ -60,6 +63,23 @@ def __init__(self, config: MemFeedbackConfig): is_reorganize=self.is_reorganize, ) + def _pure_add(self, user_name: str, feedback_content: str, feedback_time: str, info: dict): + """Directly add new memory""" + scene_data = [[{"role": "user", "content": feedback_content, "chat_time": feedback_time}]] + memories = self.mem_reader.get_memory(scene_data, type="chat", info=info) + to_add_memories = [item for scene in memories for item in scene] + added_ids = self.memory_manager.add(to_add_memories, user_name=user_name) + logger.info(f"[Feedback Core] Added {len(added_ids)} memories for user {user_name}.") + return { + "record": { + "add": [ + {"id": _id, "text": added_mem.memory} + for _id, added_mem in zip(added_ids, to_add_memories, strict=False) + ], + "update": [], + } + } + def _feedback_judgement( self, chat_history: list[MessageDict], feedback_content: str, feedback_time: str = "" ) -> dict | None: @@ -75,65 +95,105 @@ def _feedback_judgement( .replace("{feedback_time}", feedback_time) ) judge_res = self._get_llm_response(prompt) - return judge_res if judge_res else [] + if judge_res: + return judge_res + else: + logger.warning("[Feedback Core]: feedback judgement failed, return []") + return [] def _feedback_memory( self, user_name: str, feedback_memories: list[TextualMemoryItem], **kwargs ) -> dict: sync_mode = kwargs.get("sync_mode") + retrieved_memory_ids = kwargs.get("retrieved_memory_ids") or [] + retrieved_memories = [self.graph_store.get_node(_id) for _id in retrieved_memory_ids] + current_memories = [ + {"id": item["id"], "text": item["memory"]} for item in retrieved_memories + ] + + def _single_add_operation( + memory_item: TextualMemoryItem, user_name: str, sync_mode: str + ) -> dict: + """处理单个添加操作""" + added_ids = self.memory_manager.add([memory_item], user_name=user_name, mode=sync_mode) + logger.info(f"[Memory Feedback ADD] {added_ids[0]}") + + return {"id": added_ids[0], "text": memory_item.memory} + + def _single_update_operation( + op: dict, memory_item: TextualMemoryItem, user_name: str, sync_mode: str + ) -> dict: + """处理单个更新操作""" + update_id = op.get("id") + updated_ids = self.memory_manager.update( + [update_id], [memory_item], user_name=user_name, mode=sync_mode + ) + log_update_info = op.get("old_memory", "") + " >> " + op.get("text", "") + logger.info(f"[Memory Feedback UPDATE] {updated_ids[0]}, info: {log_update_info}") + + return { + "id": update_id, + "origin_memory": op.get("old_memory", ""), + "text": op.get("text", ""), + } + + def _add_or_update(memory_item: TextualMemoryItem, current_memories: list): + if current_memories == []: + current_memories = self._vec_query( + memory_item.metadata.embedding, user_name=user_name + ) - def _add_or_update(memory_item: TextualMemoryItem): - current_memories = self._vec_query(memory_item.metadata.embedding, user_name=user_name) if current_memories: lang = detect_lang("".join(memory_item.memory)) template = FEEDBACK_PROMPT_DICT["compare"][lang] prompt = template.replace("{current_memories}", str(current_memories)).replace( "{new_facts}", memory_item.memory ) - operations = self._get_llm_response(prompt).get("operation", {}) + operations = self._get_llm_response(prompt).get("operation", []) + operations = self._id_dehallucination(operations, current_memories) else: - operations = {"event": "ADD"} + operations = [{"event": "ADD"}] + logger.info(f"[Feedback memory operations]: {operations!s}") - if operations and operations["event"].lower() == "add": - added_ids = self.memory_manager.add( - [memory_item], user_name=user_name, mode=sync_mode - ) - logger.info(f"[Memory Feedback ADD] {added_ids!s}") - - return { - "record": { - "add": [{"id": added_ids[0], "text": memory_item.memory}], - "update": [], - } - } - elif operations and operations["event"].lower() == "update": - to_update_id = operations["id"] - updated_ids = self.memory_manager.update( - [to_update_id], [memory_item], user_name=user_name, mode=sync_mode - ) - log_update_info = operations["old_memory"] + " >> " + operations["text"] - logger.info(f"[Memory Feedback UPDATE] {updated_ids}, info: {log_update_info}") - - return { - "record": { - "add": [], - "update": [ - { - "id": to_update_id, - "origin_memory": operations["old_memory"], - "text": operations["text"], - } - ], - } - } - else: + if not operations: return {"record": {"add": [], "update": []}} - search_filter = {"user_name": user_name} - with ContextThreadPoolExecutor(max_workers=8) as ex: + add_results = [] + update_results = [] + + with ContextThreadPoolExecutor(max_workers=10) as executor: + future_to_op = {} + for op in operations: + event_type = op.get("event", "").lower() + + if event_type == "add": + future = executor.submit( + _single_add_operation, memory_item, user_name, sync_mode + ) + future_to_op[future] = ("add", op) + elif event_type == "update": + future = executor.submit( + _single_update_operation, op, memory_item, user_name, sync_mode + ) + future_to_op[future] = ("update", op) + + for future in concurrent.futures.as_completed(future_to_op): + result_type, original_op = future_to_op[future] + try: + result = future.result() + if result_type == "add": + add_results.append(result) + elif result_type == "update": + update_results.append(result) + except Exception as e: + logger.error(f"Operation failed for {original_op}: {e}") + + return {"record": {"add": add_results, "update": update_results}} + + with ContextThreadPoolExecutor(max_workers=3) as ex: futures = { - ex.submit(_add_or_update, mem, search_filter): i + ex.submit(_add_or_update, mem, current_memories): i for i, mem in enumerate(feedback_memories) } results = [None] * len(futures) @@ -173,14 +233,42 @@ def _get_llm_response(self, prompt: str, dsl: bool = True) -> dict: try: response_text = self.llm.generate(messages) if dsl: + response_text = response_text.replace("```", "").replace("json", "") response_json = json.loads(response_text) else: return response_text except Exception as e: - logger.error(f"[LLM] Exception during chat generation: {e}") + logger.error(f"[Feedback Core LLM] Exception during chat generation: {e}") response_json = None return response_json + def _id_dehallucination(self, operations, current_memories): + right_ids = [item["id"] for item in current_memories] + right_lower_map = {x.lower(): x for x in right_ids} + + def correct_item(data): + if data.get("event", "").lower() != "update": + return data + + original_id = data["id"] + if original_id in right_ids: + return data + + lower_id = original_id.lower() + if lower_id in right_lower_map: + data["id"] = right_lower_map[lower_id] + return data + + matches = difflib.get_close_matches(original_id, right_ids, n=1, cutoff=0.8) + if matches: + data["id"] = matches[0] + return data + + return None + + dehallu_res = [correct_item(item) for item in operations] + return [item for item in dehallu_res if item] + def _generate_answer( self, chat_history: list[MessageDict], feedback_content: str, corrected_answer: bool ) -> str: @@ -205,7 +293,6 @@ def process_feedback_core( user_name: str, chat_history: list[MessageDict], feedback_content: str, - mem_reader: SimpleStructMemReader | None = None, **kwargs, ) -> dict: """ @@ -221,31 +308,14 @@ def process_feedback_core( info = {"user_id": user_name, "session_id": session_id} logger.info(f"[Feedback Core] Starting memory feedback process for user {user_name}") - if mem_reader and not chat_history: - scene_data = [ - [{"role": "user", "content": feedback_content, "chat_time": feedback_time}] - ] - memories = mem_reader.get_memory(scene_data, type="chat", info=info) - to_add_memories = [item for scene in memories for item in scene] - added_ids = self.memory_manager.add(to_add_memories, user_name=user_name) - logger.info( - f"[Feedback Core] Added {len(added_ids)} memories for user {user_name}." - ) - return { - "record": { - "add": [ - {"id": _id, "text": added_mem.memory} - for _id, added_mem in zip(added_ids, to_add_memories, strict=False) - ], - "update": [], - } - } - - elif chat_history: + if not chat_history: + return self._pure_add(user_name, feedback_content, feedback_time, info) + + else: raw_judge = self._feedback_judgement( chat_history, feedback_content, feedback_time=feedback_time ) - judge_res = ( + valid_feedback = ( [ item for item in raw_judge @@ -254,7 +324,14 @@ def process_feedback_core( if raw_judge else [] ) - if not judge_res: + if ( + raw_judge + and raw_judge[0]["validity"].lower() == "false" + and raw_judge[0]["user_attitude"].lower() == "irrelevant" + ): + return self._pure_add(user_name, feedback_content, feedback_time, info) + + if not valid_feedback: logger.warning( f"[Feedback Core] No valid judgements for user {user_name}: {raw_judge}." ) @@ -262,9 +339,11 @@ def process_feedback_core( feedback_memories = [] feedback_memories_embeddings = self.embedder.embed( - [item["corrected_info"] for item in judge_res] + [item["corrected_info"] for item in valid_feedback] ) - for item, embedding in zip(judge_res, feedback_memories_embeddings, strict=False): + for item, embedding in zip( + valid_feedback, feedback_memories_embeddings, strict=False + ): value = item["corrected_info"] key = item["key"] tags = item["tags"] @@ -287,16 +366,13 @@ def process_feedback_core( ), ) ) + mem_record = self._feedback_memory(user_name, feedback_memories, **kwargs) logger.info( f"[Feedback Core] Processed {len(feedback_memories)} feedback memories for user {user_name}." ) return mem_record - else: - logger.info("[Feedback Core] Empty chat_history and no mem_reader, skipping.") - return {"record": {"add": [], "update": []}} - except Exception as e: logger.error(f"[Feedback Core] Error for user {user_name}: {e}") return {"record": {"add": [], "update": []}} @@ -306,7 +382,6 @@ def process_feedback( user_name: str, chat_history: list[MessageDict], feedback_content: str, - mem_reader: SimpleStructMemReader | None = None, **kwargs, ): """ @@ -316,7 +391,6 @@ def process_feedback( user_name: User identifier chat_history: List of chat messages feedback_content: Feedback content from user - mem_reader: Memory reader instance **kwargs: Additional arguments including sync_mode Returns: @@ -338,7 +412,6 @@ def process_feedback( user_name, chat_history, feedback_content, - mem_reader, **kwargs, ) concurrent.futures.wait([answer_future, core_future]) @@ -365,7 +438,6 @@ def process_feedback( user_name, chat_history, feedback_content, - mem_reader, **kwargs, ) ex.shutdown(wait=False) diff --git a/src/memos/mem_feedback/simple_feedback.py b/src/memos/mem_feedback/simple_feedback.py index 23e25ff2c..59ee38438 100644 --- a/src/memos/mem_feedback/simple_feedback.py +++ b/src/memos/mem_feedback/simple_feedback.py @@ -3,6 +3,7 @@ from memos.graph_dbs.factory import PolarDBGraphDB from memos.llms.factory import AzureLLM, OllamaLLM, OpenAILLM from memos.mem_feedback.feedback import MemFeedback +from memos.mem_reader.simple_struct import SimpleStructMemReader from memos.memories.textual.tree_text_memory.organize.manager import MemoryManager @@ -16,8 +17,10 @@ def __init__( embedder: OllamaEmbedder, graph_store: PolarDBGraphDB, memory_manager: MemoryManager, + mem_reader: SimpleStructMemReader, ): self.llm = llm self.embedder = embedder self.graph_store = graph_store self.memory_manager = memory_manager + self.mem_reader = mem_reader diff --git a/src/memos/templates/mem_feedback_prompts.py b/src/memos/templates/mem_feedback_prompts.py index 2a9a7dcb2..e1e3abfa4 100644 --- a/src/memos/templates/mem_feedback_prompts.py +++ b/src/memos/templates/mem_feedback_prompts.py @@ -101,7 +101,7 @@ 分析步骤与判定标准: 1. *有效性判定* - - 有效(true)：用户反馈的内容与对话历史的主题、任务或上一次助理的回答*相关*。例如：针对回答进行追问、纠正、补充或评价。 + - 有效(true)：用户反馈的内容与对话历史的主题、任务或上一次助理的回答*有关联*。例如：针对回答进行追问、纠正、补充或评价。 - 无效(false)：用户的反馈与对话历史*完全无关*，与任何先前内容之间不存在语义、主题或词汇上的联系。 2. *用户态度判定* - 不满意(dissatisfied)：反馈中表现出负面情绪，如直接指出错误、表达困惑、抱怨、批评，或明确表示问题未解决。 @@ -172,33 +172,40 @@ """ -UPDATE_FORMER_MEMORIES = """Please analyze the newly acquired factual information and determine how this information should be updated to the memory database: add or update, and provide final operation recommendations. +UPDATE_FORMER_MEMORIES = """Please analyze the newly acquired factual information and determine how this information should be updated to the memory database: add, update, or keep unchanged, and provide final operation recommendations. You must strictly return the response in the following JSON format: { "operation": - { - "id": "", - "text": "", - "event": "", - "old_memory": "" - } + [ + { + "id": "", + "text": "", + "event": "", + "old_memory": "" + }, + ... + ] } *Requirements*: -1. If the new fact adds no supplemental value and the existing memory supersedes it, no operation is performed. +1. If the new fact does not provide additional information to the existing memory item, the existing memory can override the new fact, and the operation is set to "NONE." 2. If the new fact is similar to existing memory but the information is more accurate, complete, or requires correction, set operation to "UPDATE" 3. If the new fact contradicts existing memory in key information (such as time, location, status, etc.), update the original memory based on the new fact and set operation to "UPDATE" -4. If there is completely new information to add, set operation to "ADD" +4. If there is no existing memory that requires updating, the new fact is added as entirely new information, and the operation is set to "ADD." Therefore, in the same operation list, ADD and UPDATE will not coexist. + *ID Management Rules*: - Update operation: Keep the original ID unchanged - Add operation: Generate a new unique ID in the format of a 4-digit string (e.g., "0001", "0002", etc.) *Important Requirements*: -- Return only the JSON format response, without any other content - For update operations, you must provide the old_memory field to show the original content +- Compare the existing memories one by one and do not miss any content that needs to be updated. When multiple existing memories need to be updated, include all relevant entries in the operation list + +If the new fact contradicts existing memory in key information (such as time, location, status, etc.), update ALL affected original memories based on the new fact and set operation to "UPDATE" for each one. Multiple memories covering the same outdated information should all be updated. +- Return only the JSON format response, without any other content - text field requirements: Use concise, complete declarative sentences that are consistent with the newly acquired factual information, avoiding redundant information - text and old_memory content should be in English @@ -206,29 +213,65 @@ Current Memories: { "memory": [ + { + "id": "0911", + "text": "The user is a senior full-stack developer working at Company B" + }, { "id": "123", - "text": "The user works as a software engineer in Company A, mainly responsible for front-end development" + "text": "The user works as a software engineer at Company A, primarily responsible for front-end development" + }, + { + "id": "648", + "text": "The user is responsible for front-end development of software at Company A" + }, + { + "id": "7210", + "text": "The user is responsible for front-end development of software at Company A" }, { "id": "908", - "text": "The user likes to go fishing with friends on weekends" + "text": "The user enjoys fishing with friends on weekends" } ] } Newly facts: -"The user is currently working as a senior full-stack development engineer at Company B" +"The user works as a senior full-stack developer at Company B" Operation recommendations: { "operation": - { - "id": "123", - "text": "The user is currently working as a senior full-stack development engineer at Company B", - "event": "UPDATE", - "old_memory": "The user works as a software engineer in Company A, mainly responsible for front-end development" - } + [ + { + "id": "0911", + "text": "The user is a senior full-stack developer working at Company B", + "event": "NONE" + }, + { + "id": "123", + "text": "The user works as a senior full-stack developer at Company B", + "event": "UPDATE", + "old_memory": "The user works as a software engineer at Company A, primarily responsible for front-end development" + }, + { + "id": "648", + "text": "The user works as a senior full-stack developer at Company B", + "event": "UPDATE", + "old_memory": "The user is responsible for front-end development of software at Company A" + }, + { + "id": "7210", + "text": "The user works as a senior full-stack developer at Company B", + "event": "UPDATE", + "old_memory": "The user is responsible for front-end development of software at Company A" + }, + { + "id": "908", + "text": "The user enjoys fishing with friends on weekends", + "event": "NONE" + } + ] } Example2: @@ -252,11 +295,23 @@ Operation recommendations: { "operation": - { - "id": "4567", - "text": "The user's residential address is Mingyue Community, Chaoyang District, Beijing", - "event": "ADD" - } + [ + { + "id": "123", + "text": "The user works as a software engineer at Company A, primarily responsible for front-end development", + "event": "NONE" + }, + { + "id": "908", + "text": "The user enjoys fishing with friends on weekends", + "event": "NONE" + }, + { + "id": "4567", + "text": "The user's residential address is Mingyue Community, Chaoyang District, Beijing", + "event": "ADD" + } + ] } Current Memories @@ -269,33 +324,37 @@ """ -UPDATE_FORMER_MEMORIES_ZH = """请分析新获取的事实信息，并决定该信息应该如何更新到记忆库中：新增或更新，并给出最终的操作建议。 +UPDATE_FORMER_MEMORIES_ZH = """请分析新获取的事实信息，并决定这些信息应该如何更新到记忆库中：新增、更新、或保持不变，并给出最终的操作建议。 你必须严格按照以下JSON格式返回响应： { "operation": - { - "id": "<记忆ID>", - "text": "<记忆内容>", - "event": "<操作类型，必须是 "ADD", "UPDATE"之一>", - "old_memory": "<原记忆内容，仅当操作为"UPDATE"时需要提供>" - } + [ + { + "id": "<记忆ID>", + "text": "<记忆内容>", + "event": "<操作类型，必须是 "ADD", "UPDATE", "NONE" 之一>", + "old_memory": "<原记忆内容，仅当操作为"UPDATE"时需要提供>" + }, + ... + ] } 要求： -1. 如果新事实对现有记忆没有额外补充，现有记忆的信息可以覆盖新事实，则不设置任何操作 -2. 如果新事实与现有记忆相似但信息更准确、完整或需要修正，设置操作为"UPDATE" +1. 如果新事实对现有记忆item没有额外补充，现有记忆的信息可以覆盖新事实，设置操作为"NONE" +2. 如果新事实与现有记忆item相似但信息更准确、完整或需要修正，设置操作为"UPDATE" 3. 如果新事实与现有记忆在关键信息上矛盾（如时间、地点、状态等），以新事实为准更新原有记忆，设置操作为"UPDATE" -4. 如果有全新信息添加，设置操作为"ADD" +4. 如果现有记忆中没有需要更新的，则新事实作为全新信息添加，设置操作为"ADD"。因此可知同一个 operation 列表中，ADD和UPDATE不会同时存在。 ID管理规则： - 更新操作：保持原有ID不变 - 新增操作：生成新的唯一ID，格式为4位数字字符串（如："0001", "0002"等） 重要要求： -- 只返回JSON格式的响应，不要包含其他任何内容 - 对于更新操作，必须提供old_memory字段显示原内容 +- 对现有记忆逐一比对，不可漏掉需要更新的内容。当多个现有记忆需要更新时，将所有的相关条目都包含在操作列表中 +- 只返回JSON格式的响应，不要包含其他任何内容 - text字段要求：使用简洁、完整的陈述句，和新获取的事实信息一致，避免冗余信息 - text和old_memory内容使用中文 @@ -303,10 +362,22 @@ 现有记忆记录： { "memory": [ + { + "id": "0911", + "text": "用户是高级全栈开发工程师，在B公司工作" + }, { "id": "123", "text": "用户在公司A担任软件工程师，主要负责前端开发" }, + { + "id": "648", + "text": "用户在公司A负责软件的前端开发工作" + }, + { + "id": "7210", + "text": "用户在公司A负责软件的前端开发工作" + }, { "id": "908", "text": "用户周末喜欢和朋友一起钓鱼" @@ -320,12 +391,36 @@ 操作建议： { "operation": - { - "id": "123", - "text": "用户在公司B担任高级全栈开发工程师", - "event": "UPDATE", - "old_memory": "用户在公司A担任软件工程师，主要负责前端开发" - } + [ + { + "id": "0911", + "text": "用户是高级全栈开发工程师，在B公司工作", + "event": "NONE" + }, + { + "id": "123", + "text": "用户现在在公司B担任高级全栈开发工程师", + "event": "UPDATE", + "old_memory": "用户在公司A担任软件工程师，主要负责前端开发" + }, + { + "id": "648", + "text": "用户现在在公司B担任高级全栈开发工程师", + "event": "UPDATE", + "old_memory": "用户在公司A负责软件的前端开发工作" + }, + { + "id": "7210", + "text": "用户现在在公司B担任高级全栈开发工程师", + "event": "UPDATE", + "old_memory": "用户在公司A负责软件的前端开发工作" + }, + { + "id": "908", + "text": "用户周末喜欢和朋友一起钓鱼", + "event": "NONE" + } + ] } 示例2： @@ -349,11 +444,23 @@ 操作建议： { "operation": - { + [ + { + "id": "123", + "text": "用户在公司A担任软件工程师，主要负责前端开发", + "event": "NONE" + }, + { + "id": "908", + "text": "用户周末喜欢和朋友一起钓鱼", + "event": "NONE" + }, + { "id": "4567", "text": "用户的居住地址是北京市朝阳区明月小区", "event": "ADD" - } + } + ] } 现有记忆记录： From aef3aad8da82af8b83da5e11f105989e82e3d9e2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 26 Nov 2025 20:55:46 +0800 Subject: [PATCH 22/72] upgrade feedback using --- src/memos/graph_dbs/polardb.py | 11 +- src/memos/mem_feedback/feedback.py | 211 ++++++-- .../tree_text_memory/organize/manager.py | 2 +- src/memos/templates/mem_feedback_prompts.py | 506 ++++++++++-------- 4 files changed, 449 insertions(+), 281 deletions(-) diff --git a/src/memos/graph_dbs/polardb.py b/src/memos/graph_dbs/polardb.py index 74bce3b97..ebaec2b90 100644 --- a/src/memos/graph_dbs/polardb.py +++ b/src/memos/graph_dbs/polardb.py @@ -525,9 +525,11 @@ def update_node(self, id: str, fields: dict[str, Any], user_name: str | None = N original_id = properties.get("id", id) # Preserve original ID original_memory = current_node.get("memory", "") # Preserve original memory updated_at = fields.pop("updated_at", datetime.utcnow().isoformat()) - usage_text = f"User: {user_name} | Time: {updated_at} | Operation: Update | Overwrite： {original_memory}" - usage_info = current_node.get("usage", []) # Preserve usage_info - usage_info.insert(0, usage_text) + + record_info = f"User：{user_name} | Time：{updated_at} | Operation：Update | Overwrite： {original_memory}" + covered_history = current_node.get("covered_history", []) + covered_history.insert(0, record_info) + logger.info(f"New GraphDB Update: {record_info}") # If fields include memory, use it; otherwise keep original memory new_memory = fields.pop("memory") if "memory" in fields else original_memory @@ -535,13 +537,14 @@ def update_node(self, id: str, fields: dict[str, Any], user_name: str | None = N properties.update(fields) properties["id"] = original_id properties["memory"] = new_memory - properties["usage"] = usage_info + properties["covered_history"] = covered_history properties["updated_at"] = updated_at # Handle embedding field embedding_vector = None if "embedding" in fields: embedding_vector = fields.pop("embedding") + assert properties["embedding"] == embedding_vector, "Embedding vector mismatch" if not isinstance(embedding_vector, list): embedding_vector = None diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index 38b6b5674..6b8520c50 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -4,6 +4,8 @@ from datetime import datetime +from tenacity import retry, stop_after_attempt, wait_exponential + from memos import log from memos.configs.memory import MemFeedbackConfig from memos.context.context import ContextThreadPoolExecutor @@ -64,12 +66,18 @@ def __init__(self, config: MemFeedbackConfig): ) def _pure_add(self, user_name: str, feedback_content: str, feedback_time: str, info: dict): - """Directly add new memory""" + """ + Directly add new memory + """ scene_data = [[{"role": "user", "content": feedback_content, "chat_time": feedback_time}]] memories = self.mem_reader.get_memory(scene_data, type="chat", info=info) to_add_memories = [item for scene in memories for item in scene] - added_ids = self.memory_manager.add(to_add_memories, user_name=user_name) - logger.info(f"[Feedback Core] Added {len(added_ids)} memories for user {user_name}.") + added_ids = self._retry_db_operation( + lambda: self.memory_manager.add(to_add_memories, user_name=user_name) + ) + logger.info( + f"[Feedback Core: _pure_add] Added {len(added_ids)} memories for user {user_name}." + ) return { "record": { "add": [ @@ -88,17 +96,21 @@ def _feedback_judgement( """ lang = detect_lang(feedback_content) template = FEEDBACK_PROMPT_DICT["judge"][lang] - chat_history_str = str(chat_history[-4:]) - prompt = ( - template.replace("{chat_history}", chat_history_str) - .replace("{user_feedback}", feedback_content) - .replace("{feedback_time}", feedback_time) + chat_history_lis = [f"""{msg["role"]}: {msg["content"]}""" for msg in chat_history[-4:]] + chat_history_str = "\n".join(chat_history_lis) + prompt = template.format( + chat_history=chat_history_str, + user_feedback=feedback_content, + feedback_time=feedback_time, ) + judge_res = self._get_llm_response(prompt) if judge_res: return judge_res else: - logger.warning("[Feedback Core]: feedback judgement failed, return []") + logger.warning( + "[Feedback Core: _feedback_judgement] feedback judgement failed, return []" + ) return [] def _feedback_memory( @@ -106,16 +118,36 @@ def _feedback_memory( ) -> dict: sync_mode = kwargs.get("sync_mode") retrieved_memory_ids = kwargs.get("retrieved_memory_ids") or [] + chat_history = kwargs.get("chat_history", []) + feedback_content = kwargs.get("feedback_content", "") + + chat_history_lis = [f"""{msg["role"]}: {msg["content"]}""" for msg in chat_history[-4:]] + fact_history = "\n".join(chat_history_lis) + f"\nuser feedback: \n{feedback_content}" + retrieved_memories = [self.graph_store.get_node(_id) for _id in retrieved_memory_ids] + filterd_ids = [ + item["id"] for item in retrieved_memories if "mode:fast" in item["metadata"]["tags"] + ] + if filterd_ids: + logger.warning( + f"[Feedback Core: _feedback_memory] Since the tags mode is fast, no modifications are made to the following memory {filterd_ids}." + ) + current_memories = [ - {"id": item["id"], "text": item["memory"]} for item in retrieved_memories + {"id": item["id"], "text": item["memory"]} + for item in retrieved_memories + if "mode:fast" not in item["metadata"]["tags"] ] def _single_add_operation( memory_item: TextualMemoryItem, user_name: str, sync_mode: str ) -> dict: - """处理单个添加操作""" - added_ids = self.memory_manager.add([memory_item], user_name=user_name, mode=sync_mode) + """ + Individual addition operations + """ + added_ids = self._retry_db_operation( + lambda: self.memory_manager.add([memory_item], user_name=user_name, mode=sync_mode) + ) logger.info(f"[Memory Feedback ADD] {added_ids[0]}") return {"id": added_ids[0], "text": memory_item.memory} @@ -123,10 +155,14 @@ def _single_add_operation( def _single_update_operation( op: dict, memory_item: TextualMemoryItem, user_name: str, sync_mode: str ) -> dict: - """处理单个更新操作""" + """ + Individual update operations + """ update_id = op.get("id") - updated_ids = self.memory_manager.update( - [update_id], [memory_item], user_name=user_name, mode=sync_mode + updated_ids = self._retry_db_operation( + lambda: self.memory_manager.update( + [update_id], [memory_item], user_name=user_name, mode=sync_mode + ) ) log_update_info = op.get("old_memory", "") + " >> " + op.get("text", "") logger.info(f"[Memory Feedback UPDATE] {updated_ids[0]}, info: {log_update_info}") @@ -137,7 +173,9 @@ def _single_update_operation( "text": op.get("text", ""), } - def _add_or_update(memory_item: TextualMemoryItem, current_memories: list): + def _add_or_update( + memory_item: TextualMemoryItem, current_memories: list, fact_history: str + ): if current_memories == []: current_memories = self._vec_query( memory_item.metadata.embedding, user_name=user_name @@ -146,9 +184,12 @@ def _add_or_update(memory_item: TextualMemoryItem, current_memories: list): if current_memories: lang = detect_lang("".join(memory_item.memory)) template = FEEDBACK_PROMPT_DICT["compare"][lang] - prompt = template.replace("{current_memories}", str(current_memories)).replace( - "{new_facts}", memory_item.memory + prompt = template.format( + current_memories=str(current_memories), + new_facts=memory_item.memory, + chat_history=fact_history, ) + operations = self._get_llm_response(prompt).get("operation", []) operations = self._id_dehallucination(operations, current_memories) else: @@ -187,13 +228,16 @@ def _add_or_update(memory_item: TextualMemoryItem, current_memories: list): elif result_type == "update": update_results.append(result) except Exception as e: - logger.error(f"Operation failed for {original_op}: {e}") + logger.error( + f"[Feedback Core: _add_or_update] Operation failed for {original_op}: {e}", + exc_info=True, + ) return {"record": {"add": add_results, "update": update_results}} with ContextThreadPoolExecutor(max_workers=3) as ex: futures = { - ex.submit(_add_or_update, mem, current_memories): i + ex.submit(_add_or_update, mem, current_memories, fact_history): i for i, mem in enumerate(feedback_memories) } results = [None] * len(futures) @@ -204,7 +248,10 @@ def _add_or_update(memory_item: TextualMemoryItem, current_memories: list): if node: results[i] = node except Exception as e: - logger.error(f"[FeedBack] error: {e}") + logger.error( + f"[Feedback Core: _feedback_memory] Error processing memory index {i}: {e}", + exc_info=True, + ) mem_res = [r for r in results if r] return { @@ -216,22 +263,34 @@ def _add_or_update(memory_item: TextualMemoryItem, current_memories: list): def _vec_query(self, new_memories_embedding: list[float], user_name=None): retrieved_ids = self.graph_store.search_by_embedding( - new_memories_embedding, user_name=user_name + new_memories_embedding, user_name=user_name, top_k=5 ) current_memories = [self.graph_store.get_node(item["id"]) for item in retrieved_ids] + if not retrieved_ids: + logger.info( + f"[Feedback Core: _vec_query] No similar memories found for embedding query for user {user_name}." + ) + filterd_ids = [ + item["id"] for item in current_memories if "mode:fast" in item["metadata"]["tags"] + ] + if filterd_ids: + logger.warning( + f"[Feedback Core: _vec_query] Since the tags mode is fast, no modifications are made to the following memory {filterd_ids}." + ) return [ { "id": item["id"], "text": item["memory"], } for item in current_memories + if "mode:fast" not in item["metadata"]["tags"] ] def _get_llm_response(self, prompt: str, dsl: bool = True) -> dict: messages = [{"role": "user", "content": prompt}] try: - response_text = self.llm.generate(messages) + response_text = self.llm.generate(messages, temperature=0.3) if dsl: response_text = response_text.replace("```", "").replace("json", "") response_json = json.loads(response_text) @@ -275,7 +334,7 @@ def _generate_answer( """ Answer generation to facilitate concurrent submission. """ - if not corrected_answer: + if not corrected_answer or feedback_content.strip() == "": return "" lang = detect_lang(feedback_content) template = FEEDBACK_PROMPT_DICT["generation"][lang] @@ -283,9 +342,8 @@ def _generate_answer( [f"{item['role']}: {item['content']}" for item in chat_history] ) chat_history_str = chat_history_str if chat_history_str else "none" - prompt = template.replace("{chat_history}", chat_history_str).replace( - "{question}", feedback_content - ) + prompt = template.format(chat_history=chat_history_str, question=feedback_content) + return self._get_llm_response(prompt, dsl=False) def process_feedback_core( @@ -298,16 +356,28 @@ def process_feedback_core( """ Core feedback processing: judgment, memory extraction, addition/update. Return record. """ + + def check_validity(item): + return ( + "validity" in item + and item["validity"].lower() == "true" + and "corrected_info" in item + and item["corrected_info"].strip() + and "key" in item + and "tags" in item + ) + try: feedback_time = kwargs.get("feedback_time") or datetime.now().isoformat() session_id = kwargs.get("session_id") allow_knowledgebase_write = bool(kwargs.get("allow_knowledgebase_write")) - if not allow_knowledgebase_write: + if feedback_content.strip() == "" or not allow_knowledgebase_write: return {"record": {"add": [], "update": []}} info = {"user_id": user_name, "session_id": session_id} - logger.info(f"[Feedback Core] Starting memory feedback process for user {user_name}") - + logger.info( + f"[Feedback Core: process_feedback_core] Starting memory feedback process for user {user_name}" + ) if not chat_history: return self._pure_add(user_name, feedback_content, feedback_time, info) @@ -316,13 +386,7 @@ def process_feedback_core( chat_history, feedback_content, feedback_time=feedback_time ) valid_feedback = ( - [ - item - for item in raw_judge - if item["validity"].lower() == "true" and item["corrected_info"].strip() - ] - if raw_judge - else [] + [item for item in raw_judge if check_validity(item)] if raw_judge else [] ) if ( raw_judge @@ -333,14 +397,25 @@ def process_feedback_core( if not valid_feedback: logger.warning( - f"[Feedback Core] No valid judgements for user {user_name}: {raw_judge}." + f"[Feedback Core: process_feedback_core] No valid judgements for user {user_name}: {raw_judge}." ) return {"record": {"add": [], "update": []}} feedback_memories = [] - feedback_memories_embeddings = self.embedder.embed( - [item["corrected_info"] for item in valid_feedback] - ) + + corrected_infos = [item["corrected_info"] for item in valid_feedback] + embed_bs = 5 + feedback_memories_embeddings = [] + for i in range(0, len(corrected_infos), embed_bs): + batch = corrected_infos[i : i + embed_bs] + try: + feedback_memories_embeddings.extend(self.embedder.embed(batch)) + except Exception as e: + logger.error( + f"[Feedback Core: process_feedback_core] Embedding batch failed: {e}", + exc_info=True, + ) + for item, embedding in zip( valid_feedback, feedback_memories_embeddings, strict=False ): @@ -367,14 +442,20 @@ def process_feedback_core( ) ) - mem_record = self._feedback_memory(user_name, feedback_memories, **kwargs) + mem_record = self._feedback_memory( + user_name, + feedback_memories, + chat_history=chat_history, + feedback_content=feedback_content, + **kwargs, + ) logger.info( - f"[Feedback Core] Processed {len(feedback_memories)} feedback memories for user {user_name}." + f"[Feedback Core: process_feedback_core] Processed {len(feedback_memories)} feedback memories for user {user_name}." ) return mem_record except Exception as e: - logger.error(f"[Feedback Core] Error for user {user_name}: {e}") + logger.error(f"[Feedback Core: process_feedback_core] Error for user {user_name}: {e}") return {"record": {"add": [], "update": []}} def process_feedback( @@ -414,17 +495,25 @@ def process_feedback( feedback_content, **kwargs, ) - concurrent.futures.wait([answer_future, core_future]) + done, pending = concurrent.futures.wait([answer_future, core_future], timeout=30) + for fut in pending: + fut.cancel() try: answer = answer_future.result() record = core_future.result() logger.info( - f"[process_feedback sync] Completed concurrently for user {user_name} with full results." + f"[MemFeedback sync] Completed concurrently for user {user_name} with full results." ) return {"answer": answer, "record": record["record"]} + except concurrent.futures.TimeoutError: + logger.error( + f"[MemFeedback sync] Timeout in sync mode for {user_name}", exc_info=True + ) + return {"answer": "", "record": {"add": [], "update": []}} except Exception as e: logger.error( - f"[process_feedback sync] Error in concurrent tasks for {user_name}: {e}" + f"[MemFeedback sync] Error in concurrent tasks for {user_name}: {e}", + exc_info=True, ) return {"answer": "", "record": {"add": [], "update": []}} else: @@ -444,14 +533,34 @@ def process_feedback( def log_completion(f): try: - result = f.result() - logger.info(f"[Background Feedback] Completed for {user_name}: {result}") + result = f.result(timeout=600) + logger.info(f"[MemFeedback async] Completed for {user_name}: {result}") + except concurrent.futures.TimeoutError: + logger.error( + f"[MemFeedback async] Background task timeout for {user_name}", + exc_info=True, + ) + f.cancel() except Exception as e: - logger.error(f"[Background Feedback] Error for {user_name}: {e}") + logger.error( + f"[MemFeedback async] Background Feedback Error for {user_name}: {e}", + exc_info=True, + ) future.add_done_callback(log_completion) logger.info( - f"[process_feedback async] Returned answer, background task started for {user_name}." + f"[MemFeedback async] Returned answer, background task started for {user_name}." ) return {"answer": answer, "record": {"add": [], "update": []}} + + # Helper for DB operations with retry + @retry(stop=stop_after_attempt(3), wait=wait_exponential(multiplier=1, min=4, max=10)) + def _retry_db_operation(self, operation): + try: + return operation() + except Exception as e: + logger.error( + f"[MemFeedback: _retry_db_operation] DB operation failed: {e}", exc_info=True + ) + raise diff --git a/src/memos/memories/textual/tree_text_memory/organize/manager.py b/src/memos/memories/textual/tree_text_memory/organize/manager.py index dadf7c321..94b86cec0 100644 --- a/src/memos/memories/textual/tree_text_memory/organize/manager.py +++ b/src/memos/memories/textual/tree_text_memory/organize/manager.py @@ -132,7 +132,7 @@ def update( self.graph_store.update_node(_id, field, user_name) graph_class_name = self.graph_store.__class__.__name__ - logger.info(f"Updated {graph_class_name}") + logger.info(f"[MemoryManager] Updated {graph_class_name}") return memories_ids def replace_working_memory( diff --git a/src/memos/templates/mem_feedback_prompts.py b/src/memos/templates/mem_feedback_prompts.py index e1e3abfa4..11e17e399 100644 --- a/src/memos/templates/mem_feedback_prompts.py +++ b/src/memos/templates/mem_feedback_prompts.py @@ -3,40 +3,33 @@ Analysis Steps and Criteria: 1. *Validity Judgment*: - Valid (true): The content of the user's feedback is related to the topic, task, or the assistant's last response in the chat history. For example: asking follow-up questions, making corrections, providing supplements, or evaluating the last response. - - Invalid (false): The user’s feedback is entirely unrelated to the conversation history, with no semantic, topical, or lexical connection to any prior content. + - Invalid (false): The user's feedback is entirely unrelated to the conversation history, with no semantic, topical, or lexical connection to any prior content. 2. *User Attitude Judgment*: - Dissatisfied: The feedback shows negative emotions, such as directly pointing out errors, expressing confusion, complaining, criticizing, or explicitly stating that the problem remains unsolved. - Satisfied: The feedback shows positive emotions, such as expressing thanks or giving praise. - Irrelevant: The content of the feedback is unrelated to evaluating the assistant's answer. -3. *Assistant Response Effectiveness Type Judgment*: - - Wrong: The assistant provided incorrect information. - - Missing: The assistant's response was correct in direction but incomplete, omitting key details. - - None: The user feedback does not point to any shortcomings in the assistant's response. - -4. *Summary Information Generation*(corrected_info field): +3. *Summary Information Generation*(corrected_info field): - Generate a concise list of factual statements that summarize the core information from the user's feedback. - — Focus on objective facts, corrections, or confirmations. - - Express time information as concrete, unambiguous date(s) or period(s) (e.g., “March 2023”, “2024-07”, or “May–June 2022”). - - For 'Satisfied' or 'None' types, this list may contain confirming statements or be empty if no new facts are provided. - - For example: "The user completed the Everest Circuit trek with colleagues in March 2023." + - When the feedback provides corrections, focus only on the corrected information. + - When the feedback provides supplements, integrate all valid information (both old and new). + - It is very important to keep any relevant time information and express time information as concrete, unambiguous date(s) or period(s) (e.g., "March 2023", "2024-07", or "May–June 2022"). + - For 'satisfied' attitude, this list may contain confirming statements or be empty if no new facts are provided. + - Focus on statement of objective facts. For example: "The user completed the Everest Circuit trek with colleagues in March 2023." Output Format: [ - { + {{ "validity": "", "user_attitude": "", - "error_type": "", "corrected_info": "", - "key": , - "tags": - }, - ... + "key": "", + "tags": "" + }} ] Example1: - Dialogue History: user: I can't eat spicy food these days. Can you recommend some suitable restaurants for me? assistant: Sure, I recommend the Fish Restaurant near you. Their signature dishes include various types of steamed seafood and sashimi of sea fish. @@ -47,147 +40,183 @@ Output: [ - { - "validity": "true", - "user_attitude": "dissatisfied", - "error_type": "wrong", - "corrected_info": "User is allergic to seafood", - "key": "allergic to seafood", - "tags": ["allergic", "seafood"] - }, - { - "validity": "true", - "user_attitude": "dissatisfied", - "error_type": "wrong", - "corrected_info": "User does not like eating raw fish.", - "key": "dislike eating raw fish.", - "tags": ["dislike", "raw fish"] - } + {{ + "validity": "true", + "user_attitude": "dissatisfied", + "corrected_info": "User is allergic to seafood and does not like eating raw fish.", + "key": "dietary restrictions", + "tags": ["allergic", "seafood", "raw fish", "food preference"] + }} ] Example2: +Dialogue History: +user: When did I bought on November 25, 2025? +assistant: A red coat +feedback time: 2025-11-28T20:45:00.875249 +User Feedback: +No, I also bought a blue shirt. + +Output: +[ + {{ + "validity": "true", + "user_attitude": "dissatisfied", + "corrected_info": "User bought a red coat and a blue shirt on November 25, 2025", + "key": "shopping record", + "tags": ["purchase", "clothing", "shopping"] + }} +] + +Example3: Dialogue History: -user: When did Jhon graduated? -assistant: 2014 -feedback time: 2025-11-18T20:45:00.875249 +user: What's my favorite food? +assistant: Pizza and sushi +feedback time: 2024-07-15T10:30:00.000000 User Feedback: -Wrong. He graduated the following year. +Wrong! I hate sushi. I like burgers. Output: [ - { - "validity": "true", - "user_attitude": "dissatisfied", - "error_type": "wrong", - "corrected_info": "Jhon was graduated at 2015", - "key": "Jhon graduated time", - "tags": ["Jhon", "graduated", "year"] - } + {{ + "validity": "true", + "user_attitude": "dissatisfied", + "corrected_info": "User likes pizza and burgers, but hates sushi.", + "key": "food preferences", + "tags": ["food preferences", "pizza", "burgers", "sushi"] + }} ] + Dialogue History: {chat_history} + feedback time: {feedback_time} User Feedback: {user_feedback} -Output: -""" +Output:""" +FEEDBACK_JUDGEMENT_PROMPT_ZH = """您是一个回答质量分析专家。请严格按照以下步骤和标准分析提供的"用户与助手聊天历史"和"用户反馈"，并将最终评估结果填入指定的JSON格式中。 -FEEDBACK_JUDGEMENT_PROMPT_ZH = """你是一个对话质量分析专家。请严格根据以下步骤和标准，对提供的“用户和助理的对话历史”和“用户反馈”进行分析，并将最终判定结果填入指定的JSON格式中。 - -分析步骤与判定标准: -1. *有效性判定* - - 有效(true)：用户反馈的内容与对话历史的主题、任务或上一次助理的回答*有关联*。例如：针对回答进行追问、纠正、补充或评价。 - - 无效(false)：用户的反馈与对话历史*完全无关*，与任何先前内容之间不存在语义、主题或词汇上的联系。 -2. *用户态度判定* - - 不满意(dissatisfied)：反馈中表现出负面情绪，如直接指出错误、表达困惑、抱怨、批评，或明确表示问题未解决。 - - 满意(satisfied)：反馈中表现出正面情绪，如表示感谢或给予称赞。 - - 无关(irrelevant)：反馈内容与评价助理回答无关。 -3. *助理回答效果类型判定* - - 错误(wrong)：助理提供了不正确的信息。 - - 缺漏(missing)：助理的回答方向正确但不完整，遗漏了关键细节。 - - 无(none)：用户反馈并未指向助理回答的任何不足。 -4. *总结信息生成* - - 生成一份简洁的事实陈述列表，该列表概括了用户反馈中的核心信息。 - - 重点放在客观事实、更正或确认上。 - - 对于“满意”或“无”类型的反馈，该列表可能包含确认性的陈述，或者如果未提供新事实，则可能为空。 - - 例如：“用户在2023年3月与同事完成了珠峰环线徒步旅行。” +分析步骤和标准： +1. *有效性判断*：(validity字段) + - 有效（true）：用户反馈的内容与聊天历史中的主题、任务或助手的最后回复相关。例如：提出后续问题、进行纠正、提供补充或评估最后回复。 + - 无效（false）：用户反馈与对话历史完全无关，与之前内容没有任何语义、主题或词汇联系。 + +2. *用户态度判断*：(user_attitude字段) + - 不满意：反馈显示负面情绪，如直接指出错误、表达困惑、抱怨、批评，或明确表示问题未解决。 + - 满意：反馈显示正面情绪，如表达感谢或给予赞扬。 + - 无关：反馈内容与评估助手回答无关。 + +3. *摘要信息生成*（corrected_info字段）： + - 从用户反馈中总结核心信息，生成简洁的事实陈述列表。 + - 当反馈提供纠正时，仅关注纠正后的信息。 + - 当反馈提供补充时，整合所有有效信息（包括旧信息和新信息）。 + - 非常重要：保留相关时间信息，并以具体、明确的日期或时间段表达（例如："2023年3月"、"2024年7月"或"2022年5月至6月"）。 + - 对于"满意"态度，此列表可能包含确认性陈述，如果没有提供新事实则为空。 + - 专注于客观事实陈述。例如："用户于2023年3月与同事完成了珠峰环线徒步。" 输出格式： [ - { - "validity": <字符串，"true" 或 "false">, - "user_attitude": <字符串，"dissatisfied" 或 "satisfied" 或 "irrelevant">, - "error_type": <字符串，"wrong" 或 "missing" 或 "irrelevant">, - "corrected_info": <字符串，中文书写正确的信息记录>, - "key": <字符串，唯一且简洁的记忆标题>, - "tags": <相关主题关键词列表（例如，["截止日期", "团队", "计划"]）> - }, - ... + {{ + "validity": "<字符串，'true' 或 'false'>", + "user_attitude": "<字符串，'dissatisfied' 或 'satisfied' 或 'irrelevant'>", + "corrected_info": "<字符串，用中文书写的事实信息记录>", + "key": "<字符串，简洁的中文记忆标题，用于快速识别该条目的核心内容（2-5个汉字）>", + "tags": "<列表，中文关键词列表（每个标签1-3个汉字），用于分类和检索>" + }} ] -示例： - -用户和助理的对话历史： -user: 这两天我吃不了辣椒，给我推荐一些适合的餐厅吧。 -assistant: 好的，推荐您附近的新荣记餐厅，黄鱼年糕以及各类清蒸海鲜是这件餐厅的招牌菜。 +示例1： +对话历史： +用户：这些天我不能吃辣。能给我推荐一些合适的餐厅吗？ +助手：好的，我推荐您附近的鱼类餐厅。他们的招牌菜包括各种蒸海鲜和海鱼生鱼片。 反馈时间：2023-1-18T14:25:00.856481 用户反馈： -你忘记我海鲜过敏这件事了吗？而且我不喜欢年糕的口感。 +哦，不！我对海鲜过敏！而且我不喜欢吃生鱼。 + +输出： +[ + {{ + "validity": "true", + "user_attitude": "dissatisfied", + "corrected_info": "用户对海鲜过敏且不喜欢吃生鱼", + "key": "饮食限制", + "tags": ["过敏", "海鲜", "生鱼", "饮食偏好"] + }} +] + +示例2： +对话历史： +用户：我2025年11月25日买了什么？ +助手：一件红色外套 +反馈时间：2025-11-28T20:45:00.875249 + +用户反馈： +不对，我还买了一件蓝色衬衫。 输出： [ - { - "validity": "true", - "user_attitude": "dissatisfied", - "error_type": "wrong", - "corrected_info": "用户对海鲜过敏。", - "key": "海鲜过敏", - "tags": ["海鲜", "过敏"] - }, - { - "validity": "true", - "user_attitude": "dissatisfied", - "error_type": "wrong", - "corrected_info": "用户不喜欢年糕的口感。", - "key": "不喜欢年糕", - "tags": ["不喜欢年糕", "年糕", "口感"] - } + {{ + "validity": "true", + "user_attitude": "dissatisfied", + "corrected_info": "用户于2025年11月25日购买了一件红色外套和一件蓝色衬衫", + "key": "购物记录", + "tags": ["红色外套", "蓝色衬衫", "服装购物"] + }} ] +示例3： +对话历史： +用户：我最喜欢的食物是什么？ +助手：披萨和寿司 +反馈时间：2024-07-15T10:30:00.000000 -用户和助理的对话历史： +用户反馈： +错了！我讨厌寿司。我喜欢汉堡。 + +输出： +[ + {{ + "validity": "true", + "user_attitude": "dissatisfied", + "corrected_info": "用户喜欢披萨和汉堡，但讨厌寿司", + "key": "食物偏好", + "tags": ["偏好", "披萨和汉堡"] + }} +] + +对话历史： {chat_history} + 反馈时间：{feedback_time} 用户反馈： {user_feedback} -输出： -""" +输出：""" UPDATE_FORMER_MEMORIES = """Please analyze the newly acquired factual information and determine how this information should be updated to the memory database: add, update, or keep unchanged, and provide final operation recommendations. You must strictly return the response in the following JSON format: -{ +{{ "operation": [ - { + {{ "id": "", "text": "", "event": "", "old_memory": "" - }, + }}, ... ] -} +}} *Requirements*: 1. If the new fact does not provide additional information to the existing memory item, the existing memory can override the new fact, and the operation is set to "NONE." @@ -211,113 +240,126 @@ Example1: Current Memories: -{ +{{ "memory": [ - { + {{ "id": "0911", "text": "The user is a senior full-stack developer working at Company B" - }, - { + }}, + {{ "id": "123", "text": "The user works as a software engineer at Company A, primarily responsible for front-end development" - }, - { + }}, + {{ "id": "648", "text": "The user is responsible for front-end development of software at Company A" - }, - { + }}, + {{ "id": "7210", "text": "The user is responsible for front-end development of software at Company A" - }, - { + }}, + {{ "id": "908", "text": "The user enjoys fishing with friends on weekends" - } + }} ] -} +}} + +The background of the new fact being put forward: +user: Do you remember where I work？ +assistant: Company A. +user feedback: I work at Company B, and I am a senior full-stack developer. Newly facts: "The user works as a senior full-stack developer at Company B" Operation recommendations: -{ +{{ "operation": [ - { + {{ "id": "0911", "text": "The user is a senior full-stack developer working at Company B", "event": "NONE" - }, - { + }}, + {{ "id": "123", "text": "The user works as a senior full-stack developer at Company B", "event": "UPDATE", "old_memory": "The user works as a software engineer at Company A, primarily responsible for front-end development" - }, - { + }}, + {{ "id": "648", "text": "The user works as a senior full-stack developer at Company B", "event": "UPDATE", "old_memory": "The user is responsible for front-end development of software at Company A" - }, - { + }}, + {{ "id": "7210", "text": "The user works as a senior full-stack developer at Company B", "event": "UPDATE", "old_memory": "The user is responsible for front-end development of software at Company A" - }, - { + }}, + {{ "id": "908", "text": "The user enjoys fishing with friends on weekends", "event": "NONE" - } + }} ] -} +}} Example2: Current Memories: -{ +{{ "memory": [ - { + {{ "id": "123", "text": "The user works as a software engineer in Company A, mainly responsible for front-end development" - }, - { + }}, + {{ "id": "908", "text": "The user likes to go fishing with friends on weekends" - } + }} ] -} +}} + +The background of the new fact being put forward: +user: Guess where I live？ +assistant: Hehuan Community. +user feedback: Wrong, update my address: Mingyue Community, Chaoyang District, Beijing Newly facts: "The user's residential address is Mingyue Community, Chaoyang District, Beijing" Operation recommendations: -{ +{{ "operation": [ - { + {{ "id": "123", "text": "The user works as a software engineer at Company A, primarily responsible for front-end development", "event": "NONE" - }, - { + }}, + {{ "id": "908", "text": "The user enjoys fishing with friends on weekends", "event": "NONE" - }, - { + }}, + {{ "id": "4567", "text": "The user's residential address is Mingyue Community, Chaoyang District, Beijing", "event": "ADD" - } + }} ] -} +}} -Current Memories +**Current Memories** {current_memories} -Newly facts: +**The background of the new fact being put forward** +{chat_history} + +**Newly facts** {new_facts} Operation recommendations: @@ -328,18 +370,18 @@ 你必须严格按照以下JSON格式返回响应： -{ +{{ "operation": [ - { + {{ "id": "<记忆ID>", "text": "<记忆内容>", "event": "<操作类型，必须是 "ADD", "UPDATE", "NONE" 之一>", "old_memory": "<原记忆内容，仅当操作为"UPDATE"时需要提供>" - }, + }}, ... ] -} +}} 要求： 1. 如果新事实对现有记忆item没有额外补充，现有记忆的信息可以覆盖新事实，设置操作为"NONE" @@ -360,113 +402,127 @@ 示例1： 现有记忆记录： -{ +{{ "memory": [ - { + {{ "id": "0911", "text": "用户是高级全栈开发工程师，在B公司工作" - }, - { + }}, + {{ "id": "123", "text": "用户在公司A担任软件工程师，主要负责前端开发" - }, - { + }}, + {{ "id": "648", "text": "用户在公司A负责软件的前端开发工作" - }, - { + }}, + {{ "id": "7210", "text": "用户在公司A负责软件的前端开发工作" - }, - { + }}, + {{ "id": "908", "text": "用户周末喜欢和朋友一起钓鱼" - } + }} ] -} +}} + +提出新事实的背景： +user: 你还记得我现在在哪里工作吗？ +assistant: A公司 +user feedback: 实际上，我在公司B工作，是一名高级全栈开发人员。 + 新获取的事实： "用户现在在公司B担任高级全栈开发工程师" 操作建议： -{ +{{ "operation": [ - { + {{ "id": "0911", "text": "用户是高级全栈开发工程师，在B公司工作", "event": "NONE" - }, - { + }}, + {{ "id": "123", "text": "用户现在在公司B担任高级全栈开发工程师", "event": "UPDATE", "old_memory": "用户在公司A担任软件工程师，主要负责前端开发" - }, - { + }}, + {{ "id": "648", "text": "用户现在在公司B担任高级全栈开发工程师", "event": "UPDATE", "old_memory": "用户在公司A负责软件的前端开发工作" - }, - { + }}, + {{ "id": "7210", "text": "用户现在在公司B担任高级全栈开发工程师", "event": "UPDATE", "old_memory": "用户在公司A负责软件的前端开发工作" - }, - { + }}, + {{ "id": "908", "text": "用户周末喜欢和朋友一起钓鱼", "event": "NONE" - } + }} ] -} +}} 示例2： 现有记忆记录： -{ +{{ "memory": [ - { + {{ "id": "123", "text": "用户在公司A担任软件工程师，主要负责前端开发" - }, - { + }}, + {{ "id": "908", "text": "用户周末喜欢和朋友一起钓鱼" - } + }} ] -} +}} + +提出新事实的背景： +user: 猜猜我住在哪里？ +assistant: 合欢社区 +user feedback: 错了，请更新我的地址：北京市朝阳区明月社区 新获取的事实： "用户的居住地址是北京市朝阳区明月小区" 操作建议： -{ +{{ "operation": [ - { + {{ "id": "123", "text": "用户在公司A担任软件工程师，主要负责前端开发", "event": "NONE" - }, - { + }}, + {{ "id": "908", "text": "用户周末喜欢和朋友一起钓鱼", "event": "NONE" - }, - { + }}, + {{ "id": "4567", "text": "用户的居住地址是北京市朝阳区明月小区", "event": "ADD" - } + }} ] -} +}} -现有记忆记录： +**现有记忆记录：** {current_memories} -新获取的事实： +**提出新事实的背景：** +{chat_history} + +**新获取的事实：** {new_facts} 操作建议： @@ -477,17 +533,17 @@ You must strictly return the response in the following JSON format: -{ +{{ "operation": [ - { + {{ "id": "", "text": "", "event": "", "old_memory": "" - }, + }}, ... ] -} +}} *Requirements*: 1. If the new fact provides no additional supplement to existing memory, set operation to "NONE" @@ -507,43 +563,43 @@ Example: Current Memories: -{ +{{ "memory": [ - { + {{ "id": "123", "text": "The user works as a software engineer in Company A, mainly responsible for front-end development" - }, - { + }}, + {{ "id": "908", "text": "The user likes to go fishing with friends on weekends" - } + }} ] -} +}} Newly facts: ["The user is currently working as a senior full-stack development engineer at Company B", "The user's residential address is Mingyue Community, Chaoyang District, Beijing", "The user goes fishing on weekends"] Operation recommendations: -{ +{{ "operation": [ - { + {{ "id": "123", "text": "The user is currently working as a senior full-stack development engineer at Company B", "event": "UPDATE", "old_memory": "The user works as a software engineer in Company A, mainly responsible for front-end development" - }, - { + }}, + {{ "id": "4567", "text": "The user's residential address is Mingyue Community, Chaoyang District, Beijing", "event": "ADD" - }, - { + }}, + {{ "id": "908", "text": "The user likes to go fishing with friends on weekends", "event": "NONE" - } + }} ] -} +}} Current Memories {current_memories} @@ -559,17 +615,17 @@ 你必须严格按照以下JSON格式返回响应： -{ +{{ "operation": [ - { + {{ "id": "<记忆ID>", "text": "<记忆内容>", "event": "<操作类型，必须是 "ADD", "UPDATE", "NONE" 之一>", "old_memory": "<原记忆内容，仅当操作为"UPDATE"时需要提供>" - }, + }}, ... ] -} +}} 要求： 1. 如果新事实对现有记忆没有额外补充，设置操作为"NONE" @@ -589,43 +645,43 @@ 示例： 现有记忆记录： -{ +{{ "memory": [ - { + {{ "id": "123", "text": "用户在公司A担任软件工程师，主要负责前端开发" - }, - { + }}, + {{ "id": "908", "text": "用户周末喜欢和朋友一起钓鱼" - } + }} ] -} +}} 新获取的事实： ["用户现在在公司B担任高级全栈开发工程师", "用户的居住地址是北京市朝阳区明月小区", "用户在周末会去钓鱼"] 操作建议： -{ +{{ "operation": [ - { + {{ "id": "123", "text": "用户在公司B担任高级全栈开发工程师", "event": "UPDATE", "old_memory": "用户在公司A担任软件工程师，主要负责前端开发" - }, - { + }}, + {{ "id": "4567", "text": "用户的居住地址是北京市朝阳区明月小区", "event": "ADD" - }, - { + }}, + {{ "id": "908", "text": "用户周末喜欢和朋友一起钓鱼", "event": "NONE" - } + }} ] -} +}} 现有记忆记录： {current_memories} From 0fa9be7e66fb7e3cd3d8abf0425393dceeacffc4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Thu, 27 Nov 2025 18:17:06 +0800 Subject: [PATCH 23/72] add threshold --- src/memos/mem_feedback/feedback.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index 6b8520c50..54d29ac88 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -263,7 +263,7 @@ def _add_or_update( def _vec_query(self, new_memories_embedding: list[float], user_name=None): retrieved_ids = self.graph_store.search_by_embedding( - new_memories_embedding, user_name=user_name, top_k=5 + new_memories_embedding, user_name=user_name, top_k=10, threshold=0.75 ) current_memories = [self.graph_store.get_node(item["id"]) for item in retrieved_ids] if not retrieved_ids: From facb7b31d17fccaae3f8e5572fa00ebc07976a4d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Thu, 27 Nov 2025 19:03:34 +0800 Subject: [PATCH 24/72] update prompt --- src/memos/mem_feedback/feedback.py | 4 ++-- src/memos/templates/mem_feedback_prompts.py | 25 ++++++++++++--------- 2 files changed, 16 insertions(+), 13 deletions(-) diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index 54d29ac88..dd5dbde4c 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -194,7 +194,7 @@ def _add_or_update( operations = self._id_dehallucination(operations, current_memories) else: operations = [{"event": "ADD"}] - + # TODO based on the operation, change memory_item memory info logger.info(f"[Feedback memory operations]: {operations!s}") if not operations: @@ -263,7 +263,7 @@ def _add_or_update( def _vec_query(self, new_memories_embedding: list[float], user_name=None): retrieved_ids = self.graph_store.search_by_embedding( - new_memories_embedding, user_name=user_name, top_k=10, threshold=0.75 + new_memories_embedding, user_name=user_name, top_k=10, threshold=0.7 ) current_memories = [self.graph_store.get_node(item["id"]) for item in retrieved_ids] if not retrieved_ids: diff --git a/src/memos/templates/mem_feedback_prompts.py b/src/memos/templates/mem_feedback_prompts.py index 11e17e399..85fbc6490 100644 --- a/src/memos/templates/mem_feedback_prompts.py +++ b/src/memos/templates/mem_feedback_prompts.py @@ -230,13 +230,14 @@ - Add operation: Generate a new unique ID in the format of a 4-digit string (e.g., "0001", "0002", etc.) *Important Requirements*: -- For update operations, you must provide the old_memory field to show the original content -- Compare the existing memories one by one and do not miss any content that needs to be updated. When multiple existing memories need to be updated, include all relevant entries in the operation list +1. For "UPDATE" operations, you must provide the old_memory field to display the original content +2. Compare existing memories one by one and do not omit any content requiring updates. When multiple existing memories need updating, include all relevant entries in the operation list +3. "text" field requirements: + - Use concise, complete declarative sentences, avoiding redundant information + - "text" should record the final adopted memory: if judged as "ADD", output text as "new fact"; if judged as "UPDATE", output text as "adjusted new fact"; if judged as "NONE", output text as "existing memory" +4. Both text and old_memory content should be in English +5. Return only the JSON format response, without any other content -If the new fact contradicts existing memory in key information (such as time, location, status, etc.), update ALL affected original memories based on the new fact and set operation to "UPDATE" for each one. Multiple memories covering the same outdated information should all be updated. -- Return only the JSON format response, without any other content -- text field requirements: Use concise, complete declarative sentences that are consistent with the newly acquired factual information, avoiding redundant information -- text and old_memory content should be in English Example1: Current Memories: @@ -394,11 +395,13 @@ - 新增操作：生成新的唯一ID，格式为4位数字字符串（如："0001", "0002"等） 重要要求： -- 对于更新操作，必须提供old_memory字段显示原内容 -- 对现有记忆逐一比对，不可漏掉需要更新的内容。当多个现有记忆需要更新时，将所有的相关条目都包含在操作列表中 -- 只返回JSON格式的响应，不要包含其他任何内容 -- text字段要求：使用简洁、完整的陈述句，和新获取的事实信息一致，避免冗余信息 -- text和old_memory内容使用中文 +1. 对于"UPDATE"更新操作，必须提供old_memory字段显示原内容 +2. 对现有记忆逐一比对，不可漏掉需要更新的内容。当多个现有记忆需要更新时，将所有的相关条目都包含在操作列表中 +3. text字段要求： + - 使用简洁、完整的陈述句，避免冗余信息 + - text要记录最终采用的记忆，如果判为"ADD"，则text输出为"新事实"；如果判为"UPDATE"，则text输出为"调整后的新事实"；如果判为"NONE"，则text输出为"现有记忆" +4. text和old_memory内容使用中文 +5. 只返回JSON格式的响应，不要包含其他任何内容 示例1： 现有记忆记录： From eab5fe6d4c62c1f6b37ca100f4993ce551c6c605 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Thu, 27 Nov 2025 19:05:30 +0800 Subject: [PATCH 25/72] update prompt --- src/memos/reranker/factory.py | 1 + 1 file changed, 1 insertion(+) diff --git a/src/memos/reranker/factory.py b/src/memos/reranker/factory.py index d2c50ba5e..1440704a6 100644 --- a/src/memos/reranker/factory.py +++ b/src/memos/reranker/factory.py @@ -2,6 +2,7 @@ from __future__ import annotations import json + from typing import TYPE_CHECKING, Any # Import singleton decorator From 7577aac491385e1063c53dca461856a5bdcfea53 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Thu, 27 Nov 2025 19:16:33 +0800 Subject: [PATCH 26/72] fix handler --- src/memos/api/handlers/feedback_handler.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/memos/api/handlers/feedback_handler.py b/src/memos/api/handlers/feedback_handler.py index 2e8a21cb1..8bf4db906 100644 --- a/src/memos/api/handlers/feedback_handler.py +++ b/src/memos/api/handlers/feedback_handler.py @@ -25,7 +25,7 @@ def __init__(self, dependencies: HandlerDependencies): dependencies: HandlerDependencies instance """ super().__init__(dependencies) - self._validate_dependencies("feedback_server", "mem_reader") + self._validate_dependencies("mem_reader") def handle_feedback_memories(self, feedback_req: APIFeedbackRequest) -> MemoryResponse: """ From cc4069da44a0321c2e0d9d197cbda9adc4c7e6be Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Sat, 29 Nov 2025 18:32:50 +0800 Subject: [PATCH 27/72] add feedback scheduler --- src/memos/api/handlers/component_init.py | 22 +- src/memos/api/handlers/feedback_handler.py | 65 ++- src/memos/api/product_models.py | 23 +- src/memos/llms/openai.py | 1 + src/memos/mem_feedback/feedback.py | 123 ++--- src/memos/mem_scheduler/base_scheduler.py | 2 + src/memos/mem_scheduler/general_scheduler.py | 59 +++ .../mem_scheduler/schemas/general_schemas.py | 1 + src/memos/multi_mem_cube/composite_cube.py | 12 +- src/memos/multi_mem_cube/single_cube.py | 44 +- src/memos/templates/mem_feedback_prompts.py | 425 +----------------- 11 files changed, 254 insertions(+), 523 deletions(-) diff --git a/src/memos/api/handlers/component_init.py b/src/memos/api/handlers/component_init.py index 9413ec2c3..982487b18 100644 --- a/src/memos/api/handlers/component_init.py +++ b/src/memos/api/handlers/component_init.py @@ -293,6 +293,15 @@ def init_server() -> dict[str, Any]: ) logger.debug("Searcher created") + # Initialize feedback server + feedback_server = SimpleMemFeedback( + llm=llm, + embedder=embedder, + graph_store=graph_db, + memory_manager=memory_manager, + mem_reader=mem_reader, + ) + # Initialize Scheduler scheduler_config_dict = APIConfig.get_scheduler_config() scheduler_config = SchedulerConfigFactory( @@ -306,7 +315,9 @@ def init_server() -> dict[str, Any]: mem_reader=mem_reader, redis_client=redis_client, ) - mem_scheduler.init_mem_cube(mem_cube=naive_mem_cube, searcher=searcher) + mem_scheduler.init_mem_cube( + mem_cube=naive_mem_cube, searcher=searcher, feedback_server=feedback_server + ) logger.debug("Scheduler initialized") # Initialize SchedulerAPIModule @@ -327,15 +338,6 @@ def init_server() -> dict[str, Any]: online_bot = get_online_bot_function() if dingding_enabled else None logger.info("DingDing bot is enabled") - # Initialize feedback server - feedback_server = SimpleMemFeedback( - llm=llm, - embedder=embedder, - graph_store=graph_db, - memory_manager=memory_manager, - mem_reader=mem_reader, - ) - deepsearch_agent = DeepSearchMemAgent( llm=llm, memory_retriever=tree_mem, diff --git a/src/memos/api/handlers/feedback_handler.py b/src/memos/api/handlers/feedback_handler.py index 8bf4db906..2d3dc4ee2 100644 --- a/src/memos/api/handlers/feedback_handler.py +++ b/src/memos/api/handlers/feedback_handler.py @@ -5,6 +5,9 @@ from memos.api.handlers.base_handler import BaseHandler, HandlerDependencies from memos.api.product_models import APIFeedbackRequest, MemoryResponse from memos.log import get_logger +from memos.multi_mem_cube.composite_cube import CompositeCubeView +from memos.multi_mem_cube.single_cube import SingleCubeView +from memos.multi_mem_cube.views import MemCubeView logger = get_logger(__name__) @@ -25,7 +28,7 @@ def __init__(self, dependencies: HandlerDependencies): dependencies: HandlerDependencies instance """ super().__init__(dependencies) - self._validate_dependencies("mem_reader") + self._validate_dependencies("mem_reader", "mem_scheduler") def handle_feedback_memories(self, feedback_req: APIFeedbackRequest) -> MemoryResponse: """ @@ -37,16 +40,54 @@ def handle_feedback_memories(self, feedback_req: APIFeedbackRequest) -> MemoryRe Returns: MemoryResponse with formatted results """ - process_record = self.feedback_server.process_feedback( - user_name=feedback_req.mem_cube_id, - session_id=feedback_req.session_id, - chat_history=feedback_req.history, - retrieved_memory_ids=feedback_req.retrieved_memory_ids, - feedback_content=feedback_req.feedback_content, - feedback_time=feedback_req.feedback_time, - allow_knowledgebase_write=feedback_req.allow_knowledgebase_write, - sync_mode=feedback_req.sync_mode, - corrected_answer=feedback_req.corrected_answer, + cube_view = self._build_cube_view(feedback_req) + + process_record = cube_view.feedback_memories(feedback_req) + + self.logger.info(f"[FeedbackHandler] Final feedback results count={len(process_record)}") + + return MemoryResponse( + message="Memory feedback successfully", + data=[process_record], ) - return MemoryResponse(message="Feedback process successfully", data=[process_record]) + def _resolve_cube_ids(self, feedback_req: APIFeedbackRequest) -> list[str]: + """ + Normalize target cube ids from feedback_req. + """ + if feedback_req.writable_cube_ids: + return list(dict.fromkeys(feedback_req.writable_cube_ids)) + + return [feedback_req.user_id] + + def _build_cube_view(self, feedback_req: APIFeedbackRequest) -> MemCubeView: + cube_ids = self._resolve_cube_ids(feedback_req) + + if len(cube_ids) == 1: + cube_id = cube_ids[0] + return SingleCubeView( + cube_id=cube_id, + naive_mem_cube=None, + mem_reader=None, + mem_scheduler=self.mem_scheduler, + logger=self.logger, + searcher=None, + feedback_server=self.feedback_server, + ) + else: + single_views = [ + SingleCubeView( + cube_id=cube_id, + naive_mem_cube=None, + mem_reader=None, + mem_scheduler=self.mem_scheduler, + logger=self.logger, + searcher=None, + feedback_server=self.feedback_server, + ) + for cube_id in cube_ids + ] + return CompositeCubeView( + cube_views=single_views, + logger=self.logger, + ) diff --git a/src/memos/api/product_models.py b/src/memos/api/product_models.py index 8035ff3f8..52d5fe6f6 100644 --- a/src/memos/api/product_models.py +++ b/src/memos/api/product_models.py @@ -625,29 +625,30 @@ class APIFeedbackRequest(BaseRequest): """Request model for processing feedback info.""" user_id: str = Field(..., description="User ID") - mem_cube_id: str | None = Field(..., description="Cube ID to use for chat") session_id: str | None = Field( "default_session", description="Session ID for soft-filtering memories" ) - agent_id: str = Field(None, description="Agent ID") - app_id: str = Field(None, description="App ID") + task_id: str | None = Field(None, description="Task ID for monitering async tasks") history: list[MessageDict] | None = Field(..., description="Chat history") retrieved_memory_ids: list[str] | None = Field( None, description="Retrieved memory ids at last turn" ) feedback_content: str | None = Field(..., description="Feedback content to process") feedback_time: str | None = Field(None, description="Feedback time") - allow_public: bool = Field( - False, description="Whether to allow writing to the public memory repository" - ) - allow_knowledgebase_write: bool = Field( - False, description="Whether to allow writing into the user memory repository" - ) - allow_knowledgebase_ids: bool = Field( - False, description="Write to the specified memory repository ID" + # ==== Multi-cube writing ==== + writable_cube_ids: list[str] | None = Field( + None, description="List of cube IDs user can write for multi-cube add" ) sync_mode: Literal["sync", "async"] = Field("async", description="feedback mode: sync or async") corrected_answer: bool = Field(False, description="Whether need return corrected answer") + # ==== Backward compatibility ==== + mem_cube_id: str | None = Field( + None, + description=( + "(Deprecated) Single cube ID to search in. " + "Prefer `readable_cube_ids` for multi-cube search." + ), + ) class APIChatCompleteRequest(BaseRequest): diff --git a/src/memos/llms/openai.py b/src/memos/llms/openai.py index 9b348adcf..19d7a60fe 100644 --- a/src/memos/llms/openai.py +++ b/src/memos/llms/openai.py @@ -39,6 +39,7 @@ def generate(self, messages: MessageList, **kwargs) -> str: top_p=kwargs.get("top_p", self.config.top_p), extra_body=kwargs.get("extra_body", self.config.extra_body), tools=kwargs.get("tools", NOT_GIVEN), + timeout=kwargs.get("timeout", 30), ) logger.info(f"Response from OpenAI: {response.model_dump_json()}") tool_calls = getattr(response.choices[0].message, "tool_calls", None) diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index dd5dbde4c..bf668852f 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -114,7 +114,7 @@ def _feedback_judgement( return [] def _feedback_memory( - self, user_name: str, feedback_memories: list[TextualMemoryItem], **kwargs + self, user_id: str, user_name: str, feedback_memories: list[TextualMemoryItem], **kwargs ) -> dict: sync_mode = kwargs.get("sync_mode") retrieved_memory_ids = kwargs.get("retrieved_memory_ids") or [] @@ -190,10 +190,11 @@ def _add_or_update( chat_history=fact_history, ) - operations = self._get_llm_response(prompt).get("operation", []) + operations = self._get_llm_response(prompt).get("operations", []) operations = self._id_dehallucination(operations, current_memories) else: - operations = [{"event": "ADD"}] + operations = [{"operation": "ADD"}] + # TODO based on the operation, change memory_item memory info logger.info(f"[Feedback memory operations]: {operations!s}") @@ -206,7 +207,7 @@ def _add_or_update( with ContextThreadPoolExecutor(max_workers=10) as executor: future_to_op = {} for op in operations: - event_type = op.get("event", "").lower() + event_type = op.get("operation", "").lower() if event_type == "add": future = executor.submit( @@ -290,7 +291,7 @@ def _vec_query(self, new_memories_embedding: list[float], user_name=None): def _get_llm_response(self, prompt: str, dsl: bool = True) -> dict: messages = [{"role": "user", "content": prompt}] try: - response_text = self.llm.generate(messages, temperature=0.3) + response_text = self.llm.generate(messages, temperature=0.3, timeout=60) if dsl: response_text = response_text.replace("```", "").replace("json", "") response_json = json.loads(response_text) @@ -306,7 +307,7 @@ def _id_dehallucination(self, operations, current_memories): right_lower_map = {x.lower(): x for x in right_ids} def correct_item(data): - if data.get("event", "").lower() != "update": + if data.get("operation", "").lower() != "update": return data original_id = data["id"] @@ -348,6 +349,7 @@ def _generate_answer( def process_feedback_core( self, + user_id: str, user_name: str, chat_history: list[MessageDict], feedback_content: str, @@ -370,11 +372,10 @@ def check_validity(item): try: feedback_time = kwargs.get("feedback_time") or datetime.now().isoformat() session_id = kwargs.get("session_id") - allow_knowledgebase_write = bool(kwargs.get("allow_knowledgebase_write")) - if feedback_content.strip() == "" or not allow_knowledgebase_write: + if feedback_content.strip() == "": return {"record": {"add": [], "update": []}} - info = {"user_id": user_name, "session_id": session_id} + info = {"user_id": user_id, "user_name": user_name, "session_id": session_id} logger.info( f"[Feedback Core: process_feedback_core] Starting memory feedback process for user {user_name}" ) @@ -435,6 +436,7 @@ def check_validity(item): embedding=embedding, usage=[], sources=[{"type": "chat"}], + user_name=user_name, background="", confidence=0.99, type="fine", @@ -443,6 +445,7 @@ def check_validity(item): ) mem_record = self._feedback_memory( + user_id, user_name, feedback_memories, chat_history=chat_history, @@ -460,6 +463,7 @@ def check_validity(item): def process_feedback( self, + user_id: str, user_name: str, chat_history: list[MessageDict], feedback_content: str, @@ -469,7 +473,7 @@ def process_feedback( Process feedback with different modes. Args: - user_name: User identifier + user_name: cube_ids chat_history: List of chat messages feedback_content: Feedback content from user **kwargs: Additional arguments including sync_mode @@ -477,82 +481,45 @@ def process_feedback( Returns: Dict with answer and/or memory operation records """ - sync_mode = kwargs.get("sync_mode") - corrected_answer = kwargs.get("corrected_answer") - - if sync_mode == "sync": - with ContextThreadPoolExecutor(max_workers=2) as ex: - answer_future = ex.submit( - self._generate_answer, - chat_history, - feedback_content, - corrected_answer=corrected_answer, - ) - core_future = ex.submit( - self.process_feedback_core, - user_name, - chat_history, - feedback_content, - **kwargs, - ) - done, pending = concurrent.futures.wait([answer_future, core_future], timeout=30) - for fut in pending: - fut.cancel() - try: - answer = answer_future.result() - record = core_future.result() - logger.info( - f"[MemFeedback sync] Completed concurrently for user {user_name} with full results." - ) - return {"answer": answer, "record": record["record"]} - except concurrent.futures.TimeoutError: - logger.error( - f"[MemFeedback sync] Timeout in sync mode for {user_name}", exc_info=True - ) - return {"answer": "", "record": {"add": [], "update": []}} - except Exception as e: - logger.error( - f"[MemFeedback sync] Error in concurrent tasks for {user_name}: {e}", - exc_info=True, - ) - return {"answer": "", "record": {"add": [], "update": []}} - else: - answer = self._generate_answer( - chat_history, feedback_content, corrected_answer=corrected_answer - ) + corrected_answer = kwargs.get("corrected_answer", False) - ex = ContextThreadPoolExecutor(max_workers=1) - future = ex.submit( + with ContextThreadPoolExecutor(max_workers=2) as ex: + answer_future = ex.submit( + self._generate_answer, + chat_history, + feedback_content, + corrected_answer=corrected_answer, + ) + core_future = ex.submit( self.process_feedback_core, + user_id, user_name, chat_history, feedback_content, **kwargs, ) - ex.shutdown(wait=False) - - def log_completion(f): - try: - result = f.result(timeout=600) - logger.info(f"[MemFeedback async] Completed for {user_name}: {result}") - except concurrent.futures.TimeoutError: - logger.error( - f"[MemFeedback async] Background task timeout for {user_name}", - exc_info=True, - ) - f.cancel() - except Exception as e: - logger.error( - f"[MemFeedback async] Background Feedback Error for {user_name}: {e}", - exc_info=True, - ) - - future.add_done_callback(log_completion) + done, pending = concurrent.futures.wait([answer_future, core_future], timeout=30) + for fut in pending: + fut.cancel() + try: + answer = answer_future.result() + record = core_future.result() + logger.info( + f"[MemFeedback process] Completed concurrently for user {user_name} with full results." + ) - logger.info( - f"[MemFeedback async] Returned answer, background task started for {user_name}." - ) - return {"answer": answer, "record": {"add": [], "update": []}} + return {"answer": answer, "record": record["record"]} + except concurrent.futures.TimeoutError: + logger.error( + f"[MemFeedback process] Timeout in sync mode for {user_name}", exc_info=True + ) + return {"answer": "", "record": {"add": [], "update": []}} + except Exception as e: + logger.error( + f"[MemFeedback process] Error in concurrent tasks for {user_name}: {e}", + exc_info=True, + ) + return {"answer": "", "record": {"add": [], "update": []}} # Helper for DB operations with retry @retry(stop=stop_after_attempt(3), wait=wait_exponential(multiplier=1, min=4, max=10)) diff --git a/src/memos/mem_scheduler/base_scheduler.py b/src/memos/mem_scheduler/base_scheduler.py index f641fc442..731655753 100644 --- a/src/memos/mem_scheduler/base_scheduler.py +++ b/src/memos/mem_scheduler/base_scheduler.py @@ -158,6 +158,7 @@ def init_mem_cube( self, mem_cube: BaseMemCube, searcher: Searcher | None = None, + feedback_server: Searcher | None = None, ): self.mem_cube = mem_cube self.text_mem: TreeTextMemory = self.mem_cube.text_mem @@ -170,6 +171,7 @@ def init_mem_cube( ) else: self.searcher = searcher + self.feedback_server = feedback_server def initialize_modules( self, diff --git a/src/memos/mem_scheduler/general_scheduler.py b/src/memos/mem_scheduler/general_scheduler.py index ac2ea2bfa..4190b8e5c 100644 --- a/src/memos/mem_scheduler/general_scheduler.py +++ b/src/memos/mem_scheduler/general_scheduler.py @@ -13,6 +13,7 @@ ANSWER_LABEL, DEFAULT_MAX_QUERY_KEY_WORDS, LONG_TERM_MEMORY_TYPE, + MEM_FEEDBACK_LABEL, MEM_ORGANIZE_LABEL, MEM_READ_LABEL, NOT_APPLICABLE_TYPE, @@ -56,6 +57,7 @@ def __init__(self, config: GeneralSchedulerConfig): MEM_READ_LABEL: self._mem_read_message_consumer, MEM_ORGANIZE_LABEL: self._mem_reorganize_message_consumer, PREF_ADD_LABEL: self._pref_add_message_consumer, + MEM_FEEDBACK_LABEL: self._mem_feedback_message_consumer, } self.dispatcher.register_handlers(handlers) @@ -359,6 +361,63 @@ def _add_message_consumer(self, messages: list[ScheduleMessageItem]) -> None: except Exception as e: logger.error(f"Error: {e}", exc_info=True) + def _mem_feedback_message_consumer(self, messages: list[ScheduleMessageItem]) -> None: + try: + message = messages[0] + mem_cube = self.current_mem_cube + + user_id = message.user_id + mem_cube_id = message.mem_cube_id + content = message.content + + feedback_data = json.loads(content) + + feedback_result = self.feedback_server.process_feedback( + user_id=user_id, + user_name=mem_cube_id, + session_id=feedback_data["session_id"], + chat_history=feedback_data["history"], + retrieved_memory_ids=feedback_data["retrieved_memory_ids"], + feedback_content=feedback_data["feedback_content"], + feedback_time=feedback_data["feedback_time"], + ) + + logger.info( + f"Successfully feedback memories for user_id={user_id}, mem_cube_id={mem_cube_id}" + ) + + should_send_log = ( + self.rabbitmq_config is not None + and hasattr(self.rabbitmq_config, "exchange_type") + and self.rabbitmq_config.exchange_type == "direct" + ) + if feedback_result and should_send_log: + feedback_content = [] + for _i, mem_item in enumerate(feedback_result): + feedback_content.append( + { + "content": mem_item.memory, + "id": mem_item["id"], + } + ) + event = self.create_event_log( + label="feedbackMemory", + from_memory_type=USER_INPUT_TYPE, + to_memory_type=LONG_TERM_MEMORY_TYPE, + user_id=user_id, + mem_cube_id=mem_cube_id, + mem_cube=mem_cube, + memcube_log_content=feedback_content, + metadata=[], + memory_len=len(feedback_content), + memcube_name=self._map_memcube_name(mem_cube_id), + ) + event.task_id = message.task_id + self._submit_web_logs([event]) + + except Exception as e: + logger.error(f"Error processing feedbackMemory message: {e}", exc_info=True) + def _mem_read_message_consumer(self, messages: list[ScheduleMessageItem]) -> None: logger.info(f"Messages {messages} assigned to {MEM_READ_LABEL} handler.") diff --git a/src/memos/mem_scheduler/schemas/general_schemas.py b/src/memos/mem_scheduler/schemas/general_schemas.py index 91d442720..e76728286 100644 --- a/src/memos/mem_scheduler/schemas/general_schemas.py +++ b/src/memos/mem_scheduler/schemas/general_schemas.py @@ -13,6 +13,7 @@ MEM_ARCHIVE_LABEL = "mem_archive" API_MIX_SEARCH_LABEL = "api_mix_search" PREF_ADD_LABEL = "pref_add" +MEM_FEEDBACK_LABEL = "mem_feedback" TreeTextMemory_SEARCH_METHOD = "tree_text_memory_search" TreeTextMemory_FINE_SEARCH_METHOD = "tree_text_memory_fine_search" diff --git a/src/memos/multi_mem_cube/composite_cube.py b/src/memos/multi_mem_cube/composite_cube.py index 8f892d60d..6db6ca3d7 100644 --- a/src/memos/multi_mem_cube/composite_cube.py +++ b/src/memos/multi_mem_cube/composite_cube.py @@ -7,7 +7,7 @@ if TYPE_CHECKING: - from memos.api.product_models import APIADDRequest, APISearchRequest + from memos.api.product_models import APIADDRequest, APIFeedbackRequest, APISearchRequest from memos.multi_mem_cube.single_cube import SingleCubeView @@ -61,3 +61,13 @@ def search_memories(self, search_req: APISearchRequest) -> dict[str, Any]: merged_results["pref_note"] = note return merged_results + + def feedback_memories(self, feedback_req: APIFeedbackRequest) -> list[dict[str, Any]]: + all_results: list[dict[str, Any]] = [] + + for view in self.cube_views: + self.logger.info(f"[CompositeCubeView] fan-out add to cube={view.cube_id}") + results = view.feedback_memories(feedback_req) + all_results.extend(results) + + return all_results diff --git a/src/memos/multi_mem_cube/single_cube.py b/src/memos/multi_mem_cube/single_cube.py index 2b79a416c..21d4fd8d1 100644 --- a/src/memos/multi_mem_cube/single_cube.py +++ b/src/memos/multi_mem_cube/single_cube.py @@ -16,6 +16,7 @@ from memos.log import get_logger from memos.mem_scheduler.schemas.general_schemas import ( ADD_LABEL, + MEM_FEEDBACK_LABEL, MEM_READ_LABEL, PREF_ADD_LABEL, ) @@ -34,7 +35,7 @@ if TYPE_CHECKING: - from memos.api.product_models import APIADDRequest, APISearchRequest + from memos.api.product_models import APIADDRequest, APIFeedbackRequest, APISearchRequest @dataclass @@ -45,6 +46,7 @@ class SingleCubeView(MemCubeView): mem_scheduler: Any logger: Any searcher: Any + feedback_server: Any | None = None deepsearch_agent: Any | None = None def add_memories(self, add_req: APIADDRequest) -> list[dict[str, Any]]: @@ -131,6 +133,46 @@ def search_memories(self, search_req: APISearchRequest) -> dict[str, Any]: self.logger.info(f"Search memories result: {memories_result}") return memories_result + def feedback_memories(self, feedback_req: APIFeedbackRequest) -> dict[str, Any]: + target_session_id = feedback_req.session_id or "default_session" + if feedback_req.sync_mode == "async": + try: + feedback_req_str = json.dumps(feedback_req.model_dump()) + message_item_feedback = ScheduleMessageItem( + user_id=feedback_req.user_id, + task_id=feedback_req.task_id, + session_id=target_session_id, + mem_cube_id=self.cube_id, + mem_cube=self.naive_mem_cube, + label=MEM_FEEDBACK_LABEL, + content=feedback_req_str, + timestamp=datetime.utcnow(), + ) + self.mem_scheduler.memos_message_queue.submit_messages( + messages=[message_item_feedback] + ) + self.logger.info(f"[SingleCubeView] cube={self.cube_id} Submitted FEEDBACK async") + except Exception as e: + self.logger.error( + f"[SingleCubeView] cube={self.cube_id} Failed to submit FEEDBACK: {e}", + exc_info=True, + ) + return [] + else: + feedback_result = self.feedback_server.process_feedback( + user_id=feedback_req.user_id, + user_name=self.cube_id, + session_id=feedback_req.session_id, + chat_history=feedback_req.history, + retrieved_memory_ids=feedback_req.retrieved_memory_ids, + feedback_content=feedback_req.feedback_content, + feedback_time=feedback_req.feedback_time, + sync_mode=feedback_req.sync_mode, + corrected_answer=feedback_req.corrected_answer, + ) + self.logger.info(f"Feedback memories result: {feedback_result}") + return feedback_result + def _get_search_mode(self, mode: str) -> str: """ Get search mode with environment variable fallback. diff --git a/src/memos/templates/mem_feedback_prompts.py b/src/memos/templates/mem_feedback_prompts.py index 85fbc6490..2e9a0e9c9 100644 --- a/src/memos/templates/mem_feedback_prompts.py +++ b/src/memos/templates/mem_feedback_prompts.py @@ -200,18 +200,17 @@ 输出：""" - -UPDATE_FORMER_MEMORIES = """Please analyze the newly acquired factual information and determine how this information should be updated to the memory database: add, update, or keep unchanged, and provide final operation recommendations. - +UPDATE_FORMER_MEMORIES = """Operation recommendations: +Please analyze the newly acquired factual information and determine how this information should be updated to the memory database: add, update, or keep unchanged, and provide final operation recommendations. You must strictly return the response in the following JSON format: {{ - "operation": + "operations": [ {{ "id": "", "text": "", - "event": "", + "operation": "", "old_memory": "" }}, ... @@ -221,10 +220,9 @@ *Requirements*: 1. If the new fact does not provide additional information to the existing memory item, the existing memory can override the new fact, and the operation is set to "NONE." 2. If the new fact is similar to existing memory but the information is more accurate, complete, or requires correction, set operation to "UPDATE" -3. If the new fact contradicts existing memory in key information (such as time, location, status, etc.), update the original memory based on the new fact and set operation to "UPDATE" +3. If the new fact contradicts existing memory in key information (such as time, location, status, etc.), update the original memory based on the new fact and set operation to "UPDATE", only modifying the relevant error segments in the existing memory paragraphs while keeping other text completely unchanged. 4. If there is no existing memory that requires updating, the new fact is added as entirely new information, and the operation is set to "ADD." Therefore, in the same operation list, ADD and UPDATE will not coexist. - *ID Management Rules*: - Update operation: Keep the original ID unchanged - Add operation: Generate a new unique ID in the format of a 4-digit string (e.g., "0001", "0002", etc.) @@ -235,125 +233,11 @@ 3. "text" field requirements: - Use concise, complete declarative sentences, avoiding redundant information - "text" should record the final adopted memory: if judged as "ADD", output text as "new fact"; if judged as "UPDATE", output text as "adjusted new fact"; if judged as "NONE", output text as "existing memory" + - When updating, ensure that only the related error segments are modified, and other text remains completely unchanged. 4. Both text and old_memory content should be in English 5. Return only the JSON format response, without any other content -Example1: -Current Memories: -{{ - "memory": [ - {{ - "id": "0911", - "text": "The user is a senior full-stack developer working at Company B" - }}, - {{ - "id": "123", - "text": "The user works as a software engineer at Company A, primarily responsible for front-end development" - }}, - {{ - "id": "648", - "text": "The user is responsible for front-end development of software at Company A" - }}, - {{ - "id": "7210", - "text": "The user is responsible for front-end development of software at Company A" - }}, - {{ - "id": "908", - "text": "The user enjoys fishing with friends on weekends" - }} - ] -}} - -The background of the new fact being put forward: -user: Do you remember where I work？ -assistant: Company A. -user feedback: I work at Company B, and I am a senior full-stack developer. - -Newly facts: -"The user works as a senior full-stack developer at Company B" - -Operation recommendations: -{{ - "operation": - [ - {{ - "id": "0911", - "text": "The user is a senior full-stack developer working at Company B", - "event": "NONE" - }}, - {{ - "id": "123", - "text": "The user works as a senior full-stack developer at Company B", - "event": "UPDATE", - "old_memory": "The user works as a software engineer at Company A, primarily responsible for front-end development" - }}, - {{ - "id": "648", - "text": "The user works as a senior full-stack developer at Company B", - "event": "UPDATE", - "old_memory": "The user is responsible for front-end development of software at Company A" - }}, - {{ - "id": "7210", - "text": "The user works as a senior full-stack developer at Company B", - "event": "UPDATE", - "old_memory": "The user is responsible for front-end development of software at Company A" - }}, - {{ - "id": "908", - "text": "The user enjoys fishing with friends on weekends", - "event": "NONE" - }} - ] -}} - -Example2: -Current Memories: -{{ - "memory": [ - {{ - "id": "123", - "text": "The user works as a software engineer in Company A, mainly responsible for front-end development" - }}, - {{ - "id": "908", - "text": "The user likes to go fishing with friends on weekends" - }} - ] -}} - -The background of the new fact being put forward: -user: Guess where I live？ -assistant: Hehuan Community. -user feedback: Wrong, update my address: Mingyue Community, Chaoyang District, Beijing - -Newly facts: -"The user's residential address is Mingyue Community, Chaoyang District, Beijing" - -Operation recommendations: -{{ - "operation": - [ - {{ - "id": "123", - "text": "The user works as a software engineer at Company A, primarily responsible for front-end development", - "event": "NONE" - }}, - {{ - "id": "908", - "text": "The user enjoys fishing with friends on weekends", - "event": "NONE" - }}, - {{ - "id": "4567", - "text": "The user's residential address is Mingyue Community, Chaoyang District, Beijing", - "event": "ADD" - }} - ] -}} - **Current Memories** {current_memories} @@ -366,7 +250,6 @@ Operation recommendations: """ - UPDATE_FORMER_MEMORIES_ZH = """请分析新获取的事实信息，并决定这些信息应该如何更新到记忆库中：新增、更新、或保持不变，并给出最终的操作建议。 你必须严格按照以下JSON格式返回响应： @@ -377,7 +260,7 @@ {{ "id": "<记忆ID>", "text": "<记忆内容>", - "event": "<操作类型，必须是 "ADD", "UPDATE", "NONE" 之一>", + "operation": "<操作类型，必须是 "ADD", "UPDATE", "NONE" 之一>", "old_memory": "<原记忆内容，仅当操作为"UPDATE"时需要提供>" }}, ... @@ -385,10 +268,10 @@ }} 要求： -1. 如果新事实对现有记忆item没有额外补充，现有记忆的信息可以覆盖新事实，设置操作为"NONE" -2. 如果新事实与现有记忆item相似但信息更准确、完整或需要修正，设置操作为"UPDATE" -3. 如果新事实与现有记忆在关键信息上矛盾（如时间、地点、状态等），以新事实为准更新原有记忆，设置操作为"UPDATE" -4. 如果现有记忆中没有需要更新的，则新事实作为全新信息添加，设置操作为"ADD"。因此可知同一个 operation 列表中，ADD和UPDATE不会同时存在。 +1. 若新事实未对现有记忆条目提供额外信息，现有记忆可覆盖新事实，操作设为"NONE" +2. 若新事实与现有记忆相似但信息更准确、完整或需修正，操作设为"UPDATE" +3. 若新事实在关键信息（如时间、地点、状态等）上与现有记忆矛盾，则根据新事实更新原记忆，操作设为"UPDATE"，仅修改现有记忆段落中的相关错误片段，其余文本完全保持不变 +4. 若无需要更新的现有记忆，则将新事实作为全新信息添加，操作设为"ADD"。因此在同一操作列表中，ADD与UPDATE不会同时存在 ID管理规则： - 更新操作：保持原有ID不变 @@ -400,296 +283,18 @@ 3. text字段要求： - 使用简洁、完整的陈述句，避免冗余信息 - text要记录最终采用的记忆，如果判为"ADD"，则text输出为"新事实"；如果判为"UPDATE"，则text输出为"调整后的新事实"；如果判为"NONE"，则text输出为"现有记忆" + - 更新时确保仅修改相关错误片段，其余文本完全保持不变 4. text和old_memory内容使用中文 5. 只返回JSON格式的响应，不要包含其他任何内容 -示例1： -现有记忆记录： -{{ - "memory": [ - {{ - "id": "0911", - "text": "用户是高级全栈开发工程师，在B公司工作" - }}, - {{ - "id": "123", - "text": "用户在公司A担任软件工程师，主要负责前端开发" - }}, - {{ - "id": "648", - "text": "用户在公司A负责软件的前端开发工作" - }}, - {{ - "id": "7210", - "text": "用户在公司A负责软件的前端开发工作" - }}, - {{ - "id": "908", - "text": "用户周末喜欢和朋友一起钓鱼" - }} - ] -}} - -提出新事实的背景： -user: 你还记得我现在在哪里工作吗？ -assistant: A公司 -user feedback: 实际上，我在公司B工作，是一名高级全栈开发人员。 - - -新获取的事实： -"用户现在在公司B担任高级全栈开发工程师" - -操作建议： -{{ - "operation": - [ - {{ - "id": "0911", - "text": "用户是高级全栈开发工程师，在B公司工作", - "event": "NONE" - }}, - {{ - "id": "123", - "text": "用户现在在公司B担任高级全栈开发工程师", - "event": "UPDATE", - "old_memory": "用户在公司A担任软件工程师，主要负责前端开发" - }}, - {{ - "id": "648", - "text": "用户现在在公司B担任高级全栈开发工程师", - "event": "UPDATE", - "old_memory": "用户在公司A负责软件的前端开发工作" - }}, - {{ - "id": "7210", - "text": "用户现在在公司B担任高级全栈开发工程师", - "event": "UPDATE", - "old_memory": "用户在公司A负责软件的前端开发工作" - }}, - {{ - "id": "908", - "text": "用户周末喜欢和朋友一起钓鱼", - "event": "NONE" - }} - ] -}} - -示例2： -现有记忆记录： -{{ - "memory": [ - {{ - "id": "123", - "text": "用户在公司A担任软件工程师，主要负责前端开发" - }}, - {{ - "id": "908", - "text": "用户周末喜欢和朋友一起钓鱼" - }} - ] -}} - -提出新事实的背景： -user: 猜猜我住在哪里？ -assistant: 合欢社区 -user feedback: 错了，请更新我的地址：北京市朝阳区明月社区 - -新获取的事实： -"用户的居住地址是北京市朝阳区明月小区" -操作建议： -{{ - "operation": - [ - {{ - "id": "123", - "text": "用户在公司A担任软件工程师，主要负责前端开发", - "event": "NONE" - }}, - {{ - "id": "908", - "text": "用户周末喜欢和朋友一起钓鱼", - "event": "NONE" - }}, - {{ - "id": "4567", - "text": "用户的居住地址是北京市朝阳区明月小区", - "event": "ADD" - }} - ] -}} - -**现有记忆记录：** +**当前记忆：** {current_memories} -**提出新事实的背景：** +**新事实提出的背景：** {chat_history} -**新获取的事实：** -{new_facts} - -操作建议： -""" - - -GROUP_UPDATE_FORMER_MEMORIES = """Please analyze the newly acquired factual information and determine how this information should be updated to the memory database: add, update, or keep unchanged, and provide final operation recommendations. - -You must strictly return the response in the following JSON format: - -{{ - "operation": [ - {{ - "id": "", - "text": "", - "event": "", - "old_memory": "" - }}, - ... - ] -}} - -*Requirements*: -1. If the new fact provides no additional supplement to existing memory, set operation to "NONE" -2. If the new fact is similar to existing memory but the information is more accurate, complete, or requires correction, set operation to "UPDATE" -3. If the new fact contradicts existing memory in key information (such as time, location, status, etc.), update the original memory based on the new fact and set operation to "UPDATE" -4. If there is completely new information to add, set operation to "ADD" - -*ID Management Rules*: -- Update operation: Keep the original ID unchanged -- Add operation: Generate a new unique ID in the format of a 4-digit string (e.g., "0001", "0002", etc.) - -*Important Requirements*: -- Return only the JSON format response, without any other content -- For update operations, you must provide the old_memory field to show the original content -- text field requirements: Use concise, complete declarative sentences that are consistent with the newly acquired factual information, avoiding redundant information -- text and old_memory content should be in English - -Example: -Current Memories: -{{ - "memory": [ - {{ - "id": "123", - "text": "The user works as a software engineer in Company A, mainly responsible for front-end development" - }}, - {{ - "id": "908", - "text": "The user likes to go fishing with friends on weekends" - }} - ] -}} - -Newly facts: -["The user is currently working as a senior full-stack development engineer at Company B", "The user's residential address is Mingyue Community, Chaoyang District, Beijing", "The user goes fishing on weekends"] - -Operation recommendations: -{{ - "operation": [ - {{ - "id": "123", - "text": "The user is currently working as a senior full-stack development engineer at Company B", - "event": "UPDATE", - "old_memory": "The user works as a software engineer in Company A, mainly responsible for front-end development" - }}, - {{ - "id": "4567", - "text": "The user's residential address is Mingyue Community, Chaoyang District, Beijing", - "event": "ADD" - }}, - {{ - "id": "908", - "text": "The user likes to go fishing with friends on weekends", - "event": "NONE" - }} - ] -}} - -Current Memories -{current_memories} - -Newly facts: -{new_facts} - -Operation recommendations: -""" - - -GROUP_UPDATE_FORMER_MEMORIES_ZH = """请分析新获取的事实信息，并决定这些信息应该如何更新到记忆库中：新增、更新、或保持不变，并给出最终的操作建议。 - -你必须严格按照以下JSON格式返回响应： - -{{ - "operation": [ - {{ - "id": "<记忆ID>", - "text": "<记忆内容>", - "event": "<操作类型，必须是 "ADD", "UPDATE", "NONE" 之一>", - "old_memory": "<原记忆内容，仅当操作为"UPDATE"时需要提供>" - }}, - ... - ] -}} - -要求： -1. 如果新事实对现有记忆没有额外补充，设置操作为"NONE" -2. 如果新事实与现有记忆相似但信息更准确、完整或需要修正，设置操作为"UPDATE" -3. 如果新事实与现有记忆在关键信息上矛盾（如时间、地点、状态等），以新事实为准更新原有记忆，设置操作为"UPDATE" -4. 如果有全新信息添加，设置操作为"ADD" - -ID管理规则： -- 更新操作：保持原有ID不变 -- 新增操作：生成新的唯一ID，格式为4位数字字符串（如："0001", "0002"等） - -重要要求： -- 只返回JSON格式的响应，不要包含其他任何内容 -- 对于更新操作，必须提供old_memory字段显示原内容 -- text字段要求：使用简洁、完整的陈述句，和新获取的事实信息一致，避免冗余信息 -- text和old_memory内容使用中文 - -示例： -现有记忆记录： -{{ - "memory": [ - {{ - "id": "123", - "text": "用户在公司A担任软件工程师，主要负责前端开发" - }}, - {{ - "id": "908", - "text": "用户周末喜欢和朋友一起钓鱼" - }} - ] -}} - -新获取的事实： -["用户现在在公司B担任高级全栈开发工程师", "用户的居住地址是北京市朝阳区明月小区", "用户在周末会去钓鱼"] - -操作建议： -{{ - "operation": [ - {{ - "id": "123", - "text": "用户在公司B担任高级全栈开发工程师", - "event": "UPDATE", - "old_memory": "用户在公司A担任软件工程师，主要负责前端开发" - }}, - {{ - "id": "4567", - "text": "用户的居住地址是北京市朝阳区明月小区", - "event": "ADD" - }}, - {{ - "id": "908", - "text": "用户周末喜欢和朋友一起钓鱼", - "event": "NONE" - }} - ] -}} - -现有记忆记录： -{current_memories} - -新获取的事实： +**新事实：** {new_facts} 操作建议： From 2529db24ac2a23889a2b795d6bef4c98a99eb8a8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Mon, 1 Dec 2025 20:45:27 +0800 Subject: [PATCH 28/72] add handler change node update --- examples/api/product_api.py | 30 ++- src/memos/api/handlers/add_handler.py | 31 ++- src/memos/api/handlers/component_init.py | 1 + src/memos/api/handlers/feedback_handler.py | 2 +- src/memos/api/product_models.py | 4 +- src/memos/graph_dbs/polardb.py | 16 +- src/memos/mem_feedback/feedback.py | 251 ++++++++++++++---- src/memos/mem_feedback/simple_feedback.py | 3 + src/memos/mem_scheduler/general_scheduler.py | 1 + .../tree_text_memory/organize/manager.py | 20 -- src/memos/multi_mem_cube/single_cube.py | 5 +- src/memos/multi_mem_cube/views.py | 15 +- src/memos/templates/mem_feedback_prompts.py | 179 ++++++++++++- 13 files changed, 458 insertions(+), 100 deletions(-) diff --git a/examples/api/product_api.py b/examples/api/product_api.py index b98f3b8e5..687c4c02c 100644 --- a/examples/api/product_api.py +++ b/examples/api/product_api.py @@ -12,7 +12,7 @@ import requests -BASE_URL = "http://0.0.0.0:8001/product" +BASE_URL = "http://0.0.0.0:8002/product" HEADERS = {"Content-Type": "application/json"} index = "24" @@ -119,6 +119,23 @@ def chat_stream(query: str, session_id: str, history: list | None = None): print(payload) +def feedback_memory(feedback_content: str, history: list | None = None): + url = f"{BASE_URL}/feedback" + data = { + "user_id": USER_ID, + "writable_cube_ids": [MEM_CUBE_ID], + "history": history, + "feedback_content": feedback_content, + "async_mode": "sync", + "corrected_answer": "false", + } + + print("[*] Feedbacking memory ...") + resp = requests.post(url, headers=HEADERS, data=json.dumps(data), timeout=30) + print(resp.status_code, resp.text) + return resp.json() + + if __name__ == "__main__": print("===== STEP 1: Register User =====") register_user() @@ -140,5 +157,14 @@ def chat_stream(query: str, session_id: str, history: list | None = None): ], ) - print("\n===== STEP 4: Stream Chat =====") + print("\n===== STEP 5: Stream Chat =====") chat_stream("我刚和你说什么了呢", SESSION_ID2, history=[]) + + print("\n===== STEP 6: Feedback Memory =====") + feedback_memory( + feedback_content="错啦，我今天没有吃拉面", + history=[ + {"role": "user", "content": "我刚和你说什么了呢"}, + {"role": "assistant", "content": "你今天吃了好吃的拉面"}, + ], + ) diff --git a/src/memos/api/handlers/add_handler.py b/src/memos/api/handlers/add_handler.py index 1bd83eae7..340a57e4a 100644 --- a/src/memos/api/handlers/add_handler.py +++ b/src/memos/api/handlers/add_handler.py @@ -6,7 +6,7 @@ """ from memos.api.handlers.base_handler import BaseHandler, HandlerDependencies -from memos.api.product_models import APIADDRequest, MemoryResponse +from memos.api.product_models import APIADDRequest, APIFeedbackRequest, MemoryResponse from memos.memories.textual.item import ( list_all_fields, ) @@ -56,6 +56,35 @@ def handle_add_memories(self, add_req: APIADDRequest) -> MemoryResponse: cube_view = self._build_cube_view(add_req) + if add_req.is_feedback: + chat_history = add_req.chat_history + messages = add_req.messages + concatenate_chat = chat_history + messages + + last_user_index = max(i for i, d in enumerate(concatenate_chat) if d["role"] == "user") + feedback_content = concatenate_chat[last_user_index]["content"] + feedback_history = concatenate_chat[:last_user_index] + + feedback_req = APIFeedbackRequest( + user_id=add_req.user_id, + session_id=add_req.session_id, + task_id=add_req.task_id, + history=feedback_history, + feedback_content=feedback_content, + writable_cube_ids=add_req.writable_cube_ids, + async_mode=add_req.async_mode, + ) + process_record = cube_view.feedback_memories(feedback_req) + + self.logger.info( + f"[FeedbackHandler] Final feedback results count={len(process_record)}" + ) + + return MemoryResponse( + message="Memory feedback successfully", + data=[process_record], + ) + results = cube_view.add_memories(add_req) self.logger.info(f"[AddHandler] Final add results count={len(results)}") diff --git a/src/memos/api/handlers/component_init.py b/src/memos/api/handlers/component_init.py index 982487b18..548f01887 100644 --- a/src/memos/api/handlers/component_init.py +++ b/src/memos/api/handlers/component_init.py @@ -300,6 +300,7 @@ def init_server() -> dict[str, Any]: graph_store=graph_db, memory_manager=memory_manager, mem_reader=mem_reader, + searcher=searcher, ) # Initialize Scheduler diff --git a/src/memos/api/handlers/feedback_handler.py b/src/memos/api/handlers/feedback_handler.py index 2d3dc4ee2..cf5c536ea 100644 --- a/src/memos/api/handlers/feedback_handler.py +++ b/src/memos/api/handlers/feedback_handler.py @@ -28,7 +28,7 @@ def __init__(self, dependencies: HandlerDependencies): dependencies: HandlerDependencies instance """ super().__init__(dependencies) - self._validate_dependencies("mem_reader", "mem_scheduler") + self._validate_dependencies("mem_reader", "mem_scheduler", "searcher") def handle_feedback_memories(self, feedback_req: APIFeedbackRequest) -> MemoryResponse: """ diff --git a/src/memos/api/product_models.py b/src/memos/api/product_models.py index 52d5fe6f6..f2b723565 100644 --- a/src/memos/api/product_models.py +++ b/src/memos/api/product_models.py @@ -639,7 +639,9 @@ class APIFeedbackRequest(BaseRequest): writable_cube_ids: list[str] | None = Field( None, description="List of cube IDs user can write for multi-cube add" ) - sync_mode: Literal["sync", "async"] = Field("async", description="feedback mode: sync or async") + async_mode: Literal["sync", "async"] = Field( + "async", description="feedback mode: sync or async" + ) corrected_answer: bool = Field(False, description="Whether need return corrected answer") # ==== Backward compatibility ==== mem_cube_id: str | None = Field( diff --git a/src/memos/graph_dbs/polardb.py b/src/memos/graph_dbs/polardb.py index ebaec2b90..da1635296 100644 --- a/src/memos/graph_dbs/polardb.py +++ b/src/memos/graph_dbs/polardb.py @@ -524,27 +524,19 @@ def update_node(self, id: str, fields: dict[str, Any], user_name: str | None = N properties = current_node["metadata"].copy() original_id = properties.get("id", id) # Preserve original ID original_memory = current_node.get("memory", "") # Preserve original memory - updated_at = fields.pop("updated_at", datetime.utcnow().isoformat()) - - record_info = f"User：{user_name} | Time：{updated_at} | Operation：Update | Overwrite： {original_memory}" - covered_history = current_node.get("covered_history", []) - covered_history.insert(0, record_info) - logger.info(f"New GraphDB Update: {record_info}") # If fields include memory, use it; otherwise keep original memory - new_memory = fields.pop("memory") if "memory" in fields else original_memory + if "memory" in fields: + original_memory = fields.pop("memory") properties.update(fields) - properties["id"] = original_id - properties["memory"] = new_memory - properties["covered_history"] = covered_history - properties["updated_at"] = updated_at + properties["id"] = original_id # Ensure ID is not overwritten + properties["memory"] = original_memory # Ensure memory is not overwritten # Handle embedding field embedding_vector = None if "embedding" in fields: embedding_vector = fields.pop("embedding") - assert properties["embedding"] == embedding_vector, "Embedding vector mismatch" if not isinstance(embedding_vector, list): embedding_vector = None diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index bf668852f..02b737451 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -3,6 +3,7 @@ import json from datetime import datetime +from typing import TYPE_CHECKING from tenacity import retry, stop_after_attempt, wait_exponential @@ -16,7 +17,14 @@ from memos.mem_reader.factory import MemReaderFactory from memos.mem_reader.simple_struct import detect_lang from memos.memories.textual.item import TextualMemoryItem, TreeNodeTextualMemoryMetadata -from memos.memories.textual.tree_text_memory.organize.manager import MemoryManager +from memos.memories.textual.tree_text_memory.organize.manager import ( + MemoryManager, + extract_working_binding_ids, +) + + +if TYPE_CHECKING: + from memos.memories.textual.tree_text_memory.retrieve.searcher import Searcher from memos.templates.mem_feedback_prompts import ( FEEDBACK_ANSWER_PROMPT, FEEDBACK_ANSWER_PROMPT_ZH, @@ -64,6 +72,7 @@ def __init__(self, config: MemFeedbackConfig): }, is_reorganize=self.is_reorganize, ) + self.searcher: Searcher = self.memory_manager.searcher def _pure_add(self, user_name: str, feedback_content: str, feedback_time: str, info: dict): """ @@ -113,10 +122,119 @@ def _feedback_judgement( ) return [] + def _single_add_operation( + self, + old_memory_item: TextualMemoryItem | None, + new_memory_item: TextualMemoryItem, + user_id: str, + user_name: str, + async_mode: str, + ) -> dict: + """ + Individual addition operations + """ + if old_memory_item: + to_add_memory = old_memory_item.model_copy(deep=True) + to_add_memory.metadata.key = new_memory_item.metadata.key + to_add_memory.metadata.tags = new_memory_item.metadata.tags + to_add_memory.memory = new_memory_item.memory + to_add_memory.metadata.embedding = new_memory_item.metadata.embedding + + to_add_memory.metadata.user_id = new_memory_item.metadata.user_id + to_add_memory.metadata.created_at = to_add_memory.metadata.updated_at = ( + datetime.now().isoformat() + ) + to_add_memory.metadata.background = new_memory_item.metadata.background + else: + to_add_memory = new_memory_item.model_copy(deep=True) + to_add_memory.metadata.created_at = to_add_memory.metadata.updated_at = ( + datetime.now().isoformat() + ) + to_add_memory.metadata.background = new_memory_item.metadata.background + + to_add_memory.id = "" + added_ids = self._retry_db_operation( + lambda: self.memory_manager.add([to_add_memory], user_name=user_name, mode=async_mode) + ) + + logger.info(f"[Memory Feedback ADD] {added_ids[0]}") + return {"id": added_ids[0], "text": to_add_memory.memory} + + def _single_update_operation( + self, + old_memory_item: TextualMemoryItem, + new_memory_item: TextualMemoryItem, + user_id: str, + user_name: str, + async_mode: str, + ) -> dict: + """ + Individual update operations + """ + memory_type = old_memory_item.metadata.memory_type + if memory_type == "WorkingMemory": + fields = { + "memory": new_memory_item.memory, + "key": new_memory_item.metadata.key, + "tags": new_memory_item.metadata.tags, + "embedding": new_memory_item.metadata.embedding, + "background": new_memory_item.metadata.background, + "covered_history": old_memory_item.id, + } + self.graph_store.update_node(old_memory_item.id, fields=fields, user_name=user_name) + item_id = old_memory_item.id + else: + done = self._single_add_operation( + old_memory_item, new_memory_item, user_id, user_name, async_mode + ) + item_id = done.get("id") + self.graph_store.update_node( + item_id, {"covered_history": old_memory_item.id}, user_name=user_name + ) + self.graph_store.update_node( + old_memory_item.id, {"status": "archived"}, user_name=user_name + ) + + logger.info( + f"[Memory Feedback UPDATE] New Add:{item_id} | Set archived:{old_memory_item.id} | memory_type: {memory_type}" + ) + + return { + "id": item_id, + "text": new_memory_item.memory, + "archived_id": old_memory_item.id, + "origin_memory": old_memory_item.memory, + } + + def _del_working_binding(self, user_name, mem_items: list[TextualMemoryItem]) -> set[str]: + """Delete working memory bindings""" + bindings_to_delete = extract_working_binding_ids(mem_items) + + logger.info( + f"[Memory Feedback UPDATE] Extracted {len(bindings_to_delete)} working_binding ids to cleanup: {list(bindings_to_delete)}" + ) + + delete_ids = [] + if bindings_to_delete: + delete_ids = list({bindings_to_delete}) + + for mid in delete_ids: + try: + print("del", mid) + self.graph_store.delete_node(mid, user_name=user_name) + + logger.info( + f"[Feedback Core:_del_working_binding] Delete raw/working mem_ids: {delete_ids} for user_name: {user_name}" + ) + except Exception as e: + logger.warning( + f"[Feedback Core:_del_working_binding] TreeTextMemory.delete_hard: failed to delete {mid}: {e}" + ) + def _feedback_memory( self, user_id: str, user_name: str, feedback_memories: list[TextualMemoryItem], **kwargs ) -> dict: - sync_mode = kwargs.get("sync_mode") + async_mode = kwargs.get("async_mode") retrieved_memory_ids = kwargs.get("retrieved_memory_ids") or [] chat_history = kwargs.get("chat_history", []) feedback_content = kwargs.get("feedback_content", "") @@ -124,7 +242,9 @@ def _feedback_memory( chat_history_lis = [f"""{msg["role"]}: {msg["content"]}""" for msg in chat_history[-4:]] fact_history = "\n".join(chat_history_lis) + f"\nuser feedback: \n{feedback_content}" - retrieved_memories = [self.graph_store.get_node(_id) for _id in retrieved_memory_ids] + retrieved_memories = [ + self.graph_store.get_node(_id, user_name=user_name) for _id in retrieved_memory_ids + ] filterd_ids = [ item["id"] for item in retrieved_memories if "mode:fast" in item["metadata"]["tags"] ] @@ -134,58 +254,29 @@ def _feedback_memory( ) current_memories = [ - {"id": item["id"], "text": item["memory"]} + TextualMemoryItem(**item) for item in retrieved_memories if "mode:fast" not in item["metadata"]["tags"] ] - def _single_add_operation( - memory_item: TextualMemoryItem, user_name: str, sync_mode: str - ) -> dict: - """ - Individual addition operations - """ - added_ids = self._retry_db_operation( - lambda: self.memory_manager.add([memory_item], user_name=user_name, mode=sync_mode) - ) - logger.info(f"[Memory Feedback ADD] {added_ids[0]}") - - return {"id": added_ids[0], "text": memory_item.memory} - - def _single_update_operation( - op: dict, memory_item: TextualMemoryItem, user_name: str, sync_mode: str - ) -> dict: - """ - Individual update operations - """ - update_id = op.get("id") - updated_ids = self._retry_db_operation( - lambda: self.memory_manager.update( - [update_id], [memory_item], user_name=user_name, mode=sync_mode - ) - ) - log_update_info = op.get("old_memory", "") + " >> " + op.get("text", "") - logger.info(f"[Memory Feedback UPDATE] {updated_ids[0]}, info: {log_update_info}") - - return { - "id": update_id, - "origin_memory": op.get("old_memory", ""), - "text": op.get("text", ""), - } - def _add_or_update( - memory_item: TextualMemoryItem, current_memories: list, fact_history: str + memory_item: TextualMemoryItem, + current_memories: list[TextualMemoryItem], + fact_history: str, ): if current_memories == []: - current_memories = self._vec_query( - memory_item.metadata.embedding, user_name=user_name + current_memories = self._retrieve( + memory_item.memory, info={"user_id": user_id}, user_name=user_name ) if current_memories: lang = detect_lang("".join(memory_item.memory)) template = FEEDBACK_PROMPT_DICT["compare"][lang] + current_memories_str = "\n".join( + [f"{item.id}: {item.memory}" for item in current_memories] + ) prompt = template.format( - current_memories=str(current_memories), + current_memories=current_memories_str, new_facts=memory_item.memory, chat_history=fact_history, ) @@ -195,7 +286,7 @@ def _add_or_update( else: operations = [{"operation": "ADD"}] - # TODO based on the operation, change memory_item memory info + # TODO based on the operation, change memory_item memory info ; change source info logger.info(f"[Feedback memory operations]: {operations!s}") if not operations: @@ -203,7 +294,7 @@ def _add_or_update( add_results = [] update_results = [] - + id_to_item = {item.id: item for item in current_memories} with ContextThreadPoolExecutor(max_workers=10) as executor: future_to_op = {} for op in operations: @@ -211,12 +302,22 @@ def _add_or_update( if event_type == "add": future = executor.submit( - _single_add_operation, memory_item, user_name, sync_mode + self._single_add_operation, + None, + memory_item, + user_id, + user_name, + async_mode, ) future_to_op[future] = ("add", op) elif event_type == "update": future = executor.submit( - _single_update_operation, op, memory_item, user_name, sync_mode + self._single_update_operation, + id_to_item[op["id"]], + memory_item, + user_id, + user_name, + async_mode, ) future_to_op[future] = ("update", op) @@ -224,15 +325,18 @@ def _add_or_update( result_type, original_op = future_to_op[future] try: result = future.result() - if result_type == "add": + if result_type == "add" and result: add_results.append(result) - elif result_type == "update": + elif result_type == "update" and result: update_results.append(result) except Exception as e: logger.error( f"[Feedback Core: _add_or_update] Operation failed for {original_op}: {e}", exc_info=True, ) + if update_results: + updated_ids = [item["archived_id"] for item in update_results] + self._del_working_binding(updated_ids, user_name) return {"record": {"add": add_results, "update": update_results}} @@ -262,11 +366,38 @@ def _add_or_update( } } + def _retrieve(self, query: str, info=None, user_name=None): + """Retrieve memory items""" + retrieved_mems = self.searcher.search(query, info=info, user_name=user_name) + return retrieved_mems + def _vec_query(self, new_memories_embedding: list[float], user_name=None): - retrieved_ids = self.graph_store.search_by_embedding( - new_memories_embedding, user_name=user_name, top_k=10, threshold=0.7 + """Vector retrieval query""" + retrieved_ids = [] + retrieved_ids.extend( + self.graph_store.search_by_embedding( + new_memories_embedding, + scope="UserMemory", + user_name=user_name, + top_k=10, + threshold=0.2, + ) + ) + retrieved_ids.extend( + self.graph_store.search_by_embedding( + new_memories_embedding, + scope="LongTermMemory", + user_name=user_name, + top_k=10, + threshold=0.2, + ) ) - current_memories = [self.graph_store.get_node(item["id"]) for item in retrieved_ids] + current_memories = [ + self.graph_store.get_node(item["id"], user_name=user_name) for item in retrieved_ids + ] + + for item in current_memories: + print(item["id"], item["metadata"]["memory_type"], item["metadata"]["status"]) if not retrieved_ids: logger.info( f"[Feedback Core: _vec_query] No similar memories found for embedding query for user {user_name}." @@ -280,10 +411,7 @@ def _vec_query(self, new_memories_embedding: list[float], user_name=None): f"[Feedback Core: _vec_query] Since the tags mode is fast, no modifications are made to the following memory {filterd_ids}." ) return [ - { - "id": item["id"], - "text": item["memory"], - } + TextualMemoryItem(**item) for item in current_memories if "mode:fast" not in item["metadata"]["tags"] ] @@ -303,7 +431,7 @@ def _get_llm_response(self, prompt: str, dsl: bool = True) -> dict: return response_json def _id_dehallucination(self, operations, current_memories): - right_ids = [item["id"] for item in current_memories] + right_ids = [item.id for item in current_memories] right_lower_map = {x.lower(): x for x in right_ids} def correct_item(data): @@ -437,7 +565,10 @@ def check_validity(item): usage=[], sources=[{"type": "chat"}], user_name=user_name, - background="", + background="[Feedback update background]: " + + str(chat_history) + + "\nUser feedback: " + + str(feedback_content), confidence=0.99, type="fine", ), @@ -476,7 +607,7 @@ def process_feedback( user_name: cube_ids chat_history: List of chat messages feedback_content: Feedback content from user - **kwargs: Additional arguments including sync_mode + **kwargs: Additional arguments including async_mode Returns: Dict with answer and/or memory operation records @@ -504,8 +635,10 @@ def process_feedback( try: answer = answer_future.result() record = core_future.result() + task_id = kwargs.get("task_id", "default") + logger.info( - f"[MemFeedback process] Completed concurrently for user {user_name} with full results." + f"[MemFeedback process] Feedback Completed : user {user_name} | task_id {task_id} | record {record}." ) return {"answer": answer, "record": record["record"]} diff --git a/src/memos/mem_feedback/simple_feedback.py b/src/memos/mem_feedback/simple_feedback.py index 59ee38438..01132eb97 100644 --- a/src/memos/mem_feedback/simple_feedback.py +++ b/src/memos/mem_feedback/simple_feedback.py @@ -5,6 +5,7 @@ from memos.mem_feedback.feedback import MemFeedback from memos.mem_reader.simple_struct import SimpleStructMemReader from memos.memories.textual.tree_text_memory.organize.manager import MemoryManager +from memos.memories.textual.tree_text_memory.retrieve.searcher import Searcher logger = log.get_logger(__name__) @@ -18,9 +19,11 @@ def __init__( graph_store: PolarDBGraphDB, memory_manager: MemoryManager, mem_reader: SimpleStructMemReader, + searcher: Searcher, ): self.llm = llm self.embedder = embedder self.graph_store = graph_store self.memory_manager = memory_manager self.mem_reader = mem_reader + self.searcher = searcher diff --git a/src/memos/mem_scheduler/general_scheduler.py b/src/memos/mem_scheduler/general_scheduler.py index 4190b8e5c..65169f358 100644 --- a/src/memos/mem_scheduler/general_scheduler.py +++ b/src/memos/mem_scheduler/general_scheduler.py @@ -380,6 +380,7 @@ def _mem_feedback_message_consumer(self, messages: list[ScheduleMessageItem]) -> retrieved_memory_ids=feedback_data["retrieved_memory_ids"], feedback_content=feedback_data["feedback_content"], feedback_time=feedback_data["feedback_time"], + task_id=feedback_data["task_id"], ) logger.info( diff --git a/src/memos/memories/textual/tree_text_memory/organize/manager.py b/src/memos/memories/textual/tree_text_memory/organize/manager.py index 94b86cec0..a71fee02f 100644 --- a/src/memos/memories/textual/tree_text_memory/organize/manager.py +++ b/src/memos/memories/textual/tree_text_memory/organize/manager.py @@ -115,26 +115,6 @@ def add( self._refresh_memory_size(user_name=user_name) return added_ids - def update( - self, - memories_ids: list[str], - memories: list[TextualMemoryItem], - user_name: str | None = None, - mode: str = "sync", - ) -> list[str]: - for _id, memory in zip(memories_ids, memories, strict=False): - field = { - "memory": memory.memory, - "key": memory.metadata.key, - "tags": memory.metadata.tags, - "embedding": memory.metadata.embedding, - } - self.graph_store.update_node(_id, field, user_name) - - graph_class_name = self.graph_store.__class__.__name__ - logger.info(f"[MemoryManager] Updated {graph_class_name}") - return memories_ids - def replace_working_memory( self, memories: list[TextualMemoryItem], user_name: str | None = None ) -> None: diff --git a/src/memos/multi_mem_cube/single_cube.py b/src/memos/multi_mem_cube/single_cube.py index 21d4fd8d1..4d40798ec 100644 --- a/src/memos/multi_mem_cube/single_cube.py +++ b/src/memos/multi_mem_cube/single_cube.py @@ -135,7 +135,7 @@ def search_memories(self, search_req: APISearchRequest) -> dict[str, Any]: def feedback_memories(self, feedback_req: APIFeedbackRequest) -> dict[str, Any]: target_session_id = feedback_req.session_id or "default_session" - if feedback_req.sync_mode == "async": + if feedback_req.async_mode == "async": try: feedback_req_str = json.dumps(feedback_req.model_dump()) message_item_feedback = ScheduleMessageItem( @@ -167,8 +167,9 @@ def feedback_memories(self, feedback_req: APIFeedbackRequest) -> dict[str, Any]: retrieved_memory_ids=feedback_req.retrieved_memory_ids, feedback_content=feedback_req.feedback_content, feedback_time=feedback_req.feedback_time, - sync_mode=feedback_req.sync_mode, + async_mode=feedback_req.async_mode, corrected_answer=feedback_req.corrected_answer, + task_id=feedback_req.task_id, ) self.logger.info(f"Feedback memories result: {feedback_result}") return feedback_result diff --git a/src/memos/multi_mem_cube/views.py b/src/memos/multi_mem_cube/views.py index baf5e80e1..7247a0328 100644 --- a/src/memos/multi_mem_cube/views.py +++ b/src/memos/multi_mem_cube/views.py @@ -4,7 +4,7 @@ if TYPE_CHECKING: - from memos.api.product_models import APIADDRequest, APISearchRequest + from memos.api.product_models import APIADDRequest, APIFeedbackRequest, APISearchRequest class MemCubeView(Protocol): @@ -39,3 +39,16 @@ def search_memories(self, search_req: APISearchRequest) -> dict[str, Any]: - cube_id """ ... + + def feedback_memories(self, feedback_req: APIFeedbackRequest) -> dict[str, Any]: + """ + Process feedback_req, read memories from one or more cubes and feedback them. + + Returns: + A list of memory dicts, each item should at least contain: + - memory + - memory_id + - memory_type + - cube_id + """ + ... diff --git a/src/memos/templates/mem_feedback_prompts.py b/src/memos/templates/mem_feedback_prompts.py index 2e9a0e9c9..f7f2e8cb4 100644 --- a/src/memos/templates/mem_feedback_prompts.py +++ b/src/memos/templates/mem_feedback_prompts.py @@ -238,6 +238,94 @@ 5. Return only the JSON format response, without any other content + +Example1: +Current Memories: +"0911": "The user is a senior full-stack developer working at Company B" +"123": "The user works as a software engineer at Company A. And he has a good relationship with his wife." +"648": "The user is responsible for front-end development of software at Company A" +"7210": "The user is responsible for front-end development of software at Company A" +"908": "The user enjoys fishing with friends on weekends" + +The background of the new fact being put forward: +user: Do you remember where I work？ +assistant: Company A. +user feedback: I work at Company B, and I am a senior full-stack developer. + +Newly facts: +The user works as a senior full-stack developer at Company B + +Operation recommendations: +{{ + "operations": + [ + {{ + "id": "0911", + "text": "The user is a senior full-stack developer working at Company B", + "operation": "NONE" + }}, + {{ + "id": "123", + "text": "The user works as a senior full-stack developer at Company B. And he has a good relationship with his wife.", + "operation": "UPDATE", + "old_memory": "The user works as a software engineer at Company A. And he has a good relationship with his wife." + }}, + {{ + "id": "648", + "text": "The user works as a senior full-stack developer at Company B", + "operation": "UPDATE", + "old_memory": "The user is responsible for front-end development of software at Company A" + }}, + {{ + "id": "7210", + "text": "The user works as a senior full-stack developer at Company B", + "operation": "UPDATE", + "old_memory": "The user is responsible for front-end development of software at Company A" + }}, + {{ + "id": "908", + "text": "The user enjoys fishing with friends on weekends", + "operation": "NONE" + }} + ] +}} + +Example2: +Current Memories: +"123": "The user works as a software engineer in Company A, mainly responsible for front-end development" +"908": "The user likes to go fishing with friends on weekends" + +The background of the new fact being put forward: +user: Guess where I live？ +assistant: Hehuan Community. +user feedback: Wrong, update my address: Mingyue Community, Chaoyang District, Beijing + +Newly facts: +"The user's residential address is Mingyue Community, Chaoyang District, Beijing" + +Operation recommendations: +{{ + "operations": + [ + {{ + "id": "123", + "text": "The user works as a software engineer at Company A, primarily responsible for front-end development", + "operation": "NONE" + }}, + {{ + "id": "908", + "text": "The user enjoys fishing with friends on weekends", + "operation": "NONE" + }}, + {{ + "id": "4567", + "text": "The user's residential address is Mingyue Community, Chaoyang District, Beijing", + "operation": "ADD" + }} + ] +}} + + **Current Memories** {current_memories} @@ -255,7 +343,7 @@ 你必须严格按照以下JSON格式返回响应： {{ - "operation": + "operations": [ {{ "id": "<记忆ID>", @@ -288,6 +376,95 @@ 5. 只返回JSON格式的响应，不要包含其他任何内容 +示例1： +当前记忆： +"0911": "用户是高级全栈开发工程师，在B公司工作" +"123": "用户在公司A担任软件工程师。而且用户和同事们的关系很好，他们共同协作大项目。" +"648": "用户在公司A负责软件的前端开发工作" +"7210": "用户在公司A负责软件的前端开发工作" +"908": "用户周末喜欢和朋友一起钓鱼" + + +提出新事实的背景： +user: 你还记得我现在在哪里工作吗？ +assistant: A公司 +user feedback: 实际上，我在公司B工作，是一名高级全栈开发人员。 + + +新获取的事实： +"用户现在在公司B担任高级全栈开发工程师" + +操作建议： +{{ + "operations": + [ + {{ + "id": "0911", + "text": "用户是高级全栈开发工程师，在B公司工作", + "operation": "NONE" + }}, + {{ + "id": "123", + "text": "用户现在在公司B担任高级全栈开发工程师。而且用户和同事们的关系很好，他们共同协作大项目。", + "operation": "UPDATE", + "old_memory": "用户在公司A担任软件工程师，主要负责前端开发。而且用户和同事们的关系很好，他们共同协作大项目。" + }}, + {{ + "id": "648", + "text": "用户现在在公司B担任高级全栈开发工程师", + "operation": "UPDATE", + "old_memory": "用户在公司A负责软件的前端开发工作" + }}, + {{ + "id": "7210", + "text": "用户现在在公司B担任高级全栈开发工程师", + "operation": "UPDATE", + "old_memory": "用户在公司A负责软件的前端开发工作" + }}, + {{ + "id": "908", + "text": "用户周末喜欢和朋友一起钓鱼", + "operation": "NONE" + }} + ] +}} + +示例2： +当前记忆： +"123": "用户在公司A担任软件工程师，主要负责前端开发" +"908": "用户周末喜欢和朋友一起钓鱼" + + +提出新事实的背景： +user: 猜猜我住在哪里？ +assistant: 合欢社区 +user feedback: 错了，请更新我的地址：北京市朝阳区明月社区 + +新获取的事实： +"用户的居住地址是北京市朝阳区明月小区" + +操作建议： +{{ + "operations": + [ + {{ + "id": "123", + "text": "用户在公司A担任软件工程师，主要负责前端开发", + "operation": "NONE" + }}, + {{ + "id": "908", + "text": "用户周末喜欢和朋友一起钓鱼", + "operation": "NONE" + }}, + {{ + "id": "4567", + "text": "用户的居住地址是北京市朝阳区明月小区", + "operation": "ADD" + }} + ] +}} + **当前记忆：** {current_memories} From 898ccacb5457c24c0ae0e27d408711b28b878e0e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Mon, 1 Dec 2025 20:53:12 +0800 Subject: [PATCH 29/72] add handler change node update --- src/memos/api/handlers/add_handler.py | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) diff --git a/src/memos/api/handlers/add_handler.py b/src/memos/api/handlers/add_handler.py index 340a57e4a..46e7fd108 100644 --- a/src/memos/api/handlers/add_handler.py +++ b/src/memos/api/handlers/add_handler.py @@ -30,7 +30,9 @@ def __init__(self, dependencies: HandlerDependencies): dependencies: HandlerDependencies instance """ super().__init__(dependencies) - self._validate_dependencies("naive_mem_cube", "mem_reader", "mem_scheduler") + self._validate_dependencies( + "naive_mem_cube", "mem_reader", "mem_scheduler", "feedback_server" + ) def handle_add_memories(self, add_req: APIADDRequest) -> MemoryResponse: """ @@ -59,6 +61,10 @@ def handle_add_memories(self, add_req: APIADDRequest) -> MemoryResponse: if add_req.is_feedback: chat_history = add_req.chat_history messages = add_req.messages + if chat_history is None: + chat_history = [] + if messages is None: + messages = [] concatenate_chat = chat_history + messages last_user_index = max(i for i, d in enumerate(concatenate_chat) if d["role"] == "user") @@ -117,6 +123,7 @@ def _build_cube_view(self, add_req: APIADDRequest) -> MemCubeView: mem_reader=self.mem_reader, mem_scheduler=self.mem_scheduler, logger=self.logger, + feedback_server=self.feedback_server, searcher=None, ) else: @@ -127,6 +134,7 @@ def _build_cube_view(self, add_req: APIADDRequest) -> MemCubeView: mem_reader=self.mem_reader, mem_scheduler=self.mem_scheduler, logger=self.logger, + feedback_server=self.feedback_server, searcher=None, ) for cube_id in cube_ids From 91d063dd6186d65902157fbc0e1bcce060b04835 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Mon, 1 Dec 2025 20:59:28 +0800 Subject: [PATCH 30/72] add handler change node update --- tests/api/test_server_router.py | 1 + 1 file changed, 1 insertion(+) diff --git a/tests/api/test_server_router.py b/tests/api/test_server_router.py index 7c4b4be9d..5906697d9 100644 --- a/tests/api/test_server_router.py +++ b/tests/api/test_server_router.py @@ -38,6 +38,7 @@ def mock_init_server(): "default_cube_config": Mock(), "mos_server": Mock(), "mem_scheduler": Mock(), + "feedback_server": Mock(), "naive_mem_cube": Mock(), "searcher": Mock(), "api_module": Mock(), From 2a478802053ede23d6b882dd634d9945fd6b1066 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Mon, 1 Dec 2025 21:01:27 +0800 Subject: [PATCH 31/72] add handler change node update --- examples/api/product_api.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/examples/api/product_api.py b/examples/api/product_api.py index 687c4c02c..e364ce483 100644 --- a/examples/api/product_api.py +++ b/examples/api/product_api.py @@ -12,7 +12,7 @@ import requests -BASE_URL = "http://0.0.0.0:8002/product" +BASE_URL = "http://0.0.0.0:8001/product" HEADERS = {"Content-Type": "application/json"} index = "24" From ad9c2e73515711c1e8679fa0dffa745fb76ed44f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 2 Dec 2025 14:31:06 +0800 Subject: [PATCH 32/72] fix interface input --- src/memos/api/handlers/add_handler.py | 68 +++++++++++++++------------ src/memos/api/product_models.py | 3 +- src/memos/memories/textual/item.py | 5 ++ 3 files changed, 45 insertions(+), 31 deletions(-) diff --git a/src/memos/api/handlers/add_handler.py b/src/memos/api/handlers/add_handler.py index 46e7fd108..31f5bc35b 100644 --- a/src/memos/api/handlers/add_handler.py +++ b/src/memos/api/handlers/add_handler.py @@ -5,6 +5,8 @@ using dependency injection for better modularity and testability. """ +from pydantic import validate_call + from memos.api.handlers.base_handler import BaseHandler, HandlerDependencies from memos.api.product_models import APIADDRequest, APIFeedbackRequest, MemoryResponse from memos.memories.textual.item import ( @@ -13,6 +15,7 @@ from memos.multi_mem_cube.composite_cube import CompositeCubeView from memos.multi_mem_cube.single_cube import SingleCubeView from memos.multi_mem_cube.views import MemCubeView +from memos.types import MessageList class AddHandler(BaseHandler): @@ -58,38 +61,45 @@ def handle_add_memories(self, add_req: APIADDRequest) -> MemoryResponse: cube_view = self._build_cube_view(add_req) + @validate_call + def _check_messages(messages: MessageList) -> None: + pass + if add_req.is_feedback: - chat_history = add_req.chat_history - messages = add_req.messages - if chat_history is None: - chat_history = [] - if messages is None: - messages = [] - concatenate_chat = chat_history + messages - - last_user_index = max(i for i, d in enumerate(concatenate_chat) if d["role"] == "user") - feedback_content = concatenate_chat[last_user_index]["content"] - feedback_history = concatenate_chat[:last_user_index] - - feedback_req = APIFeedbackRequest( - user_id=add_req.user_id, - session_id=add_req.session_id, - task_id=add_req.task_id, - history=feedback_history, - feedback_content=feedback_content, - writable_cube_ids=add_req.writable_cube_ids, - async_mode=add_req.async_mode, - ) - process_record = cube_view.feedback_memories(feedback_req) + try: + messages = add_req.messages + _check_messages(messages) - self.logger.info( - f"[FeedbackHandler] Final feedback results count={len(process_record)}" - ) + chat_history = add_req.chat_history if add_req.chat_history else [] + concatenate_chat = chat_history + messages - return MemoryResponse( - message="Memory feedback successfully", - data=[process_record], - ) + last_user_index = max( + i for i, d in enumerate(concatenate_chat) if d["role"] == "user" + ) + feedback_content = concatenate_chat[last_user_index]["content"] + feedback_history = concatenate_chat[:last_user_index] + + feedback_req = APIFeedbackRequest( + user_id=add_req.user_id, + session_id=add_req.session_id, + task_id=add_req.task_id, + history=feedback_history, + feedback_content=feedback_content, + writable_cube_ids=add_req.writable_cube_ids, + async_mode=add_req.async_mode, + ) + process_record = cube_view.feedback_memories(feedback_req) + + self.logger.info( + f"[ADDFeedbackHandler] Final feedback results count={len(process_record)}" + ) + + return MemoryResponse( + message="Memory feedback successfully", + data=[process_record], + ) + except Exception as e: + self.logger.warning(f"[ADDFeedbackHandler] Running error: {e}") results = cube_view.add_memories(add_req) diff --git a/src/memos/api/product_models.py b/src/memos/api/product_models.py index d58385c2e..216664e53 100644 --- a/src/memos/api/product_models.py +++ b/src/memos/api/product_models.py @@ -642,7 +642,6 @@ class APIFeedbackRequest(BaseRequest): ) feedback_content: str | None = Field(..., description="Feedback content to process") feedback_time: str | None = Field(None, description="Feedback time") - # ==== Multi-cube writing ==== writable_cube_ids: list[str] | None = Field( None, description="List of cube IDs user can write for multi-cube add" ) @@ -650,7 +649,7 @@ class APIFeedbackRequest(BaseRequest): "async", description="feedback mode: sync or async" ) corrected_answer: bool = Field(False, description="Whether need return corrected answer") - # ==== Backward compatibility ==== + # ==== mem_cube_id is NOT enabled==== mem_cube_id: str | None = Field( None, description=( diff --git a/src/memos/memories/textual/item.py b/src/memos/memories/textual/item.py index b7956bfec..63a5b3b1d 100644 --- a/src/memos/memories/textual/item.py +++ b/src/memos/memories/textual/item.py @@ -90,6 +90,11 @@ class TextualMemoryMetadata(BaseModel): model_config = ConfigDict(extra="allow") + covered_history: str | None = Field( + default=None, + description="Record the memory id covered by the update", + ) + def __str__(self) -> str: """Pretty string representation of the metadata.""" meta = self.model_dump(exclude_none=True) From 6ad8dae2253f544401541b85d95593a42fb10cf0 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 3 Dec 2025 17:33:55 +0800 Subject: [PATCH 33/72] add chunk and ratio filter --- src/memos/graph_dbs/polardb.py | 92 ++++ src/memos/mem_feedback/feedback.py | 416 +++++++++++++----- src/memos/mem_feedback/simple_feedback.py | 2 + src/memos/mem_feedback/utils.py | 86 ++++ src/memos/memories/textual/item.py | 2 +- .../tree_text_memory/retrieve/searcher.py | 4 + src/memos/templates/mem_feedback_prompts.py | 116 +++++ 7 files changed, 599 insertions(+), 119 deletions(-) create mode 100644 src/memos/mem_feedback/utils.py diff --git a/src/memos/graph_dbs/polardb.py b/src/memos/graph_dbs/polardb.py index 74dd38fc1..580811a50 100644 --- a/src/memos/graph_dbs/polardb.py +++ b/src/memos/graph_dbs/polardb.py @@ -1451,6 +1451,98 @@ def get_context_chain(self, id: str, type: str = "FOLLOWS") -> list[str]: """Get the ordered context chain starting from a node.""" raise NotImplementedError + @timed + def seach_by_keywords( + self, + query_words: list[str], + scope: str | None = None, + status: str | None = None, + search_filter: dict | None = None, + user_name: str | None = None, + filter: dict | None = None, + knowledgebase_ids: list[str] | None = None, + tsvector_field: str = "properties_tsvector_zh", + tsquery_config: str = "jiebaqry", + **kwargs, + ) -> list[dict]: + where_clauses = [] + + if scope: + where_clauses.append( + f"ag_catalog.agtype_access_operator(properties, '\"memory_type\"'::agtype) = '\"{scope}\"'::agtype" + ) + if status: + where_clauses.append( + f"ag_catalog.agtype_access_operator(properties, '\"status\"'::agtype) = '\"{status}\"'::agtype" + ) + else: + where_clauses.append( + "ag_catalog.agtype_access_operator(properties, '\"status\"'::agtype) = '\"activated\"'::agtype" + ) + + # Build user_name filter with knowledgebase_ids support (OR relationship) using common method + user_name_conditions = self._build_user_name_and_kb_ids_conditions_sql( + user_name=user_name, + knowledgebase_ids=knowledgebase_ids, + default_user_name=self.config.user_name, + ) + + # Add OR condition if we have any user_name conditions + if user_name_conditions: + if len(user_name_conditions) == 1: + where_clauses.append(user_name_conditions[0]) + else: + where_clauses.append(f"({' OR '.join(user_name_conditions)})") + + # Add search_filter conditions + if search_filter: + for key, value in search_filter.items(): + if isinstance(value, str): + where_clauses.append( + f"ag_catalog.agtype_access_operator(properties, '\"{key}\"'::agtype) = '\"{value}\"'::agtype" + ) + else: + where_clauses.append( + f"ag_catalog.agtype_access_operator(properties, '\"{key}\"'::agtype) = {value}::agtype" + ) + + # Build filter conditions using common method + filter_conditions = self._build_filter_conditions_sql(filter) + where_clauses.extend(filter_conditions) + # Add fulltext search condition + # Convert query_text to OR query format: "word1 | word2 | word3" + tsquery_string = " | ".join(query_words) + + where_clauses.append(f"{tsvector_field} @@ to_tsquery('{tsquery_config}', %s)") + + where_clause = f"WHERE {' AND '.join(where_clauses)}" if where_clauses else "" + + # Build fulltext search query + query = f""" + SELECT + ag_catalog.agtype_access_operator(properties, '"id"'::agtype) AS old_id, + agtype_object_field_text(properties, 'memory') as memory_text + FROM "{self.db_name}_graph"."Memory" + {where_clause} + """ + + params = (tsquery_string,) + logger.info(f"[search_by_fulltext] query: {query}, params: {params}") + conn = self._get_connection() + try: + with conn.cursor() as cursor: + cursor.execute(query, params) + results = cursor.fetchall() + output = [] + for row in results: + oldid = row[0] + id_val = str(oldid) + output.append({"id": id_val}) + + return output + finally: + self._return_connection(conn) + @timed def search_by_fulltext( self, diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index 02b737451..eed43d66e 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -10,10 +10,12 @@ from memos import log from memos.configs.memory import MemFeedbackConfig from memos.context.context import ContextThreadPoolExecutor +from memos.dependency import require_python_package from memos.embedders.factory import EmbedderFactory, OllamaEmbedder from memos.graph_dbs.factory import GraphStoreFactory, PolarDBGraphDB from memos.llms.factory import AzureLLM, LLMFactory, OllamaLLM, OpenAILLM from memos.mem_feedback.base import BaseMemFeedback +from memos.mem_feedback.utils import should_keep_update, split_into_chunks from memos.mem_reader.factory import MemReaderFactory from memos.mem_reader.simple_struct import detect_lang from memos.memories.textual.item import TextualMemoryItem, TreeNodeTextualMemoryMetadata @@ -30,6 +32,8 @@ FEEDBACK_ANSWER_PROMPT_ZH, FEEDBACK_JUDGEMENT_PROMPT, FEEDBACK_JUDGEMENT_PROMPT_ZH, + KEYWORDS_REPLACE, + KEYWORDS_REPLACE_ZH, UPDATE_FORMER_MEMORIES, UPDATE_FORMER_MEMORIES_ZH, ) @@ -37,6 +41,7 @@ FEEDBACK_PROMPT_DICT = { + "if_kw_replace": {"en": KEYWORDS_REPLACE, "zh": KEYWORDS_REPLACE_ZH}, "judge": {"en": FEEDBACK_JUDGEMENT_PROMPT, "zh": FEEDBACK_JUDGEMENT_PROMPT_ZH}, "compare": {"en": UPDATE_FORMER_MEMORIES, "zh": UPDATE_FORMER_MEMORIES_ZH}, "generation": {"en": FEEDBACK_ANSWER_PROMPT, "zh": FEEDBACK_ANSWER_PROMPT_ZH}, @@ -74,6 +79,20 @@ def __init__(self, config: MemFeedbackConfig): ) self.searcher: Searcher = self.memory_manager.searcher + def _batch_embed(self, texts: list[str], embed_bs: int = 5): + embed_bs = 5 + texts_embeddings = [] + for i in range(0, len(texts), embed_bs): + batch = texts[i : i + embed_bs] + try: + texts_embeddings.extend(self.embedder.embed(batch)) + except Exception as e: + logger.error( + f"[Feedback Core: process_feedback_core] Embedding batch failed: {e}", + exc_info=True, + ) + return texts_embeddings + def _pure_add(self, user_name: str, feedback_content: str, feedback_time: str, info: dict): """ Directly add new memory @@ -97,6 +116,25 @@ def _pure_add(self, user_name: str, feedback_content: str, feedback_time: str, i } } + def _keyword_replace_judgement(self, feedback_content: str) -> dict | None: + """ + Determine whether it is keyword replacement + """ + lang = detect_lang(feedback_content) + template = FEEDBACK_PROMPT_DICT["if_kw_replace"][lang] + prompt = template.format( + user_feedback=feedback_content, + ) + + judge_res = self._get_llm_response(prompt) + if judge_res: + return judge_res + else: + logger.warning( + "[Feedback Core: _feedback_judgement] feedback judgement failed, return []" + ) + return {} + def _feedback_judgement( self, chat_history: list[MessageDict], feedback_content: str, feedback_time: str = "" ) -> dict | None: @@ -128,7 +166,7 @@ def _single_add_operation( new_memory_item: TextualMemoryItem, user_id: str, user_name: str, - async_mode: str, + async_mode: str = "sync", ) -> dict: """ Individual addition operations @@ -166,7 +204,7 @@ def _single_update_operation( new_memory_item: TextualMemoryItem, user_id: str, user_name: str, - async_mode: str, + async_mode: str = "sync", ) -> dict: """ Individual update operations @@ -231,10 +269,111 @@ def _del_working_binding(self, user_name, mem_items: list[TextualMemoryItem]) -> f"[Feedback Core:_del_working_binding] TreeTextMemory.delete_hard: failed to delete {mid}: {e}" ) + def semantics_feedback( + self, + user_id: str, + user_name: str, + memory_item: TextualMemoryItem, + current_memories: list[TextualMemoryItem], + fact_history: str, + ): + lang = detect_lang("".join(memory_item.memory)) + template = FEEDBACK_PROMPT_DICT["compare"][lang] + if current_memories == []: + current_memories = self._retrieve( + memory_item.memory, info={"user_id": user_id}, user_name=user_name + ) + + if not current_memories: + operations = [{"operation": "ADD"}] + else: + memory_chunks = split_into_chunks(current_memories, max_tokens_per_chunk=500) + + all_operations = [] + with ContextThreadPoolExecutor(max_workers=10) as executor: + future_to_chunk_idx = {} + for chunk in memory_chunks: + current_memories_str = "\n".join( + [f"{item.id}: {item.memory}" for item in chunk] + ) + prompt = template.format( + current_memories=current_memories_str, + new_facts=memory_item.memory, + chat_history=fact_history, + ) + + future = executor.submit(self._get_llm_response, prompt) + future_to_chunk_idx[future] = chunk + for future in concurrent.futures.as_completed(future_to_chunk_idx): + try: + chunk_operations = future.result() + if ( + chunk_operations + and "operations" in chunk_operations + and isinstance(chunk_operations["operations"], list) + ): + all_operations.extend(chunk_operations["operations"]) + except Exception as e: + logger.error(f"[Feedback Core: semantics_feedback] Operation failed: {e}") + + operations = self.standard_operations(all_operations, current_memories) + + # TODO based on the operation, change memory_item memory info ; change source info + logger.info(f"[Feedback memory operations]: {operations!s}") + + if not operations: + return {"record": {"add": [], "update": []}} + + add_results = [] + update_results = [] + id_to_item = {item.id: item for item in current_memories} + + with ContextThreadPoolExecutor(max_workers=10) as executor: + future_to_op = {} + for op in operations: + event_type = op.get("operation", "").lower() + + if event_type == "add": + future = executor.submit( + self._single_add_operation, + None, + memory_item, + user_id, + user_name, + ) + future_to_op[future] = ("add", op) + elif event_type == "update": + future = executor.submit( + self._single_update_operation, + id_to_item[op["id"]], + memory_item, + user_id, + user_name, + ) + future_to_op[future] = ("update", op) + + for future in concurrent.futures.as_completed(future_to_op): + result_type, original_op = future_to_op[future] + try: + result = future.result() + if result_type == "add" and result: + add_results.append(result) + elif result_type == "update" and result: + update_results.append(result) + except Exception as e: + logger.error( + f"[Feedback Core: semantics_feedback] Operation failed for {original_op}: {e}", + exc_info=True, + ) + if update_results: + updated_ids = [item["archived_id"] for item in update_results] + self._del_working_binding(updated_ids, user_name) + + return {"record": {"add": add_results, "update": update_results}} + def _feedback_memory( self, user_id: str, user_name: str, feedback_memories: list[TextualMemoryItem], **kwargs ) -> dict: - async_mode = kwargs.get("async_mode") retrieved_memory_ids = kwargs.get("retrieved_memory_ids") or [] chat_history = kwargs.get("chat_history", []) feedback_content = kwargs.get("feedback_content", "") @@ -259,90 +398,11 @@ def _feedback_memory( if "mode:fast" not in item["metadata"]["tags"] ] - def _add_or_update( - memory_item: TextualMemoryItem, - current_memories: list[TextualMemoryItem], - fact_history: str, - ): - if current_memories == []: - current_memories = self._retrieve( - memory_item.memory, info={"user_id": user_id}, user_name=user_name - ) - - if current_memories: - lang = detect_lang("".join(memory_item.memory)) - template = FEEDBACK_PROMPT_DICT["compare"][lang] - current_memories_str = "\n".join( - [f"{item.id}: {item.memory}" for item in current_memories] - ) - prompt = template.format( - current_memories=current_memories_str, - new_facts=memory_item.memory, - chat_history=fact_history, - ) - - operations = self._get_llm_response(prompt).get("operations", []) - operations = self._id_dehallucination(operations, current_memories) - else: - operations = [{"operation": "ADD"}] - - # TODO based on the operation, change memory_item memory info ; change source info - logger.info(f"[Feedback memory operations]: {operations!s}") - - if not operations: - return {"record": {"add": [], "update": []}} - - add_results = [] - update_results = [] - id_to_item = {item.id: item for item in current_memories} - with ContextThreadPoolExecutor(max_workers=10) as executor: - future_to_op = {} - for op in operations: - event_type = op.get("operation", "").lower() - - if event_type == "add": - future = executor.submit( - self._single_add_operation, - None, - memory_item, - user_id, - user_name, - async_mode, - ) - future_to_op[future] = ("add", op) - elif event_type == "update": - future = executor.submit( - self._single_update_operation, - id_to_item[op["id"]], - memory_item, - user_id, - user_name, - async_mode, - ) - future_to_op[future] = ("update", op) - - for future in concurrent.futures.as_completed(future_to_op): - result_type, original_op = future_to_op[future] - try: - result = future.result() - if result_type == "add" and result: - add_results.append(result) - elif result_type == "update" and result: - update_results.append(result) - except Exception as e: - logger.error( - f"[Feedback Core: _add_or_update] Operation failed for {original_op}: {e}", - exc_info=True, - ) - if update_results: - updated_ids = [item["archived_id"] for item in update_results] - self._del_working_binding(updated_ids, user_name) - - return {"record": {"add": add_results, "update": update_results}} - with ContextThreadPoolExecutor(max_workers=3) as ex: futures = { - ex.submit(_add_or_update, mem, current_memories, fact_history): i + ex.submit( + self.semantics_feedback, user_id, user_name, mem, current_memories, fact_history + ): i for i, mem in enumerate(feedback_memories) } results = [None] * len(futures) @@ -368,7 +428,10 @@ def _add_or_update( def _retrieve(self, query: str, info=None, user_name=None): """Retrieve memory items""" - retrieved_mems = self.searcher.search(query, info=info, user_name=user_name) + retrieved_mems = self.searcher.search( + query, info=info, user_name=user_name, topk=50, full_recall=True + ) + retrieved_mems = [item[0] for item in retrieved_mems] return retrieved_mems def _vec_query(self, new_memories_embedding: list[float], user_name=None): @@ -430,28 +493,51 @@ def _get_llm_response(self, prompt: str, dsl: bool = True) -> dict: response_json = None return response_json - def _id_dehallucination(self, operations, current_memories): + def standard_operations(self, operations, current_memories): right_ids = [item.id for item in current_memories] right_lower_map = {x.lower(): x for x in right_ids} def correct_item(data): - if data.get("operation", "").lower() != "update": - return data - - original_id = data["id"] - if original_id in right_ids: - return data - - lower_id = original_id.lower() - if lower_id in right_lower_map: - data["id"] = right_lower_map[lower_id] - return data - - matches = difflib.get_close_matches(original_id, right_ids, n=1, cutoff=0.8) - if matches: - data["id"] = matches[0] - return data + try: + assert "operation" in data + if data.get("operation", "").lower() == "add": + return data + + if data.get("operation", "").lower() == "none": + return None + + assert ( + "id" in data + and "text" in data + and "old_memory" in data + and data["operation"].lower() == "update" + ) + if not should_keep_update(data["text"], data["old_memory"]): + logger.warning( + f"[Feedback Core: semantics_feedback] Due to the excessive proportion of changes, skip update: {data}" + ) + return None + + # id dehallucination + original_id = data["id"] + if original_id in right_ids: + return data + + lower_id = original_id.lower() + if lower_id in right_lower_map: + data["id"] = right_lower_map[lower_id] + return data + + matches = difflib.get_close_matches(original_id, right_ids, n=1, cutoff=0.8) + if matches: + data["id"] = matches[0] + return data + except Exception: + logger.error( + f"[Feedback Core: standard_operations] Error processing operation item: {data}", + exc_info=True, + ) return None dehallu_res = [correct_item(item) for item in operations] @@ -475,6 +561,86 @@ def _generate_answer( return self._get_llm_response(prompt, dsl=False) + def process_keyword_replace(self, user_id: str, user_name: str, kwp_judge: dict | None = None): + """ + memory keyword replace process + """ + doc_scope = kwp_judge.get("doc_scope", "NONE") + original_word = kwp_judge.get("original") + target_word = kwp_judge.get("target") + + # retrieve + lang = detect_lang(original_word) + queries = self._tokenize_chinese(original_word) if lang == "zh" else original_word.split() + + must_part = f"{' & '.join(queries)}" if len(queries) > 1 else queries[0] + retrieved_ids = self.graph_store.seach_by_keywords([must_part], user_name=user_name) + if len(retrieved_ids) < 1: + retrieved_ids = self.graph_store.search_by_fulltext( + queries, top_k=100, user_name=user_name + ) + + # filter by doc scope + mem_data = [ + self.graph_store.get_node(item["id"], user_name=user_name) for item in retrieved_ids + ] + retrieved_memories = [TextualMemoryItem(**item) for item in mem_data] + + if doc_scope != "NONE": + retrieved_memories = [ + item + for item in retrieved_memories + if doc_scope in item.metadata.sources # TODO + ] + + if not retrieved_memories: + return {"record": {"add": [], "update": []}} + + # replace keywords + pick_index = [] + update_memories = [] + for i, old_mem in enumerate(retrieved_memories): + if original_word in old_mem.memory: + mem = old_mem.model_copy(deep=True) + mem.memory = mem.memory.replace(original_word, target_word) + if target_word not in mem.metadata.tags: + mem.metadata.tags.append(target_word) + pick_index.append(i) + update_memories.append(mem) + + update_memories_embed = self._retry_db_operation( + lambda: self._batch_embed([mem.memory for mem in update_memories]) + ) + for _i, embed in zip(range(len(update_memories)), update_memories_embed, strict=False): + update_memories[_i].metadata.embedding = embed + + update_results = [] + with ContextThreadPoolExecutor(max_workers=10) as executor: + future_to_info = {} + for new_mem, old_idx in zip(update_memories, pick_index, strict=False): + old_mem = retrieved_memories[old_idx] + + future = executor.submit( + self._single_update_operation, + old_mem, + new_mem, + user_id, + user_name, + ) + future_to_info[future] = old_mem.id + + for future in future_to_info: + try: + result = future.result() + update_results.append(result) + except Exception as e: + mem_id = future_to_info[future][0] + self.logger.error( + f"[Feedback Core DB] Exception during update operation for memory {mem_id}: {e}" + ) + + return {"record": {"add": [], "update": update_results}} + def process_feedback_core( self, user_id: str, @@ -497,19 +663,28 @@ def check_validity(item): and "tags" in item ) + if feedback_content.strip() == "": + return {"record": {"add": [], "update": []}} try: feedback_time = kwargs.get("feedback_time") or datetime.now().isoformat() session_id = kwargs.get("session_id") - if feedback_content.strip() == "": - return {"record": {"add": [], "update": []}} - info = {"user_id": user_id, "user_name": user_name, "session_id": session_id} logger.info( f"[Feedback Core: process_feedback_core] Starting memory feedback process for user {user_name}" ) + # feedback keywords update + kwp_judge = self._keyword_replace_judgement(feedback_content) + if ( + kwp_judge + and kwp_judge["if_keyword_replace"].lower() == "true" + and kwp_judge.get("original", "NONE") != "NONE" + and kwp_judge.get("target", "NONE") != "NONE" + ): + return self.process_keyword_replace(user_id, user_name, kwp_judge=kwp_judge) + + # llm update memory if not chat_history: return self._pure_add(user_name, feedback_content, feedback_time, info) - else: raw_judge = self._feedback_judgement( chat_history, feedback_content, feedback_time=feedback_time @@ -533,17 +708,9 @@ def check_validity(item): feedback_memories = [] corrected_infos = [item["corrected_info"] for item in valid_feedback] - embed_bs = 5 - feedback_memories_embeddings = [] - for i in range(0, len(corrected_infos), embed_bs): - batch = corrected_infos[i : i + embed_bs] - try: - feedback_memories_embeddings.extend(self.embedder.embed(batch)) - except Exception as e: - logger.error( - f"[Feedback Core: process_feedback_core] Embedding batch failed: {e}", - exc_info=True, - ) + feedback_memories_embeddings = self._retry_db_operation( + lambda: self._batch_embed(corrected_infos) + ) for item, embedding in zip( valid_feedback, feedback_memories_embeddings, strict=False @@ -664,3 +831,16 @@ def _retry_db_operation(self, operation): f"[MemFeedback: _retry_db_operation] DB operation failed: {e}", exc_info=True ) raise + + @require_python_package( + import_name="jieba", + install_command="pip install jieba", + install_link="https://github.com/fxsjy/jieba", + ) + def _tokenize_chinese(self, text): + """split zh jieba""" + import jieba + + tokens = jieba.lcut(text) + tokens = [token.strip() for token in tokens if token.strip()] + return self.stopword_manager.filter_words(tokens) diff --git a/src/memos/mem_feedback/simple_feedback.py b/src/memos/mem_feedback/simple_feedback.py index 01132eb97..bb5a1c552 100644 --- a/src/memos/mem_feedback/simple_feedback.py +++ b/src/memos/mem_feedback/simple_feedback.py @@ -5,6 +5,7 @@ from memos.mem_feedback.feedback import MemFeedback from memos.mem_reader.simple_struct import SimpleStructMemReader from memos.memories.textual.tree_text_memory.organize.manager import MemoryManager +from memos.memories.textual.tree_text_memory.retrieve.retrieve_utils import StopwordManager from memos.memories.textual.tree_text_memory.retrieve.searcher import Searcher @@ -27,3 +28,4 @@ def __init__( self.memory_manager = memory_manager self.mem_reader = mem_reader self.searcher = searcher + self.stopword_manager = StopwordManager diff --git a/src/memos/mem_feedback/utils.py b/src/memos/mem_feedback/utils.py new file mode 100644 index 000000000..b290993cd --- /dev/null +++ b/src/memos/mem_feedback/utils.py @@ -0,0 +1,86 @@ +from memos.memories.textual.item import TextualMemoryItem + + +def estimate_tokens(text: str) -> int: + """ + Estimate the approximate number of tokens for the text + """ + if not text: + return 0 + + chinese_chars = sum(1 for char in text if "\u4e00" <= char <= "\u9fff") + + english_parts = text.split() + english_words = 0 + for part in english_parts: + has_chinese = any("\u4e00" <= char <= "\u9fff" for char in part) + if not has_chinese and any(c.isalpha() for c in part): + english_words += 1 + + other_chars = len(text) - chinese_chars + + estimated_tokens = int(chinese_chars * 1.5 + english_words * 1.33 + other_chars * 0.5) + + return max(1, estimated_tokens) + + +def should_keep_update(new_text: str, old_text: str) -> bool: + """ + Determine whether the update should be skipped + Rule: + 1. If the length of old_text is less than 50 and the modification ratio is less than 50% => returns True + 2. If the length of old_text is greater than or equal to 50 and the modification ratio is less than 15% => returns True + 3. Return False in other cases + """ + + old_len = estimate_tokens(old_text) + + def calculate_similarity(text1: str, text2: str) -> float: + set1 = set(text1) + set2 = set(text2) + if not set1 and not set2: + return 1.0 + + intersection = len(set1.intersection(set2)) + union = len(set1.union(set2)) + return intersection / union if union > 0 else 0.0 + + similarity = calculate_similarity(old_text, new_text) + change_ratio = 1 - similarity + + if old_len < 50: + return change_ratio < 0.5 + else: + return change_ratio < 0.15 + + +def split_into_chunks(memories: list[TextualMemoryItem], max_tokens_per_chunk=500): + chunks = [] + current_chunk = [] + current_tokens = 0 + + for item in memories: + item_text = f"{item.id}: {item.memory}" + item_tokens = estimate_tokens(item_text) + + if item_tokens > max_tokens_per_chunk: + if current_chunk: + chunks.append(current_chunk) + current_chunk = [] + + chunks.append([item]) + current_tokens = 0 + + elif current_tokens + item_tokens <= max_tokens_per_chunk: + current_chunk.append(item) + current_tokens += item_tokens + else: + if current_chunk: + chunks.append(current_chunk) + current_chunk = [item] + current_tokens = item_tokens + + if current_chunk: + chunks.append(current_chunk) + + return chunks diff --git a/src/memos/memories/textual/item.py b/src/memos/memories/textual/item.py index 63a5b3b1d..8067c7f72 100644 --- a/src/memos/memories/textual/item.py +++ b/src/memos/memories/textual/item.py @@ -90,7 +90,7 @@ class TextualMemoryMetadata(BaseModel): model_config = ConfigDict(extra="allow") - covered_history: str | None = Field( + covered_history: Any | None = Field( default=None, description="Record the memory id covered by the update", ) diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py b/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py index f428bf5c0..4ebf73a23 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/searcher.py @@ -173,6 +173,10 @@ def search( user_name=user_name, ) + full_recall = kwargs.get("full_recall", False) + if full_recall: + return retrieved_results + final_results = self.post_retrieve( retrieved_results=retrieved_results, top_k=top_k, diff --git a/src/memos/templates/mem_feedback_prompts.py b/src/memos/templates/mem_feedback_prompts.py index f7f2e8cb4..cd0c46a61 100644 --- a/src/memos/templates/mem_feedback_prompts.py +++ b/src/memos/templates/mem_feedback_prompts.py @@ -1,3 +1,119 @@ +KEYWORDS_REPLACE = """ +**Instruction:** +Please analyze the user's input text to determine if it is a "keyword replacement" request. If yes, follow these steps: + +1. **Identify the request type**: Confirm whether the user is asking to replace a specific word or phrase with another **within a specified scope**. +2. **Extract the modification scope**: Determine the scope where the modification should apply. + - If the user mentions a specific **document, file, or material identifier** (e.g., "in the Q1 operations plan", "in the prospectus numbered BT7868"), extract this description as the document scope. + - **If the user does not explicitly specify any scope, mark the scope as "NONE"**. +3. **Extract the original term (A)**: Identify the original word or phrase the user wants to be replaced. +4. **Extract the target term (B)**: Identify the target word or phrase the user wants to replace it with. + +**Output JSON Format**: +{{ + "if_keyword_replace": "true" | "false", + "doc_scope": "[Extracted specific file or document description]" | "NONE" | null, + "original": "[Extracted original word or phrase A]" | null, + "target": "[Extracted target word or phrase B]" | null +}} +- **If it is NOT a replacement request**, set `if_keyword_replace` to `"false"`, and set the values for `doc_scope`, `original`, and `target` to `null`. +- **If it IS a replacement request**, set `if_keyword_replace` to `"true"` and fill in the remaining fields. If the user did not specify a scope, set `doc_scope` to `"NONE"`. + +**Examples**: + +1. **User Input**: "In the file `User_Agreement.docx`, replace 'Party B' with 'User'." + **Output**: + {{ + "if_keyword_replace": "true", + "doc_scope": "User_Agreement.docx", + "original": "Party B", + "target": "User" + }} + +2. **User Input**: "Change 'Homepage' to 'Front Page'." + **Output**: + {{ + "if_keyword_replace": "true", + "doc_scope": "NONE", + "original": "Homepage", + "target": "Front Page" + }} + +3. **User Input**: "Does this sentence need modification?" + **Output**: + {{ + "if_keyword_replace": "false", + "doc_scope": null, + "original": null, + "target": null + }} + +**User Input** +{user_feedback} + +**Output**: +""" + + +KEYWORDS_REPLACE_ZH = """ +**指令：** +请分析用户输入的文本，判断是否为“关键词替换”需求。 如果是，请按以下步骤处理： + +1. **识别需求类型**：确认用户是否要求将**特定范围**内的某个词或短语替换为另一个词或短语。 +2. **提取修改范围**：确定用户指定的修改生效范围。 + - 如果用户提及了具体的**文档、文件或资料标识**（如“在第一季运营方案”、“编号为BT7868的招股书”），则提取此描述作为文件范围。 + - **如果用户未明确指定任何范围，则范围标记为 "NONE"**。 +3. **提取原始词汇（A）**：找出用户希望被替换的原始词或短语。 +4. **提取目标词汇（B）**：找出用户希望替换成的目标词或短语。 + +**输出JSON格式**： +{{ + "if_keyword_replace": "true" | "false", + "doc_scope": "[提取的具体文件或文档描述]" | "NONE" | null, + "original": "[提取的原始词或短语A]" | null, + "target": "[提取的目标词或短语B]" | null +}} +- **如果不是替换需求**，将 `if_keyword_replace` 设为 `"false"`，并将 `doc_scope`、`original`、`target` 三个键的值都设为 `null`。 +- **如果是替换需求**，将 `if_keyword_replace` 设为 `"true"`，并填充其余字段。如果用户未指定范围，`doc_scope` 设为 `"NONE"`。 + + +**示例**： + +1. **用户输入**：“在`用户协议.docx`这个文件中，把‘乙方’替换为‘用户’。” + **输出**： + {{ + "if_keyword_replace": "true", + "doc_scope": "用户协议.docx", + "original": "乙方", + "target": "用户" + }} + +2. **用户输入**：“把‘主页’改成‘首页’。” + **输出**： + {{ + "if_keyword_replace": "true", + "doc_scope": "NONE", + "original": "主页", + "target": "首页" + }} + +3. **用户输入**：“这个句子需要修改吗？” + **输出**： + {{ + "if_keyword_replace": "false", + "doc_scope": null, + "original": null, + "target": null + }} + + +**用户输入** +{user_feedback} + +**输出**： +""" + + FEEDBACK_JUDGEMENT_PROMPT = """You are a answer quality analysis expert. Please strictly follow the steps and criteria below to analyze the provided "User and Assistant Chat History" and "User Feedback," and fill the final evaluation results into the specified JSON format. Analysis Steps and Criteria: From 699cdf7bb07ddaafc451ba49d3a29f9b5d393d3d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 3 Dec 2025 18:10:34 +0800 Subject: [PATCH 34/72] update stopwords --- .../retrieve/retrieve_utils.py | 22 +------------------ 1 file changed, 1 insertion(+), 21 deletions(-) diff --git a/src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py b/src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py index 9e1e6c240..5a82883c8 100644 --- a/src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py +++ b/src/memos/memories/textual/tree_text_memory/retrieve/retrieve_utils.py @@ -93,12 +93,6 @@ def find_project_root(marker=".git"): return Path(".") -PROJECT_ROOT = find_project_root() -DEFAULT_STOPWORD_FILE = ( - PROJECT_ROOT / "examples" / "data" / "config" / "stopwords.txt" -) # cause time delay - - class StopwordManager: _stopwords = None @@ -109,13 +103,7 @@ def _load_stopwords(cls): return cls._stopwords stopwords = set() - try: - with open(DEFAULT_STOPWORD_FILE, encoding="utf-8") as f: - stopwords = {line.strip() for line in f if line.strip()} - logger.info("Stopwords loaded successfully.") - except Exception as e: - logger.warning(f"Error loading stopwords: {e}, using default stopwords.") - stopwords = cls._load_default_stopwords() + stopwords = cls._load_default_stopwords() cls._stopwords = stopwords return stopwords @@ -370,14 +358,6 @@ def is_stopword(cls, word): cls._load_stopwords() return word in cls._stopwords - @classmethod - def reload_stopwords(cls, file_path=None): - cls._stopwords = None - if file_path: - global DEFAULT_STOPWORD_FILE - DEFAULT_STOPWORD_FILE = file_path - cls._load_stopwords() - class FastTokenizer: def __init__(self, use_jieba=True, use_stopwords=True): From 343eeb399ed60d19f71bdca9a26549c77c2f9c83 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Thu, 4 Dec 2025 11:19:48 +0800 Subject: [PATCH 35/72] fix messages queue --- src/memos/mem_scheduler/general_scheduler.py | 14 ++------------ 1 file changed, 2 insertions(+), 12 deletions(-) diff --git a/src/memos/mem_scheduler/general_scheduler.py b/src/memos/mem_scheduler/general_scheduler.py index ad34530bc..46b6aba1f 100644 --- a/src/memos/mem_scheduler/general_scheduler.py +++ b/src/memos/mem_scheduler/general_scheduler.py @@ -626,18 +626,8 @@ def _extract_fields(mem_item): or mem_item.get("original_content") ) source_doc_id = None - if isinstance(mem_item, dict): - source_doc_id = ( - mem_item.get("source_doc_id") - or mem_item.get("doc_id") - or (mem_item.get("metadata") or {}).get("source_doc_id") - ) - else: - metadata = getattr(mem_item, "metadata", None) - if metadata: - source_doc_id = getattr(metadata, "source_doc_id", None) or getattr( - metadata, "doc_id", None - ) + if "archived_id" in mem_item: + source_doc_id = mem_item.get("archived_id") return mem_id, mem_memory, original_content, source_doc_id From d66e8ce9fb7af08f641b47a86153639fa76ec33b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Sun, 7 Dec 2025 15:46:36 +0800 Subject: [PATCH 36/72] add seach_by_keywords_LIKE --- src/memos/graph_dbs/polardb.py | 99 ++++++++++++++++++++++- src/memos/mem_feedback/feedback.py | 26 ++++-- src/memos/mem_feedback/simple_feedback.py | 1 + 3 files changed, 116 insertions(+), 10 deletions(-) diff --git a/src/memos/graph_dbs/polardb.py b/src/memos/graph_dbs/polardb.py index 638eac9c2..458bc29aa 100644 --- a/src/memos/graph_dbs/polardb.py +++ b/src/memos/graph_dbs/polardb.py @@ -1458,7 +1458,97 @@ def get_context_chain(self, id: str, type: str = "FOLLOWS") -> list[str]: raise NotImplementedError @timed - def seach_by_keywords( + def seach_by_keywords_like( + self, + query_word: str, + scope: str | None = None, + status: str | None = None, + search_filter: dict | None = None, + user_name: str | None = None, + filter: dict | None = None, + knowledgebase_ids: list[str] | None = None, + **kwargs, + ) -> list[dict]: + where_clauses = [] + + if scope: + where_clauses.append( + f"ag_catalog.agtype_access_operator(properties, '\"memory_type\"'::agtype) = '\"{scope}\"'::agtype" + ) + if status: + where_clauses.append( + f"ag_catalog.agtype_access_operator(properties, '\"status\"'::agtype) = '\"{status}\"'::agtype" + ) + else: + where_clauses.append( + "ag_catalog.agtype_access_operator(properties, '\"status\"'::agtype) = '\"activated\"'::agtype" + ) + + # Build user_name filter with knowledgebase_ids support (OR relationship) using common method + user_name_conditions = self._build_user_name_and_kb_ids_conditions_sql( + user_name=user_name, + knowledgebase_ids=knowledgebase_ids, + default_user_name=self.config.user_name, + ) + + # Add OR condition if we have any user_name conditions + if user_name_conditions: + if len(user_name_conditions) == 1: + where_clauses.append(user_name_conditions[0]) + else: + where_clauses.append(f"({' OR '.join(user_name_conditions)})") + + # Add search_filter conditions + if search_filter: + for key, value in search_filter.items(): + if isinstance(value, str): + where_clauses.append( + f"ag_catalog.agtype_access_operator(properties, '\"{key}\"'::agtype) = '\"{value}\"'::agtype" + ) + else: + where_clauses.append( + f"ag_catalog.agtype_access_operator(properties, '\"{key}\"'::agtype) = {value}::agtype" + ) + + # Build filter conditions using common method + filter_conditions = self._build_filter_conditions_sql(filter) + where_clauses.extend(filter_conditions) + + # Build key + where_clauses.append("""(properties -> '"memory"')::text LIKE %s""") + where_clause = f"WHERE {' AND '.join(where_clauses)}" if where_clauses else "" + + query = f""" + SELECT + ag_catalog.agtype_access_operator(properties, '"id"'::agtype) AS old_id, + agtype_object_field_text(properties, 'memory') as memory_text + FROM "{self.db_name}_graph"."Memory" + {where_clause} + """ + + params = (query_word,) + logger.info( + f"[seach_by_keywords_LIKE start:] user_name: {user_name}, query: {query}, params: {params}" + ) + conn = self._get_connection() + try: + with conn.cursor() as cursor: + cursor.execute(query, params) + results = cursor.fetchall() + output = [] + for row in results: + oldid = row[0] + id_val = str(oldid) + output.append({"id": id_val}) + logger.info( + f"[seach_by_keywords_LIKE end:] user_name: {user_name}, query: {query}, params: {params} recalled: {output}" + ) + return output + finally: + self._return_connection(conn) + + @timed + def seach_by_keywords_tfidf( self, query_words: list[str], scope: str | None = None, @@ -1533,7 +1623,9 @@ def seach_by_keywords( """ params = (tsquery_string,) - logger.info(f"[search_by_fulltext] query: {query}, params: {params}") + logger.info( + f"[seach_by_keywords_TFIDF start:] user_name: {user_name}, query: {query}, params: {params}" + ) conn = self._get_connection() try: with conn.cursor() as cursor: @@ -1545,6 +1637,9 @@ def seach_by_keywords( id_val = str(oldid) output.append({"id": id_val}) + logger.info( + f"[seach_by_keywords_TFIDF end:] user_name: {user_name}, query: {query}, params: {params} recalled: {output}" + ) return output finally: self._return_connection(conn) diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index eed43d66e..370470534 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -78,6 +78,7 @@ def __init__(self, config: MemFeedbackConfig): is_reorganize=self.is_reorganize, ) self.searcher: Searcher = self.memory_manager.searcher + self.DB_IDX_READY = False def _batch_embed(self, texts: list[str], embed_bs: int = 5): embed_bs = 5 @@ -569,15 +570,24 @@ def process_keyword_replace(self, user_id: str, user_name: str, kwp_judge: dict original_word = kwp_judge.get("original") target_word = kwp_judge.get("target") - # retrieve - lang = detect_lang(original_word) - queries = self._tokenize_chinese(original_word) if lang == "zh" else original_word.split() + if self.DB_IDX_READY: + # retrieve + lang = detect_lang(original_word) + queries = ( + self._tokenize_chinese(original_word) if lang == "zh" else original_word.split() + ) - must_part = f"{' & '.join(queries)}" if len(queries) > 1 else queries[0] - retrieved_ids = self.graph_store.seach_by_keywords([must_part], user_name=user_name) - if len(retrieved_ids) < 1: - retrieved_ids = self.graph_store.search_by_fulltext( - queries, top_k=100, user_name=user_name + must_part = f"{' & '.join(queries)}" if len(queries) > 1 else queries[0] + retrieved_ids = self.graph_store.seach_by_keywords_tfidf( + [must_part], user_name=user_name + ) + if len(retrieved_ids) < 1: + retrieved_ids = self.graph_store.search_by_fulltext( + queries, top_k=100, user_name=user_name + ) + else: + retrieved_ids = self.graph_store.seach_by_keywords_like( + f"%{original_word}%", user_name=user_name ) # filter by doc scope diff --git a/src/memos/mem_feedback/simple_feedback.py b/src/memos/mem_feedback/simple_feedback.py index bb5a1c552..478fa104f 100644 --- a/src/memos/mem_feedback/simple_feedback.py +++ b/src/memos/mem_feedback/simple_feedback.py @@ -29,3 +29,4 @@ def __init__( self.mem_reader = mem_reader self.searcher = searcher self.stopword_manager = StopwordManager + self.DB_IDX_READY = False From ae60994c998b0af1ffc8b26c2f8d9c578be5862a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 9 Dec 2025 16:57:10 +0800 Subject: [PATCH 37/72] add doc filter --- src/memos/api/handlers/component_init.py | 1 + src/memos/api/handlers/feedback_handler.py | 2 +- src/memos/graph_dbs/polardb.py | 6 +- src/memos/mem_feedback/feedback.py | 82 ++++++++++++++-------- src/memos/mem_feedback/simple_feedback.py | 3 + src/memos/mem_feedback/utils.py | 24 ++++++- 6 files changed, 82 insertions(+), 36 deletions(-) diff --git a/src/memos/api/handlers/component_init.py b/src/memos/api/handlers/component_init.py index 632c2ed4c..670a19110 100644 --- a/src/memos/api/handlers/component_init.py +++ b/src/memos/api/handlers/component_init.py @@ -304,6 +304,7 @@ def init_server() -> dict[str, Any]: memory_manager=memory_manager, mem_reader=mem_reader, searcher=searcher, + reranker=reranker, ) # Initialize Scheduler diff --git a/src/memos/api/handlers/feedback_handler.py b/src/memos/api/handlers/feedback_handler.py index cf5c536ea..217bca7cd 100644 --- a/src/memos/api/handlers/feedback_handler.py +++ b/src/memos/api/handlers/feedback_handler.py @@ -28,7 +28,7 @@ def __init__(self, dependencies: HandlerDependencies): dependencies: HandlerDependencies instance """ super().__init__(dependencies) - self._validate_dependencies("mem_reader", "mem_scheduler", "searcher") + self._validate_dependencies("mem_reader", "mem_scheduler", "searcher", "reranker") def handle_feedback_memories(self, feedback_req: APIFeedbackRequest) -> MemoryResponse: """ diff --git a/src/memos/graph_dbs/polardb.py b/src/memos/graph_dbs/polardb.py index 517005c9d..df365fc80 100644 --- a/src/memos/graph_dbs/polardb.py +++ b/src/memos/graph_dbs/polardb.py @@ -1597,9 +1597,7 @@ def seach_by_keywords_like( """ params = (query_word,) - logger.info( - f"[seach_by_keywords_LIKE start:] user_name: {user_name}, query: {query}, params: {params}" - ) + logger.info(f"[seach_by_keywords_LIKE start:] user_name: {user_name}, params: {params}") conn = self._get_connection() try: with conn.cursor() as cursor: @@ -1611,7 +1609,7 @@ def seach_by_keywords_like( id_val = str(oldid) output.append({"id": id_val}) logger.info( - f"[seach_by_keywords_LIKE end:] user_name: {user_name}, query: {query}, params: {params} recalled: {output}" + f"[seach_by_keywords_LIKE end:] user_name: {user_name}, params: {params} recalled: {output}" ) return output finally: diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index b986f7f13..6ff2e0df5 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -15,10 +15,10 @@ from memos.graph_dbs.factory import GraphStoreFactory, PolarDBGraphDB from memos.llms.factory import AzureLLM, LLMFactory, OllamaLLM, OpenAILLM from memos.mem_feedback.base import BaseMemFeedback -from memos.mem_feedback.utils import should_keep_update, split_into_chunks +from memos.mem_feedback.utils import make_mem_item, should_keep_update, split_into_chunks from memos.mem_reader.factory import MemReaderFactory from memos.mem_reader.read_multi_modal import detect_lang -from memos.memories.textual.item import TextualMemoryItem, TreeNodeTextualMemoryMetadata +from memos.memories.textual.item import TextualMemoryItem from memos.memories.textual.tree_text_memory.organize.manager import ( MemoryManager, extract_working_binding_ids, @@ -77,7 +77,8 @@ def __init__(self, config: MemFeedbackConfig): }, is_reorganize=self.is_reorganize, ) - self.searcher: Searcher = self.memory_manager.searcher + self.searcher: Searcher = None + self.reranker = None self.DB_IDX_READY = False def _batch_embed(self, texts: list[str], embed_bs: int = 5): @@ -562,6 +563,35 @@ def _generate_answer( return self._get_llm_response(prompt, dsl=False) + def _doc_filter(self, doc_scope: str, memories: list[TextualMemoryItem]): + """ + Filter the memory based on filename + """ + filename2_memid = {} + filename_mems = [] + + for item in memories: + for file_info in item.metadata.sources: + if file_info.type == "file": + file_dict = file_info.original_part + filename = file_dict["file"]["filename"] + if filename not in filename2_memid: + filename2_memid[filename] = [] + filename_mems.append(make_mem_item(filename)) + filename2_memid[filename].append(item.id) + + rerank_res = self.reranker.rerank(doc_scope, filename_mems, top_k=100) + inscope_docs = [item[0].memory for item in rerank_res if item[1] > 0.95] + + inscope_ids = [ + memid for inscope_file in inscope_docs for memid in filename2_memid[inscope_file] + ] + logger.info( + f"[Feedback Core: process_keyword_replace] These docs are in scope : {inscope_docs}, relared memids: {inscope_ids}" + ) + filter_memories = [mem for mem in memories if mem.id in inscope_ids] + return filter_memories + def process_keyword_replace(self, user_id: str, user_name: str, kwp_judge: dict | None = None): """ memory keyword replace process @@ -597,11 +627,7 @@ def process_keyword_replace(self, user_id: str, user_name: str, kwp_judge: dict retrieved_memories = [TextualMemoryItem(**item) for item in mem_data] if doc_scope != "NONE": - retrieved_memories = [ - item - for item in retrieved_memories - if doc_scope in item.metadata.sources # TODO - ] + retrieved_memories = self._doc_filter(doc_scope, retrieved_memories) if not retrieved_memories: return {"record": {"add": [], "update": []}} @@ -728,29 +754,25 @@ def check_validity(item): value = item["corrected_info"] key = item["key"] tags = item["tags"] - feedback_memories.append( - TextualMemoryItem( - memory=value, - metadata=TreeNodeTextualMemoryMetadata( - user_id=info.get("user_id", ""), - session_id=info.get("session_id", ""), - memory_type="LongTermMemory", - status="activated", - tags=tags, - key=key, - embedding=embedding, - usage=[], - sources=[{"type": "chat"}], - user_name=user_name, - background="[Feedback update background]: " - + str(chat_history) - + "\nUser feedback: " - + str(feedback_content), - confidence=0.99, - type="fine", - ), - ) + background = ( + "[Feedback update background]: " + + str(chat_history) + + "\nUser feedback: " + + str(feedback_content) + ) + mem_item = make_mem_item( + value, + user_id=user_id, + user_name=user_name, + session_id=session_id, + tags=tags, + key=key, + embedding=embedding, + sources=[{"type": "chat"}], + background=background, + type="fine", ) + feedback_memories.append(mem_item) mem_record = self._feedback_memory( user_id, diff --git a/src/memos/mem_feedback/simple_feedback.py b/src/memos/mem_feedback/simple_feedback.py index 478fa104f..429c2ea20 100644 --- a/src/memos/mem_feedback/simple_feedback.py +++ b/src/memos/mem_feedback/simple_feedback.py @@ -7,6 +7,7 @@ from memos.memories.textual.tree_text_memory.organize.manager import MemoryManager from memos.memories.textual.tree_text_memory.retrieve.retrieve_utils import StopwordManager from memos.memories.textual.tree_text_memory.retrieve.searcher import Searcher +from memos.reranker.base import BaseReranker logger = log.get_logger(__name__) @@ -21,6 +22,7 @@ def __init__( memory_manager: MemoryManager, mem_reader: SimpleStructMemReader, searcher: Searcher, + reranker: BaseReranker, ): self.llm = llm self.embedder = embedder @@ -29,4 +31,5 @@ def __init__( self.mem_reader = mem_reader self.searcher = searcher self.stopword_manager = StopwordManager + self.reranker = reranker self.DB_IDX_READY = False diff --git a/src/memos/mem_feedback/utils.py b/src/memos/mem_feedback/utils.py index b290993cd..24817a1d3 100644 --- a/src/memos/mem_feedback/utils.py +++ b/src/memos/mem_feedback/utils.py @@ -1,4 +1,4 @@ -from memos.memories.textual.item import TextualMemoryItem +from memos.memories.textual.item import TextualMemoryItem, TreeNodeTextualMemoryMetadata def estimate_tokens(text: str) -> int: @@ -84,3 +84,25 @@ def split_into_chunks(memories: list[TextualMemoryItem], max_tokens_per_chunk=50 chunks.append(current_chunk) return chunks + + +def make_mem_item(text: str, **kwargs) -> TextualMemoryItem: + """Build a minimal TextualMemoryItem.""" + return TextualMemoryItem( + memory=text, + metadata=TreeNodeTextualMemoryMetadata( + user_id=kwargs.get("user_id", ""), + session_id=kwargs.get("session_id", ""), + memory_type="LongTermMemory", + status="activated", + tags=kwargs.get("tags", []), + key=kwargs.get("key", ""), + embedding=kwargs.get("embedding", []), + usage=[], + sources=kwargs.get("sources", []), + user_name=kwargs.get("user_name", ""), + background=kwargs.get("background", ""), + confidence=0.99, + type=kwargs.get("type", ""), + ), + ) From 7b0f2f4d1489c1a7a87c2f78572f096c581c32e8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 9 Dec 2025 19:57:01 +0800 Subject: [PATCH 38/72] add retrieve query --- src/memos/mem_feedback/feedback.py | 43 +++++++++++++++++++++++++----- src/memos/mem_feedback/utils.py | 6 ++--- 2 files changed, 40 insertions(+), 9 deletions(-) diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index 6ff2e0df5..ebe7b9f24 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -277,15 +277,30 @@ def semantics_feedback( user_name: str, memory_item: TextualMemoryItem, current_memories: list[TextualMemoryItem], - fact_history: str, + history_str: str, + chat_history_list: list, ): + """Modify memory at the semantic level""" lang = detect_lang("".join(memory_item.memory)) template = FEEDBACK_PROMPT_DICT["compare"][lang] if current_memories == []: - current_memories = self._retrieve( + # retrieve feedback + feedback_retrieved = self._retrieve( memory_item.memory, info={"user_id": user_id}, user_name=user_name ) + # retrieve question + last_user_index = max(i for i, d in enumerate(chat_history_list) if d["role"] == "user") + last_qa = " ".join([item["content"] for item in chat_history_list[last_user_index:]]) + supplementary_retrieved = self._retrieve( + last_qa, info={"user_id": user_id}, user_name=user_name + ) + ids = [] + for item in feedback_retrieved + supplementary_retrieved: + if item.id not in ids: + ids.append(item.id) + current_memories.append(item) + if not current_memories: operations = [{"operation": "ADD"}] else: @@ -301,7 +316,7 @@ def semantics_feedback( prompt = template.format( current_memories=current_memories_str, new_facts=memory_item.memory, - chat_history=fact_history, + chat_history=history_str, ) future = executor.submit(self._get_llm_response, prompt) @@ -381,7 +396,7 @@ def _feedback_memory( feedback_content = kwargs.get("feedback_content", "") chat_history_lis = [f"""{msg["role"]}: {msg["content"]}""" for msg in chat_history[-4:]] - fact_history = "\n".join(chat_history_lis) + f"\nuser feedback: \n{feedback_content}" + history_str = "\n".join(chat_history_lis) + f"\nuser feedback: \n{feedback_content}" retrieved_memories = [ self.graph_store.get_node(_id, user_name=user_name) for _id in retrieved_memory_ids @@ -403,7 +418,13 @@ def _feedback_memory( with ContextThreadPoolExecutor(max_workers=3) as ex: futures = { ex.submit( - self.semantics_feedback, user_id, user_name, mem, current_memories, fact_history + self.semantics_feedback, + user_id, + user_name, + mem, + current_memories, + history_str, + chat_history, ): i for i, mem in enumerate(feedback_memories) } @@ -543,7 +564,17 @@ def correct_item(data): return None dehallu_res = [correct_item(item) for item in operations] - return [item for item in dehallu_res if item] + llm_operations = [item for item in dehallu_res if item] + + # Update takes precedence over add + has_update = any(item.get("operation").lower() == "update" for item in llm_operations) + if has_update: + filtered_items = [ + item for item in llm_operations if item.get("operation").lower() != "add" + ] + return filtered_items + else: + return llm_operations def _generate_answer( self, chat_history: list[MessageDict], feedback_content: str, corrected_answer: bool diff --git a/src/memos/mem_feedback/utils.py b/src/memos/mem_feedback/utils.py index 24817a1d3..ae0e0ba80 100644 --- a/src/memos/mem_feedback/utils.py +++ b/src/memos/mem_feedback/utils.py @@ -48,13 +48,13 @@ def calculate_similarity(text1: str, text2: str) -> float: similarity = calculate_similarity(old_text, new_text) change_ratio = 1 - similarity - if old_len < 50: + if old_len < 200: return change_ratio < 0.5 else: - return change_ratio < 0.15 + return change_ratio < 0.2 -def split_into_chunks(memories: list[TextualMemoryItem], max_tokens_per_chunk=500): +def split_into_chunks(memories: list[TextualMemoryItem], max_tokens_per_chunk: int = 500): chunks = [] current_chunk = [] current_tokens = 0 From 005a5bb42e1cd2983d070d63011fdba5146c8a88 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 10 Dec 2025 14:47:26 +0800 Subject: [PATCH 39/72] add retrieve queies --- src/memos/mem_feedback/feedback.py | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index ebe7b9f24..c8e081937 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -23,6 +23,7 @@ MemoryManager, extract_working_binding_ids, ) +from memos.memories.textual.tree_text_memory.retrieve.retrieve_utils import StopwordManager if TYPE_CHECKING: @@ -77,6 +78,7 @@ def __init__(self, config: MemFeedbackConfig): }, is_reorganize=self.is_reorganize, ) + self.stopword_manager = StopwordManager self.searcher: Searcher = None self.reranker = None self.DB_IDX_READY = False @@ -260,7 +262,6 @@ def _del_working_binding(self, user_name, mem_items: list[TextualMemoryItem]) -> for mid in delete_ids: try: - print("del", mid) self.graph_store.delete_node(mid, user_name=user_name) logger.info( @@ -482,8 +483,6 @@ def _vec_query(self, new_memories_embedding: list[float], user_name=None): self.graph_store.get_node(item["id"], user_name=user_name) for item in retrieved_ids ] - for item in current_memories: - print(item["id"], item["metadata"]["memory_type"], item["metadata"]["status"]) if not retrieved_ids: logger.info( f"[Feedback Core: _vec_query] No similar memories found for embedding query for user {user_name}." From d69e7f4cbffa6174210551fc7aaee1d31f1b3baa Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 10 Dec 2025 21:14:33 +0800 Subject: [PATCH 40/72] patch info filter --- src/memos/api/handlers/add_handler.py | 1 + src/memos/api/product_models.py | 13 ++++ src/memos/mem_feedback/feedback.py | 69 +++++++++++++++----- src/memos/mem_feedback/utils.py | 10 ++- src/memos/mem_scheduler/general_scheduler.py | 1 + src/memos/multi_mem_cube/single_cube.py | 1 + 6 files changed, 76 insertions(+), 19 deletions(-) diff --git a/src/memos/api/handlers/add_handler.py b/src/memos/api/handlers/add_handler.py index 2758c9e32..3cdbedabf 100644 --- a/src/memos/api/handlers/add_handler.py +++ b/src/memos/api/handlers/add_handler.py @@ -89,6 +89,7 @@ def _check_messages(messages: MessageList) -> None: feedback_content=feedback_content, writable_cube_ids=add_req.writable_cube_ids, async_mode=add_req.async_mode, + info=add_req.info, ) process_record = cube_view.feedback_memories(feedback_req) diff --git a/src/memos/api/product_models.py b/src/memos/api/product_models.py index 06cc29729..d583f3e1f 100644 --- a/src/memos/api/product_models.py +++ b/src/memos/api/product_models.py @@ -684,6 +684,19 @@ class APIFeedbackRequest(BaseRequest): "async", description="feedback mode: sync or async" ) corrected_answer: bool = Field(False, description="Whether need return corrected answer") + info: dict[str, Any] | None = Field( + None, + description=( + "Additional metadata for the add request. " + "All keys can be used as filters in search. " + "Example: " + "{'agent_id': 'xxxxxx', " + "'app_id': 'xxxx', " + "'source_type': 'web', " + "'source_url': 'https://www.baidu.com', " + "'source_content': 'West Lake is the most famous scenic spot in Hangzhou'}." + ), + ) # ==== mem_cube_id is NOT enabled==== mem_cube_id: str | None = Field( None, diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index c8e081937..3d650c17b 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -3,7 +3,7 @@ import json from datetime import datetime -from typing import TYPE_CHECKING +from typing import TYPE_CHECKING, Any from tenacity import retry, stop_after_attempt, wait_exponential @@ -280,27 +280,28 @@ def semantics_feedback( current_memories: list[TextualMemoryItem], history_str: str, chat_history_list: list, + info: dict, ): """Modify memory at the semantic level""" lang = detect_lang("".join(memory_item.memory)) template = FEEDBACK_PROMPT_DICT["compare"][lang] if current_memories == []: # retrieve feedback - feedback_retrieved = self._retrieve( - memory_item.memory, info={"user_id": user_id}, user_name=user_name - ) + feedback_retrieved = self._retrieve(memory_item.memory, info=info, user_name=user_name) # retrieve question last_user_index = max(i for i, d in enumerate(chat_history_list) if d["role"] == "user") last_qa = " ".join([item["content"] for item in chat_history_list[last_user_index:]]) - supplementary_retrieved = self._retrieve( - last_qa, info={"user_id": user_id}, user_name=user_name - ) + supplementary_retrieved = self._retrieve(last_qa, info=info, user_name=user_name) ids = [] for item in feedback_retrieved + supplementary_retrieved: if item.id not in ids: ids.append(item.id) current_memories.append(item) + include_keys = ["agent_id", "app_id"] + current_memories = [ + item for item in current_memories if self._info_comparison(item, info, include_keys) + ] if not current_memories: operations = [{"operation": "ADD"}] @@ -336,7 +337,6 @@ def semantics_feedback( operations = self.standard_operations(all_operations, current_memories) - # TODO based on the operation, change memory_item memory info ; change source info logger.info(f"[Feedback memory operations]: {operations!s}") if not operations: @@ -395,6 +395,7 @@ def _feedback_memory( retrieved_memory_ids = kwargs.get("retrieved_memory_ids") or [] chat_history = kwargs.get("chat_history", []) feedback_content = kwargs.get("feedback_content", "") + info = kwargs.get("info", {}) chat_history_lis = [f"""{msg["role"]}: {msg["content"]}""" for msg in chat_history[-4:]] history_str = "\n".join(chat_history_lis) + f"\nuser feedback: \n{feedback_content}" @@ -426,6 +427,7 @@ def _feedback_memory( current_memories, history_str, chat_history, + info, ): i for i, mem in enumerate(feedback_memories) } @@ -450,6 +452,17 @@ def _feedback_memory( } } + def _info_comparison(self, memory: TextualMemoryItem, _info: dict, include_keys: list) -> bool: + if not _info and not memory.metadata.info: + return True + + record = [] + for key in include_keys: + info_v = _info.get(key) + mem_v = memory.metadata.info.get(key, None) + record.append(info_v == mem_v) + return all(record) + def _retrieve(self, query: str, info=None, user_name=None): """Retrieve memory items""" retrieved_mems = self.searcher.search( @@ -622,13 +635,20 @@ def _doc_filter(self, doc_scope: str, memories: list[TextualMemoryItem]): filter_memories = [mem for mem in memories if mem.id in inscope_ids] return filter_memories - def process_keyword_replace(self, user_id: str, user_name: str, kwp_judge: dict | None = None): + def process_keyword_replace( + self, user_id: str, user_name: str, kwp_judge: dict | None = None, info: dict | None = None + ): """ - memory keyword replace process + Memory keyword replace process """ + info = info or {} doc_scope = kwp_judge.get("doc_scope", "NONE") original_word = kwp_judge.get("original") target_word = kwp_judge.get("target") + include_keys = ["agent_id", "app_id"] + + mem_info = {key: info[key] for key in info if key in include_keys} + filter_dict = {f"info.{key}": info[key] for key in mem_info} if self.DB_IDX_READY: # retrieve @@ -639,22 +659,26 @@ def process_keyword_replace(self, user_id: str, user_name: str, kwp_judge: dict must_part = f"{' & '.join(queries)}" if len(queries) > 1 else queries[0] retrieved_ids = self.graph_store.seach_by_keywords_tfidf( - [must_part], user_name=user_name + [must_part], user_name=user_name, filter=filter_dict ) if len(retrieved_ids) < 1: retrieved_ids = self.graph_store.search_by_fulltext( - queries, top_k=100, user_name=user_name + queries, top_k=100, user_name=user_name, filter=filter_dict ) else: retrieved_ids = self.graph_store.seach_by_keywords_like( - f"%{original_word}%", user_name=user_name + f"%{original_word}%", user_name=user_name, filter=filter_dict ) - # filter by doc scope mem_data = [ self.graph_store.get_node(item["id"], user_name=user_name) for item in retrieved_ids ] retrieved_memories = [TextualMemoryItem(**item) for item in mem_data] + retrieved_memories = [ + item + for item in retrieved_memories + if self._info_comparison(item, mem_info, include_keys) + ] if doc_scope != "NONE": retrieved_memories = self._doc_filter(doc_scope, retrieved_memories) @@ -701,7 +725,7 @@ def process_keyword_replace(self, user_id: str, user_name: str, kwp_judge: dict update_results.append(result) except Exception as e: mem_id = future_to_info[future][0] - self.logger.error( + logger.error( f"[Feedback Core DB] Exception during update operation for memory {mem_id}: {e}" ) @@ -713,6 +737,7 @@ def process_feedback_core( user_name: str, chat_history: list[MessageDict], feedback_content: str, + info: dict | None = None, **kwargs, ) -> dict: """ @@ -734,7 +759,11 @@ def check_validity(item): try: feedback_time = kwargs.get("feedback_time") or datetime.now().isoformat() session_id = kwargs.get("session_id") - info = {"user_id": user_id, "user_name": user_name, "session_id": session_id} + if not info: + info = {"user_id": user_id, "user_name": user_name, "session_id": session_id} + else: + info.update({"user_id": user_id, "user_name": user_name, "session_id": session_id}) + logger.info( f"[Feedback Core: process_feedback_core] Starting memory feedback process for user {user_name}" ) @@ -746,7 +775,9 @@ def check_validity(item): and kwp_judge.get("original", "NONE") != "NONE" and kwp_judge.get("target", "NONE") != "NONE" ): - return self.process_keyword_replace(user_id, user_name, kwp_judge=kwp_judge) + return self.process_keyword_replace( + user_id, user_name, kwp_judge=kwp_judge, info=info + ) # llm update memory if not chat_history: @@ -801,6 +832,7 @@ def check_validity(item): sources=[{"type": "chat"}], background=background, type="fine", + info=info, ) feedback_memories.append(mem_item) @@ -810,6 +842,7 @@ def check_validity(item): feedback_memories, chat_history=chat_history, feedback_content=feedback_content, + info=info, **kwargs, ) logger.info( @@ -827,6 +860,7 @@ def process_feedback( user_name: str, chat_history: list[MessageDict], feedback_content: str, + info: dict[str, Any] | None = None, **kwargs, ): """ @@ -856,6 +890,7 @@ def process_feedback( user_name, chat_history, feedback_content, + info, **kwargs, ) done, pending = concurrent.futures.wait([answer_future, core_future], timeout=30) diff --git a/src/memos/mem_feedback/utils.py b/src/memos/mem_feedback/utils.py index ae0e0ba80..0033d85b4 100644 --- a/src/memos/mem_feedback/utils.py +++ b/src/memos/mem_feedback/utils.py @@ -88,11 +88,16 @@ def split_into_chunks(memories: list[TextualMemoryItem], max_tokens_per_chunk: i def make_mem_item(text: str, **kwargs) -> TextualMemoryItem: """Build a minimal TextualMemoryItem.""" + info = kwargs.get("info", {}) + info_ = info.copy() + user_id = info_.pop("user_id", "") + session_id = info_.pop("session_id", "") + return TextualMemoryItem( memory=text, metadata=TreeNodeTextualMemoryMetadata( - user_id=kwargs.get("user_id", ""), - session_id=kwargs.get("session_id", ""), + user_id=user_id, + session_id=session_id, memory_type="LongTermMemory", status="activated", tags=kwargs.get("tags", []), @@ -104,5 +109,6 @@ def make_mem_item(text: str, **kwargs) -> TextualMemoryItem: background=kwargs.get("background", ""), confidence=0.99, type=kwargs.get("type", ""), + info=info_, ), ) diff --git a/src/memos/mem_scheduler/general_scheduler.py b/src/memos/mem_scheduler/general_scheduler.py index 8f3eccecf..a9502c394 100644 --- a/src/memos/mem_scheduler/general_scheduler.py +++ b/src/memos/mem_scheduler/general_scheduler.py @@ -604,6 +604,7 @@ def _mem_feedback_message_consumer(self, messages: list[ScheduleMessageItem]) -> feedback_content=feedback_data.get("feedback_content"), feedback_time=feedback_data.get("feedback_time"), task_id=task_id, + info=feedback_data.get("info", None), ) logger.info( diff --git a/src/memos/multi_mem_cube/single_cube.py b/src/memos/multi_mem_cube/single_cube.py index f0157952b..71a34beb4 100644 --- a/src/memos/multi_mem_cube/single_cube.py +++ b/src/memos/multi_mem_cube/single_cube.py @@ -183,6 +183,7 @@ def feedback_memories(self, feedback_req: APIFeedbackRequest) -> dict[str, Any]: async_mode=feedback_req.async_mode, corrected_answer=feedback_req.corrected_answer, task_id=feedback_req.task_id, + info=feedback_req.info, ) self.logger.info(f"Feedback memories result: {feedback_result}") return feedback_result From eab3d80bc7bddd2dac56d9d0dd773fa53b923171 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Fri, 12 Dec 2025 17:03:38 +0800 Subject: [PATCH 41/72] add log and make embedding safety net --- src/memos/mem_feedback/feedback.py | 130 ++++++++++++-------- src/memos/multi_mem_cube/single_cube.py | 2 +- src/memos/templates/mem_feedback_prompts.py | 5 + 3 files changed, 86 insertions(+), 51 deletions(-) diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index 3d650c17b..fe46fbe62 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -5,15 +5,15 @@ from datetime import datetime from typing import TYPE_CHECKING, Any -from tenacity import retry, stop_after_attempt, wait_exponential +from tenacity import retry, stop_after_attempt, wait_random_exponential -from memos import log from memos.configs.memory import MemFeedbackConfig from memos.context.context import ContextThreadPoolExecutor from memos.dependency import require_python_package from memos.embedders.factory import EmbedderFactory, OllamaEmbedder from memos.graph_dbs.factory import GraphStoreFactory, PolarDBGraphDB from memos.llms.factory import AzureLLM, LLMFactory, OllamaLLM, OpenAILLM +from memos.log import get_logger from memos.mem_feedback.base import BaseMemFeedback from memos.mem_feedback.utils import make_mem_item, should_keep_update, split_into_chunks from memos.mem_reader.factory import MemReaderFactory @@ -48,7 +48,7 @@ "generation": {"en": FEEDBACK_ANSWER_PROMPT, "zh": FEEDBACK_ANSWER_PROMPT_ZH}, } -logger = log.get_logger(__name__) +logger = get_logger(__name__) class MemFeedback(BaseMemFeedback): @@ -83,19 +83,47 @@ def __init__(self, config: MemFeedbackConfig): self.reranker = None self.DB_IDX_READY = False + @require_python_package( + import_name="jieba", + install_command="pip install jieba", + install_link="https://github.com/fxsjy/jieba", + ) + def _tokenize_chinese(self, text): + """split zh jieba""" + import jieba + + tokens = jieba.lcut(text) + tokens = [token.strip() for token in tokens if token.strip()] + return self.stopword_manager.filter_words(tokens) + + @retry(stop=stop_after_attempt(4), wait=wait_random_exponential(multiplier=1, max=10)) + def _embed_once(self, texts): + return self.embedder.embed(texts) + + @retry(stop=stop_after_attempt(3), wait=wait_random_exponential(multiplier=1, min=4, max=10)) + def _retry_db_operation(self, operation): + try: + return operation() + except Exception as e: + logger.error( + f"[Feedback Core: _retry_db_operation] DB operation failed: {e}", exc_info=True + ) + raise + def _batch_embed(self, texts: list[str], embed_bs: int = 5): - embed_bs = 5 - texts_embeddings = [] + results = [] + dim = self.embedder.config.embedding_dims + for i in range(0, len(texts), embed_bs): batch = texts[i : i + embed_bs] try: - texts_embeddings.extend(self.embedder.embed(batch)) + results.extend(self._embed_once(batch)) except Exception as e: logger.error( - f"[Feedback Core: process_feedback_core] Embedding batch failed: {e}", - exc_info=True, + f"[Feedback Core: process_feedback_core] Embedding batch failed, Cover with all zeros: {len(batch)} entries: {e}" ) - return texts_embeddings + results.extend([[0.0] * dim for _ in range(len(batch))]) + return results def _pure_add(self, user_name: str, feedback_content: str, feedback_time: str, info: dict): """ @@ -108,7 +136,7 @@ def _pure_add(self, user_name: str, feedback_content: str, feedback_time: str, i lambda: self.memory_manager.add(to_add_memories, user_name=user_name) ) logger.info( - f"[Feedback Core: _pure_add] Added {len(added_ids)} memories for user {user_name}." + f"[Feedback Core: _pure_add] Pure added {len(added_ids)} memories for user {user_name}." ) return { "record": { @@ -199,7 +227,7 @@ def _single_add_operation( lambda: self.memory_manager.add([to_add_memory], user_name=user_name, mode=async_mode) ) - logger.info(f"[Memory Feedback ADD] {added_ids[0]}") + logger.info(f"[Memory Feedback ADD] memory id: {added_ids[0]}") return {"id": added_ids[0], "text": to_add_memory.memory} def _single_update_operation( @@ -305,10 +333,14 @@ def semantics_feedback( if not current_memories: operations = [{"operation": "ADD"}] + logger.warning( + "[Feedback Core]: There was no recall of the relevant memory, so it was added directly." + ) else: memory_chunks = split_into_chunks(current_memories, max_tokens_per_chunk=500) all_operations = [] + now_time = datetime.now().isoformat() with ContextThreadPoolExecutor(max_workers=10) as executor: future_to_chunk_idx = {} for chunk in memory_chunks: @@ -316,6 +348,7 @@ def semantics_feedback( [f"{item.id}: {item.memory}" for item in chunk] ) prompt = template.format( + now_time=now_time, current_memories=current_memories_str, new_facts=memory_item.memory, chat_history=history_str, @@ -337,7 +370,7 @@ def semantics_feedback( operations = self.standard_operations(all_operations, current_memories) - logger.info(f"[Feedback memory operations]: {operations!s}") + logger.info(f"[Feedback Core Operations]: {operations!s}") if not operations: return {"record": {"add": [], "update": []}} @@ -453,6 +486,7 @@ def _feedback_memory( } def _info_comparison(self, memory: TextualMemoryItem, _info: dict, include_keys: list) -> bool: + """Filter the relevant memory items based on info""" if not _info and not memory.metadata.info: return True @@ -463,10 +497,10 @@ def _info_comparison(self, memory: TextualMemoryItem, _info: dict, include_keys: record.append(info_v == mem_v) return all(record) - def _retrieve(self, query: str, info=None, user_name=None): + def _retrieve(self, query: str, info=None, top_k=100, user_name=None): """Retrieve memory items""" retrieved_mems = self.searcher.search( - query, info=info, user_name=user_name, topk=50, full_recall=True + query, info=info, user_name=user_name, top_k=top_k, full_recall=True ) retrieved_mems = [item[0] for item in retrieved_mems] return retrieved_mems @@ -524,11 +558,19 @@ def _get_llm_response(self, prompt: str, dsl: bool = True) -> dict: else: return response_text except Exception as e: - logger.error(f"[Feedback Core LLM] Exception during chat generation: {e}") + logger.error( + f"[Feedback Core LLM Error] Exception during chat generation: {e} | response_text： {response_text}" + ) response_json = None return response_json def standard_operations(self, operations, current_memories): + """ + Regularize the operation design + 1. Map the id to the correct original memory id + 2. If there is an update, skip the memory object of add + 3. If the modified text is too long, skip the update + """ right_ids = [item.id for item in current_memories] right_lower_map = {x.lower(): x for x in right_ids} @@ -582,9 +624,16 @@ def correct_item(data): has_update = any(item.get("operation").lower() == "update" for item in llm_operations) if has_update: filtered_items = [ + item for item in llm_operations if item.get("operation").lower() == "add" + ] + update_items = [ item for item in llm_operations if item.get("operation").lower() != "add" ] - return filtered_items + if filtered_items: + logger.info( + f"[Feedback Core: semantics_feedback] Due to have update objects, skip add: {filtered_items}" + ) + return update_items else: return llm_operations @@ -683,6 +732,10 @@ def process_keyword_replace( if doc_scope != "NONE": retrieved_memories = self._doc_filter(doc_scope, retrieved_memories) + logger.info( + f"[Feedback Core: process_keyword_replace] Keywords recalled memory for user {user_name}: {len(retrieved_ids)} memories | After filtering: {len(retrieved_memories)} memories." + ) + if not retrieved_memories: return {"record": {"add": [], "update": []}} @@ -693,14 +746,14 @@ def process_keyword_replace( if original_word in old_mem.memory: mem = old_mem.model_copy(deep=True) mem.memory = mem.memory.replace(original_word, target_word) + if original_word in mem.metadata.tags: + mem.metadata.tags.remove(original_word) if target_word not in mem.metadata.tags: mem.metadata.tags.append(target_word) pick_index.append(i) update_memories.append(mem) + update_memories_embed = self._batch_embed([mem.memory for mem in update_memories]) - update_memories_embed = self._retry_db_operation( - lambda: self._batch_embed([mem.memory for mem in update_memories]) - ) for _i, embed in zip(range(len(update_memories)), update_memories_embed, strict=False): update_memories[_i].metadata.embedding = embed @@ -805,9 +858,7 @@ def check_validity(item): feedback_memories = [] corrected_infos = [item["corrected_info"] for item in valid_feedback] - feedback_memories_embeddings = self._retry_db_operation( - lambda: self._batch_embed(corrected_infos) - ) + feedback_memories_embeddings = self._batch_embed(corrected_infos) for item, embedding in zip( valid_feedback, feedback_memories_embeddings, strict=False @@ -845,8 +896,10 @@ def check_validity(item): info=info, **kwargs, ) + add_memories = mem_record["record"]["add"] + update_memories = mem_record["record"]["update"] logger.info( - f"[Feedback Core: process_feedback_core] Processed {len(feedback_memories)} feedback memories for user {user_name}." + f"[Feedback Core: process_feedback_core] Processed {len(feedback_memories)} feedback | add {len(add_memories)} memories | update {len(update_memories)} memories for user {user_name}." ) return mem_record @@ -902,42 +955,19 @@ def process_feedback( task_id = kwargs.get("task_id", "default") logger.info( - f"[MemFeedback process] Feedback Completed : user {user_name} | task_id {task_id} | record {record}." + f"[Feedback Core MemFeedback process] Feedback Completed : user {user_name} | task_id {task_id} | record {record}." ) return {"answer": answer, "record": record["record"]} except concurrent.futures.TimeoutError: logger.error( - f"[MemFeedback process] Timeout in sync mode for {user_name}", exc_info=True + f"[Feedback Core MemFeedback process] Timeout in sync mode for {user_name}", + exc_info=True, ) return {"answer": "", "record": {"add": [], "update": []}} except Exception as e: logger.error( - f"[MemFeedback process] Error in concurrent tasks for {user_name}: {e}", + f"[Feedback Core MemFeedback process] Error in concurrent tasks for {user_name}: {e}", exc_info=True, ) return {"answer": "", "record": {"add": [], "update": []}} - - # Helper for DB operations with retry - @retry(stop=stop_after_attempt(3), wait=wait_exponential(multiplier=1, min=4, max=10)) - def _retry_db_operation(self, operation): - try: - return operation() - except Exception as e: - logger.error( - f"[MemFeedback: _retry_db_operation] DB operation failed: {e}", exc_info=True - ) - raise - - @require_python_package( - import_name="jieba", - install_command="pip install jieba", - install_link="https://github.com/fxsjy/jieba", - ) - def _tokenize_chinese(self, text): - """split zh jieba""" - import jieba - - tokens = jieba.lcut(text) - tokens = [token.strip() for token in tokens if token.strip()] - return self.stopword_manager.filter_words(tokens) diff --git a/src/memos/multi_mem_cube/single_cube.py b/src/memos/multi_mem_cube/single_cube.py index bc50faab0..a36f4ff3a 100644 --- a/src/memos/multi_mem_cube/single_cube.py +++ b/src/memos/multi_mem_cube/single_cube.py @@ -185,7 +185,7 @@ def feedback_memories(self, feedback_req: APIFeedbackRequest) -> dict[str, Any]: task_id=feedback_req.task_id, info=feedback_req.info, ) - self.logger.info(f"Feedback memories result: {feedback_result}") + self.logger.info(f"[Feedback memories result:] {feedback_result}") return feedback_result def _get_search_mode(self, mode: str) -> str: diff --git a/src/memos/templates/mem_feedback_prompts.py b/src/memos/templates/mem_feedback_prompts.py index cd0c46a61..bbdb187e2 100644 --- a/src/memos/templates/mem_feedback_prompts.py +++ b/src/memos/templates/mem_feedback_prompts.py @@ -441,6 +441,8 @@ ] }} +**Current time** +{now_time} **Current Memories** {current_memories} @@ -581,6 +583,9 @@ ] }} +**当前时间：** +{now_time} + **当前记忆：** {current_memories} From 7f146e1cc7f51d2c2b9e5149b4dcf4b5d820cd37 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Fri, 12 Dec 2025 17:12:46 +0800 Subject: [PATCH 42/72] add log and make embedding safety net --- src/memos/graph_dbs/polardb.py | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-) diff --git a/src/memos/graph_dbs/polardb.py b/src/memos/graph_dbs/polardb.py index b1fc74528..c3f0297b7 100644 --- a/src/memos/graph_dbs/polardb.py +++ b/src/memos/graph_dbs/polardb.py @@ -1740,9 +1740,12 @@ def seach_by_keywords_like( """ params = (query_word,) - logger.info(f"[seach_by_keywords_LIKE start:] user_name: {user_name}, params: {params}") - conn = self._get_connection() + logger.info( + f"[seach_by_keywords_LIKE start:] user_name: {user_name}, query: {query}, params: {params}" + ) + conn = None try: + conn = self._get_connection() with conn.cursor() as cursor: cursor.execute(query, params) results = cursor.fetchall() @@ -1752,7 +1755,7 @@ def seach_by_keywords_like( id_val = str(oldid) output.append({"id": id_val}) logger.info( - f"[seach_by_keywords_LIKE end:] user_name: {user_name}, params: {params} recalled: {output}" + f"[seach_by_keywords_LIKE end:] user_name: {user_name}, query: {query}, params: {params} recalled: {output}" ) return output finally: @@ -1837,8 +1840,9 @@ def seach_by_keywords_tfidf( logger.info( f"[seach_by_keywords_TFIDF start:] user_name: {user_name}, query: {query}, params: {params}" ) - conn = self._get_connection() + conn = None try: + conn = self._get_connection() with conn.cursor() as cursor: cursor.execute(query, params) results = cursor.fetchall() @@ -1848,9 +1852,6 @@ def seach_by_keywords_tfidf( id_val = str(oldid) output.append({"id": id_val}) - logger.info( - f"[seach_by_keywords_TFIDF end:] user_name: {user_name}, query: {query}, params: {params} recalled: {output}" - ) logger.info( f"[seach_by_keywords_TFIDF end:] user_name: {user_name}, query: {query}, params: {params} recalled: {output}" ) From 4da6d315a8d94b65fba945c942957600a1e0a307 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 16 Dec 2025 11:47:58 +0800 Subject: [PATCH 43/72] deduplicate add objects --- src/memos/mem_feedback/feedback.py | 14 +++++++++++++- 1 file changed, 13 insertions(+), 1 deletion(-) diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index fe46fbe62..13b4fb036 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -618,7 +618,19 @@ def correct_item(data): return None dehallu_res = [correct_item(item) for item in operations] - llm_operations = [item for item in dehallu_res if item] + dehalluded_operations = [item for item in dehallu_res if item] + + # deduplicate add objects + add_texts = [] + llm_operations = [] + for item in dehalluded_operations: + if item["operation"].lower() == "add" and "text" in item and item["text"]: + if item["text"] in add_texts: + continue + llm_operations.append(item) + add_texts.append(item["text"]) + elif item["operation"].lower() == "update": + llm_operations.append(item) # Update takes precedence over add has_update = any(item.get("operation").lower() == "update" for item in llm_operations) From fd2816c1a12f5a94855ca8dfa43a9fbe6afc0885 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 17 Dec 2025 11:20:56 +0800 Subject: [PATCH 44/72] use _add_memories_parallel --- src/memos/mem_feedback/feedback.py | 6 +++--- .../memories/textual/tree_text_memory/organize/manager.py | 1 + 2 files changed, 4 insertions(+), 3 deletions(-) diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index 13b4fb036..b0927fa0f 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -133,7 +133,7 @@ def _pure_add(self, user_name: str, feedback_content: str, feedback_time: str, i memories = self.mem_reader.get_memory(scene_data, type="chat", info=info) to_add_memories = [item for scene in memories for item in scene] added_ids = self._retry_db_operation( - lambda: self.memory_manager.add(to_add_memories, user_name=user_name) + lambda: self.memory_manager.add(to_add_memories, user_name=user_name, use_batch=False) ) logger.info( f"[Feedback Core: _pure_add] Pure added {len(added_ids)} memories for user {user_name}." @@ -224,10 +224,10 @@ def _single_add_operation( to_add_memory.id = "" added_ids = self._retry_db_operation( - lambda: self.memory_manager.add([to_add_memory], user_name=user_name, mode=async_mode) + lambda: self.memory_manager.add([to_add_memory], user_name=user_name, use_batch=False) ) - logger.info(f"[Memory Feedback ADD] memory id: {added_ids[0]}") + logger.info(f"[Memory Feedback ADD] memory id: {added_ids!s}") return {"id": added_ids[0], "text": to_add_memory.memory} def _single_update_operation( diff --git a/src/memos/memories/textual/tree_text_memory/organize/manager.py b/src/memos/memories/textual/tree_text_memory/organize/manager.py index c8c3cb01c..95f4e780d 100644 --- a/src/memos/memories/textual/tree_text_memory/organize/manager.py +++ b/src/memos/memories/textual/tree_text_memory/organize/manager.py @@ -131,6 +131,7 @@ def _add_memories_parallel( added_ids.extend(ids) except Exception as e: logger.exception("Memory processing error: ", exc_info=e) + logger.info(f"[MemoryManager: _add_memories_parallel] Added {len(added_ids)} memories") return added_ids def _add_memories_batch( From 39b0b2002563091a1449d1356c3a8f09a71ef6f4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 17 Dec 2025 14:21:38 +0800 Subject: [PATCH 45/72] delete Special characters --- src/memos/mem_feedback/feedback.py | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index b0927fa0f..26aa93a3f 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -1,6 +1,7 @@ import concurrent.futures import difflib import json +import re from datetime import datetime from typing import TYPE_CHECKING, Any @@ -493,7 +494,7 @@ def _info_comparison(self, memory: TextualMemoryItem, _info: dict, include_keys: record = [] for key in include_keys: info_v = _info.get(key) - mem_v = memory.metadata.info.get(key, None) + mem_v = memory.metadata.info.get(key, None) if memory.metadata.info else None record.append(info_v == mem_v) return all(record) @@ -554,7 +555,8 @@ def _get_llm_response(self, prompt: str, dsl: bool = True) -> dict: response_text = self.llm.generate(messages, temperature=0.3, timeout=60) if dsl: response_text = response_text.replace("```", "").replace("json", "") - response_json = json.loads(response_text) + cleaned_text = re.sub(r"[\x00-\x08\x0B\x0C\x0E-\x1F\x7F]", "", response_text) + response_json = json.loads(cleaned_text) else: return response_text except Exception as e: @@ -631,7 +633,6 @@ def correct_item(data): add_texts.append(item["text"]) elif item["operation"].lower() == "update": llm_operations.append(item) - # Update takes precedence over add has_update = any(item.get("operation").lower() == "update" for item in llm_operations) if has_update: From b7b50037bf1dfa9b40d59594f15ffbeb126a7c31 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 17 Dec 2025 14:33:50 +0800 Subject: [PATCH 46/72] delete Special characters --- src/memos/mem_feedback/feedback.py | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index 26aa93a3f..a5ab28a89 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -622,7 +622,7 @@ def correct_item(data): dehallu_res = [correct_item(item) for item in operations] dehalluded_operations = [item for item in dehallu_res if item] - # deduplicate add objects + # c add objects add_texts = [] llm_operations = [] for item in dehalluded_operations: @@ -633,6 +633,10 @@ def correct_item(data): add_texts.append(item["text"]) elif item["operation"].lower() == "update": llm_operations.append(item) + logger.info( + f"[Feedback Core: deduplicate add] {len(dehalluded_operations)} -> {len(llm_operations)} memories" + ) + # Update takes precedence over add has_update = any(item.get("operation").lower() == "update" for item in llm_operations) if has_update: From 95bb061ab2412f9099748b08ce41eca28e36e8ee Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 17 Dec 2025 15:01:54 +0800 Subject: [PATCH 47/72] delete Special characters --- src/memos/mem_feedback/feedback.py | 20 ++++++++++++++++++-- 1 file changed, 18 insertions(+), 2 deletions(-) diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index a5ab28a89..d62eab91a 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -1,4 +1,5 @@ import concurrent.futures +import copy import difflib import json import re @@ -331,7 +332,7 @@ def semantics_feedback( current_memories = [ item for item in current_memories if self._info_comparison(item, info, include_keys) ] - + operations = [] if not current_memories: operations = [{"operation": "ADD"}] logger.warning( @@ -371,6 +372,21 @@ def semantics_feedback( operations = self.standard_operations(all_operations, current_memories) + add_texts = [] + final_operations = [] + for item in operations: + if item["operation"].lower() == "add" and "text" in item and item["text"]: + if item["text"] in add_texts: + continue + final_operations.append(item) + add_texts.append(item["text"]) + elif item["operation"].lower() == "update": + final_operations.append(item) + logger.info( + f"[Feedback Core: deduplicate add] {len(operations)} -> {len(final_operations)} memories" + ) + operations = copy.deepcopy(final_operations) + logger.info(f"[Feedback Core Operations]: {operations!s}") if not operations: @@ -622,7 +638,7 @@ def correct_item(data): dehallu_res = [correct_item(item) for item in operations] dehalluded_operations = [item for item in dehallu_res if item] - # c add objects + # deduplicate add objects add_texts = [] llm_operations = [] for item in dehalluded_operations: From a2fe6ed57933b24d109ae89617a4ae180b129c4c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 17 Dec 2025 15:13:38 +0800 Subject: [PATCH 48/72] delete Special characters --- src/memos/mem_feedback/feedback.py | 1 + 1 file changed, 1 insertion(+) diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index 5f99325a0..8df04333c 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -637,6 +637,7 @@ def correct_item(data): dehallu_res = [correct_item(item) for item in operations] dehalluded_operations = [item for item in dehallu_res if item] + logger.info(f"[Feedback Core: dehalluded_operations] {dehalluded_operations}") # c add objects add_texts = [] From 62748641f8f6e383d1eb1cb2f137c6d8fe673e10 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 17 Dec 2025 18:02:08 +0800 Subject: [PATCH 49/72] add source_doc_id --- src/memos/mem_feedback/feedback.py | 32 ++++++++++++++++++-- src/memos/mem_scheduler/general_scheduler.py | 5 +-- 2 files changed, 33 insertions(+), 4 deletions(-) diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index 8df04333c..d6e0ce3c4 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -143,7 +143,17 @@ def _pure_add(self, user_name: str, feedback_content: str, feedback_time: str, i return { "record": { "add": [ - {"id": _id, "text": added_mem.memory} + { + "id": _id, + "text": added_mem.memory, + "source_doc_id": ( + added_mem.metadata.file_ids[0] + if hasattr(added_mem.metadata, "file_ids") + and isinstance(added_mem.metadata.file_ids, list) + and added_mem.metadata.file_ids + else None + ), + } for _id, added_mem in zip(added_ids, to_add_memories, strict=False) ], "update": [], @@ -230,7 +240,17 @@ def _single_add_operation( ) logger.info(f"[Memory Feedback ADD] memory id: {added_ids!s}") - return {"id": added_ids[0], "text": to_add_memory.memory} + return { + "id": added_ids[0], + "text": to_add_memory.memory, + "source_doc_id": ( + to_add_memory.metadata.file_ids[0] + if hasattr(to_add_memory.metadata, "file_ids") + and isinstance(to_add_memory.metadata.file_ids, list) + and to_add_memory.metadata.file_ids + else None + ), + } def _single_update_operation( self, @@ -244,6 +264,13 @@ def _single_update_operation( Individual update operations """ memory_type = old_memory_item.metadata.memory_type + source_doc_id = ( + old_memory_item.metadata.file_ids[0] + if hasattr(old_memory_item.metadata, "file_ids") + and isinstance(old_memory_item.metadata.file_ids, list) + and old_memory_item.metadata.file_ids + else None + ) if memory_type == "WorkingMemory": fields = { "memory": new_memory_item.memory, @@ -274,6 +301,7 @@ def _single_update_operation( return { "id": item_id, "text": new_memory_item.memory, + "source_doc_id": source_doc_id, "archived_id": old_memory_item.id, "origin_memory": old_memory_item.memory, } diff --git a/src/memos/mem_scheduler/general_scheduler.py b/src/memos/mem_scheduler/general_scheduler.py index bd7fb202d..6256467ba 100644 --- a/src/memos/mem_scheduler/general_scheduler.py +++ b/src/memos/mem_scheduler/general_scheduler.py @@ -644,8 +644,8 @@ def _extract_fields(mem_item): or mem_item.get("original_content") ) source_doc_id = None - if "archived_id" in mem_item: - source_doc_id = mem_item.get("archived_id") + if isinstance(mem_item, dict): + source_doc_id = mem_item.get("source_doc_id", None) return mem_id, mem_memory, original_content, source_doc_id @@ -699,6 +699,7 @@ def _extract_fields(mem_item): stack_info=True, ) + logger.info(f"[Feedback Scheduler] kb_log_content: {kb_log_content!s}") if kb_log_content: logger.info( "[DIAGNOSTIC] general_scheduler._mem_feedback_message_consumer: Creating knowledgeBaseUpdate event for feedback. user_id=%s mem_cube_id=%s task_id=%s items=%s", From f2aec3849fb68bb7172203be3c62cc750ceaacdc Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 17 Dec 2025 19:24:13 +0800 Subject: [PATCH 50/72] add source_doc_id --- src/memos/mem_feedback/feedback.py | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index d6e0ce3c4..e0fd6cc77 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -259,6 +259,7 @@ def _single_update_operation( user_id: str, user_name: str, async_mode: str = "sync", + operation: dict | None = None, ) -> dict: """ Individual update operations @@ -271,6 +272,9 @@ def _single_update_operation( and old_memory_item.metadata.file_ids else None ) + if operation and "text" in operation and operation["text"]: + new_memory_item.memory = operation["text"] + if memory_type == "WorkingMemory": fields = { "memory": new_memory_item.memory, @@ -445,6 +449,7 @@ def semantics_feedback( memory_item, user_id, user_name, + operation=op, ) future_to_op[future] = ("update", op) From d34812b4f84852e656f5563a6d681cf45490167d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Thu, 18 Dec 2025 11:56:10 +0800 Subject: [PATCH 51/72] add reranker in init com.. --- .../general_modules/init_components_for_scheduler.py | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/src/memos/mem_scheduler/general_modules/init_components_for_scheduler.py b/src/memos/mem_scheduler/general_modules/init_components_for_scheduler.py index 8da6a2890..5c9d556a3 100644 --- a/src/memos/mem_scheduler/general_modules/init_components_for_scheduler.py +++ b/src/memos/mem_scheduler/general_modules/init_components_for_scheduler.py @@ -39,6 +39,7 @@ if TYPE_CHECKING: from memos.memories.textual.tree_text_memory.retrieve.searcher import Searcher +from memos.api.handlers.config_builders import build_feedback_reranker_config from memos.reranker.factory import RerankerFactory from memos.vec_dbs.factory import VecDBFactory @@ -277,6 +278,7 @@ def init_components() -> dict[str, Any]: embedder_config = build_embedder_config() mem_reader_config = build_mem_reader_config() reranker_config = build_reranker_config() + feedback_reranker_config = build_feedback_reranker_config() internet_retriever_config = build_internet_retriever_config() vector_db_config = build_vec_db_config() pref_extractor_config = build_pref_extractor_config() @@ -296,6 +298,7 @@ def init_components() -> dict[str, Any]: embedder = EmbedderFactory.from_config(embedder_config) mem_reader = MemReaderFactory.from_config(mem_reader_config) reranker = RerankerFactory.from_config(reranker_config) + feedback_reranker = RerankerFactory.from_config(feedback_reranker_config) internet_retriever = InternetRetrieverFactory.from_config( internet_retriever_config, embedder=embedder ) @@ -359,7 +362,7 @@ def init_components() -> dict[str, Any]: config_factory=pref_retriever_config, llm_provider=llm, embedder=embedder, - reranker=reranker, + reranker=feedback_reranker, vector_db=vector_db, ) if os.getenv("ENABLE_PREFERENCE_MEMORY", "false") == "true" @@ -374,7 +377,7 @@ def init_components() -> dict[str, Any]: extractor_llm=llm, vector_db=vector_db, embedder=embedder, - reranker=reranker, + reranker=feedback_reranker, extractor=pref_extractor, adder=pref_adder, retriever=pref_retriever, @@ -405,6 +408,7 @@ def init_components() -> dict[str, Any]: memory_manager=memory_manager, mem_reader=mem_reader, searcher=searcher, + reranker=feedback_reranker, ) # Return all components as a dictionary for easy access and extension return {"naive_mem_cube": naive_mem_cube, "feedback_server": feedback_server} From 3919dcf96a918b5360867ae4bd767010b5688ad9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Thu, 18 Dec 2025 14:07:24 +0800 Subject: [PATCH 52/72] fix circle import --- .../general_modules/init_components_for_scheduler.py | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/src/memos/mem_scheduler/general_modules/init_components_for_scheduler.py b/src/memos/mem_scheduler/general_modules/init_components_for_scheduler.py index 5c9d556a3..ba7b558fd 100644 --- a/src/memos/mem_scheduler/general_modules/init_components_for_scheduler.py +++ b/src/memos/mem_scheduler/general_modules/init_components_for_scheduler.py @@ -39,7 +39,6 @@ if TYPE_CHECKING: from memos.memories.textual.tree_text_memory.retrieve.searcher import Searcher -from memos.api.handlers.config_builders import build_feedback_reranker_config from memos.reranker.factory import RerankerFactory from memos.vec_dbs.factory import VecDBFactory @@ -161,6 +160,16 @@ def build_reranker_config() -> dict[str, Any]: return RerankerConfigFactory.model_validate(APIConfig.get_reranker_config()) +def build_feedback_reranker_config() -> dict[str, Any]: + """ + Build reranker configuration. + + Returns: + Validated reranker configuration dictionary + """ + return RerankerConfigFactory.model_validate(APIConfig.get_feedback_reranker_config()) + + def build_internet_retriever_config() -> dict[str, Any]: """ Build internet retriever configuration. From 73106ed125b8ecc7d431c48235eae1e695b3f7b3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 23 Dec 2025 16:51:53 +0800 Subject: [PATCH 53/72] add feedback judgement --- src/memos/mem_feedback/feedback.py | 77 ++++++-- src/memos/mem_feedback/utils.py | 32 ++++ src/memos/templates/mem_feedback_prompts.py | 200 ++++++++++++++++++-- 3 files changed, 270 insertions(+), 39 deletions(-) diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index e0fd6cc77..35109dbae 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -1,5 +1,4 @@ import concurrent.futures -import copy import difflib import json import re @@ -17,7 +16,12 @@ from memos.llms.factory import AzureLLM, LLMFactory, OllamaLLM, OpenAILLM from memos.log import get_logger from memos.mem_feedback.base import BaseMemFeedback -from memos.mem_feedback.utils import make_mem_item, should_keep_update, split_into_chunks +from memos.mem_feedback.utils import ( + general_split_into_chunks, + make_mem_item, + should_keep_update, + split_into_chunks, +) from memos.mem_reader.factory import MemReaderFactory from memos.mem_reader.read_multi_modal import detect_lang from memos.memories.textual.item import TextualMemoryItem @@ -37,6 +41,8 @@ FEEDBACK_JUDGEMENT_PROMPT_ZH, KEYWORDS_REPLACE, KEYWORDS_REPLACE_ZH, + OPERATION_UPDATE_JUDGEMENT, + OPERATION_UPDATE_JUDGEMENT_ZH, UPDATE_FORMER_MEMORIES, UPDATE_FORMER_MEMORIES_ZH, ) @@ -47,6 +53,7 @@ "if_kw_replace": {"en": KEYWORDS_REPLACE, "zh": KEYWORDS_REPLACE_ZH}, "judge": {"en": FEEDBACK_JUDGEMENT_PROMPT, "zh": FEEDBACK_JUDGEMENT_PROMPT_ZH}, "compare": {"en": UPDATE_FORMER_MEMORIES, "zh": UPDATE_FORMER_MEMORIES_ZH}, + "compare_judge": {"en": OPERATION_UPDATE_JUDGEMENT, "zh": OPERATION_UPDATE_JUDGEMENT_ZH}, "generation": {"en": FEEDBACK_ANSWER_PROMPT, "zh": FEEDBACK_ANSWER_PROMPT_ZH}, } @@ -402,22 +409,8 @@ def semantics_feedback( except Exception as e: logger.error(f"[Feedback Core: semantics_feedback] Operation failed: {e}") - operations = self.standard_operations(all_operations, current_memories) - - add_texts = [] - final_operations = [] - for item in operations: - if item["operation"].lower() == "add" and "text" in item and item["text"]: - if item["text"] in add_texts: - continue - final_operations.append(item) - add_texts.append(item["text"]) - elif item["operation"].lower() == "update": - final_operations.append(item) - logger.info( - f"[Feedback Core: deduplicate add] {len(operations)} -> {len(final_operations)} memories" - ) - operations = copy.deepcopy(final_operations) + standard_operations = self.standard_operations(all_operations, current_memories) + operations = self.filter_fault_update(standard_operations) logger.info(f"[Feedback Core Operations]: {operations!s}") @@ -552,7 +545,7 @@ def _retrieve(self, query: str, info=None, top_k=100, user_name=None): retrieved_mems = self.searcher.search( query, info=info, user_name=user_name, top_k=top_k, full_recall=True ) - retrieved_mems = [item[0] for item in retrieved_mems] + retrieved_mems = [item[0] for item in retrieved_mems if float(item[1]) > 0.01] return retrieved_mems def _vec_query(self, new_memories_embedding: list[float], user_name=None): @@ -615,6 +608,52 @@ def _get_llm_response(self, prompt: str, dsl: bool = True) -> dict: response_json = None return response_json + def filter_fault_update(self, operations: list[dict]): + """To address the randomness of large model outputs, it is necessary to conduct validity evaluation on the texts used for memory override operations.""" + updated_operations = [item for item in operations if item["operation"] == "UPDATE"] + if len(updated_operations) < 5: + return operations + + lang = detect_lang("".join(updated_operations[0]["text"])) + template = FEEDBACK_PROMPT_DICT["compare_judge"][lang] + + all_judge = [] + operations_chunks = general_split_into_chunks(updated_operations) + with ContextThreadPoolExecutor(max_workers=10) as executor: + future_to_chunk_idx = {} + for chunk in operations_chunks: + raw_operations_str = {"operations": chunk} + prompt = template.format(raw_operations=str(raw_operations_str)) + + future = executor.submit(self._get_llm_response, prompt) + future_to_chunk_idx[future] = chunk + for future in concurrent.futures.as_completed(future_to_chunk_idx): + try: + judge_res = future.result() + if ( + judge_res + and "operations_judgement" in judge_res + and isinstance(judge_res["operations_judgement"], list) + ): + all_judge.extend(judge_res["operations_judgement"]) + except Exception as e: + logger.error(f"[Feedback Core: filter_fault_update] Judgement failed: {e}") + + logger.info(f"[Feedback Core: filter_fault_update] LLM judgement: {all_judge}") + id2op = {item["id"]: item for item in updated_operations} + valid_updates = [] + for judge in all_judge: + valid_update = None + if judge["judgement"] == "UPDATE_APPROVED": + valid_update = id2op.get(judge["id"], None) + if valid_update: + valid_updates.append(valid_update) + + logger.info( + f"[Feedback Core: filter_fault_update] {len(updated_operations)} -> {len(valid_updates)}" + ) + return valid_updates + [item for item in operations if item["operation"] != "UPDATE"] + def standard_operations(self, operations, current_memories): """ Regularize the operation design diff --git a/src/memos/mem_feedback/utils.py b/src/memos/mem_feedback/utils.py index 0033d85b4..c32c12328 100644 --- a/src/memos/mem_feedback/utils.py +++ b/src/memos/mem_feedback/utils.py @@ -54,6 +54,38 @@ def calculate_similarity(text1: str, text2: str) -> float: return change_ratio < 0.2 +def general_split_into_chunks(items: list[dict], max_tokens_per_chunk: int = 500): + chunks = [] + current_chunk = [] + current_tokens = 0 + + for item in items: + item_text = str(item) + item_tokens = estimate_tokens(item_text) + + if item_tokens > max_tokens_per_chunk: + if current_chunk: + chunks.append(current_chunk) + current_chunk = [] + + chunks.append([item]) + current_tokens = 0 + + elif current_tokens + item_tokens <= max_tokens_per_chunk: + current_chunk.append(item) + current_tokens += item_tokens + else: + if current_chunk: + chunks.append(current_chunk) + current_chunk = [item] + current_tokens = item_tokens + + if current_chunk: + chunks.append(current_chunk) + + return chunks + + def split_into_chunks(memories: list[TextualMemoryItem], max_tokens_per_chunk: int = 500): chunks = [] current_chunk = [] diff --git a/src/memos/templates/mem_feedback_prompts.py b/src/memos/templates/mem_feedback_prompts.py index bbdb187e2..dd30c4f92 100644 --- a/src/memos/templates/mem_feedback_prompts.py +++ b/src/memos/templates/mem_feedback_prompts.py @@ -334,10 +334,11 @@ }} *Requirements*: -1. If the new fact does not provide additional information to the existing memory item, the existing memory can override the new fact, and the operation is set to "NONE." -2. If the new fact is similar to existing memory but the information is more accurate, complete, or requires correction, set operation to "UPDATE" +1. If the new fact does not provide additional information to the existing memory item, or the existing memory can override the new fact, and the operation is set to "NONE." +2. If the new fact is similar to existing memory **about the same entity** but the information is more accurate, complete, or requires correction, set operation to "UPDATE" 3. If the new fact contradicts existing memory in key information (such as time, location, status, etc.), update the original memory based on the new fact and set operation to "UPDATE", only modifying the relevant error segments in the existing memory paragraphs while keeping other text completely unchanged. -4. If there is no existing memory that requires updating, the new fact is added as entirely new information, and the operation is set to "ADD." Therefore, in the same operation list, ADD and UPDATE will not coexist. +4. If there is no existing memory that requires updating **or if the new fact refers to a different entity**, the new fact is added as entirely new information, and the operation is set to "ADD." Therefore, in the same operation list, ADD and UPDATE will not coexist. +5. Facts about different entities that were acknowledged by the user within the same time period can coexist and are not considered contradictory. *ID Management Rules*: - Update operation: Keep the original ID unchanged @@ -408,16 +409,16 @@ Example2: Current Memories: -"123": "The user works as a software engineer in Company A, mainly responsible for front-end development" -"908": "The user likes to go fishing with friends on weekends" +"123": "On December 22, 2025, the user claim that John works at Company X" +"908": "On December 22, 2025, the user claim that Mary lives in New York" The background of the new fact being put forward: -user: Guess where I live？ -assistant: Hehuan Community. -user feedback: Wrong, update my address: Mingyue Community, Chaoyang District, Beijing +user: Guess who am I？ +assistant: You are a teacher at School ABC. +user feedback: No, I mean Peter is a teacher at School ABC. Newly facts: -"The user's residential address is Mingyue Community, Chaoyang District, Beijing" +"Peter is a teacher at School ABC." Operation recommendations: {{ @@ -425,17 +426,17 @@ [ {{ "id": "123", - "text": "The user works as a software engineer at Company A, primarily responsible for front-end development", + "text": "On December 22, 2025, the user claim that John works at Company X", "operation": "NONE" }}, {{ "id": "908", - "text": "The user enjoys fishing with friends on weekends", + "text": "On December 22, 2025, the user claim that Mary lives in New York", "operation": "NONE" }}, {{ - "id": "4567", - "text": "The user's residential address is Mingyue Community, Chaoyang District, Beijing", + "id": "001", + "text": "Peter is a teacher at School ABC.", "operation": "ADD" }} ] @@ -478,6 +479,7 @@ 2. 若新事实与现有记忆相似但信息更准确、完整或需修正，操作设为"UPDATE" 3. 若新事实在关键信息（如时间、地点、状态等）上与现有记忆矛盾，则根据新事实更新原记忆，操作设为"UPDATE"，仅修改现有记忆段落中的相关错误片段，其余文本完全保持不变 4. 若无需要更新的现有记忆，则将新事实作为全新信息添加，操作设为"ADD"。因此在同一操作列表中，ADD与UPDATE不会同时存在 +5. 同一时间段内用户所确认的不同实体的相关事实可以并存，且不会被视作相互矛盾。 ID管理规则： - 更新操作：保持原有ID不变 @@ -549,17 +551,16 @@ 示例2： 当前记忆： -"123": "用户在公司A担任软件工程师，主要负责前端开发" -"908": "用户周末喜欢和朋友一起钓鱼" - +"123": "2025年12月12日，用户声明约翰在 X 公司工作" +"908": "2025年12月12日，用户声明玛丽住在纽约" 提出新事实的背景： -user: 猜猜我住在哪里？ +user: 猜猜刘青住在哪里？ assistant: 合欢社区 -user feedback: 错了，请更新我的地址：北京市朝阳区明月社区 +user feedback: 错了，他住在明月小区 新获取的事实： -"用户的居住地址是北京市朝阳区明月小区" +"用户声明刘青住在明月小区" 操作建议： {{ @@ -577,7 +578,7 @@ }}, {{ "id": "4567", - "text": "用户的居住地址是北京市朝阳区明月小区", + "text": "用户声明刘青住在明月小区", "operation": "ADD" }} ] @@ -660,3 +661,162 @@ 回答： """ + + +OPERATION_UPDATE_JUDGEMENT = """ +# Batch UPDATE Safety Assessment Instruction + +**Background**: +This instruction serves as a supplementary safety verification layer for the memory update instruction. It evaluates each UPDATE operation in the `operations` list to ensure safety and effectiveness, preventing erroneous data overwrites. + +**Input**: The `operations` list containing multiple UPDATE proposals generated by the main instruction +**Output**: The final `operations_judgement` list after safety assessment and necessary corrections + +**Safety Assessment Process (for each UPDATE entry)**: +1. **Entity Consistency Check**: Verify that the old and new texts of this UPDATE entry describe exactly the same core entity (same person, organization, event, etc.). This is the most important check. +2. **Semantic Relevance Check**: Determine whether the new information directly corrects errors in or supplements missing information from the old information, rather than introducing completely unrelated new facts. +3. **Context Preservation Check**: Ensure that the updated text of this UPDATE only modifies the parts that need correction, while completely preserving all other valid information from the original text. + +**Batch Assessment Rules**: +- Independently assess each entry in the list and record the evaluation results + +**Key Decision Rules**: +1. If the core entities of old and new texts are different → Set `judgement` to "INVALID" (completely invalid) +2. If the core entities are the same but the information is completely unrelated → Set `judgement` to "NONE" (should not update) +3. If all three checks pass → Set `judgement` to "UPDATE_APPROVED" + +**Output Format**: +{{ + "operations_judgement": [ + {{ + "id": "...", + "text": "...", + "old_memory": "...", + "judgement": "INVALID" | "NONE" | "UPDATE_APPROVED" + }}, + ... + ] +}} + +**Example 1**: +Input operations list: +{{ + "operations": [ + {{ + "id": "275a", + "text": "On December 22, 2025 at 6:58 AM UTC, the user mentioned that Mission Terra is from Germany.", + "operation": "UPDATE", + "old_memory": "On December 13, 2025 at 4:02 PM UTC, the user mentioned that Mission Terra is a French national." + }}, + {{ + "id": "88a4", + "text": "On December 22, 2025 at 6:58 AM UTC, the user mentioned that Mission Terra is from Germany.", + "operation": "UPDATE", + "old_memory": "On December 22, 2025 at 6:52 AM UTC, the user confirmed that Gladys Liu is an Italian citizen." + }} + ] +}} + +Safety assessment output: +{{ + "operations_judgement": [ + {{ + "id": "275a", + "text": "On December 22, 2025 at 6:58 AM UTC, the user mentioned that Mission Terra is from Germany.", + "old_memory": "On December 13, 2025 at 4:02 PM UTC, the user mentioned that Mission Terra is a French national.", + "judgement": "UPDATE_APPROVED" + }}, + {{ + "id": "88a4", + "text": "On December 22, 2025 at 6:58 AM UTC, the user mentioned that Mission Terra is from Germany.", + "old_memory": "On December 22, 2025 at 6:52 AM UTC, the user confirmed that Gladys Liu is an Italian citizen.", + "judgement": "INVALID" + }} + ] +}} + +**For actual execution**: +Input operations list: +{raw_operations} + +Safety assessment output:""" + + +OPERATION_UPDATE_JUDGEMENT_ZH = """## 批量UPDATE安全评估指令 + +**背景说明**： +本指令作为记忆更新指令的补充安全验证层。针对`operations`列表，评估每个UPDATE操作都安全有效，防止错误的数据覆盖。 + +**输入**：主指令生成的包含多个UPDATE提议的`operations`列表 +**输出**：经过安全评估和必要修正后的最终`operations_judgement`列表 + +**安全评估流程（针对每个UPDATE条目）**： +1. **实体一致性检查**：确认该UPDATE条目的新旧文本是否描述完全相同的核心实体（同一人物、组织、事件等）。这是最重要的检查。 +2. **语义相关性检查**：判断该UPDATE的新信息是否直接修正旧信息中的错误部分或补充缺失信息，而非引入完全不相关的新事实。 +3. **上下文保留检查**：确保该UPDATE更新后的文本只修改需要纠正的部分，完全保留原始文本中其他所有有效信息。 + +**批量评估规则**： +- 对列表中的每个条目独立评估，记录评估结果 + +**关键决策规则**： +1. 如果新旧文本核心实体不同 → `judgement`置为"INVALID"（完全无效） +2. 如果新旧文本核心实体相同但信息完全不相关 → `judgement`置为"NONE"（不应更新） +3. 如果通过全部三项检查 → `judgement`置为"UPDATE_APPROVED" + + +**输出格式**： +{{ + "operations_judgement": [ + // 评估后的完整operations列表 + {{ + "id": "...", + "text": "...", + "old_memory": "...", + "judgement": "INVALID" | "NONE" | "UPDATE_APPROVED" + }}, + ... + ] +}} + + +示例1： +输入operations列表： +{{ + "operations": [ + {{ + "id": "275a", + "text": "2025年12月22日 UTC 时间6:58，用户提到Mission Terra 来自德国。", + "operation": "UPDATE", + "old_memory": "2025年12月13日 UTC 时间16:02，用户提及 Mission Terra 是法国国籍。" + }}, + {{ + "id": "88a4", + "text": "2025年12月22日 UTC 时间6:58，用户提到Mission Terra 来自德国。", + "operation": "UPDATE", + "old_memory": "2025年12月22日 UTC 时间6:52，用户确认 Gladys Liu 是意大利公民。" + }} + ] +}} +安全评估输出： +{{ + "operations_judgement": [ + {{ + "id": "275a", + "text": "2025年12月22日 UTC 时间6:58，用户提到Mission Terra 来自德国。", + "old_memory": "2025年12月13日 UTC 时间16:02，用户提及 Mission Terra 是法国国籍。", + "judgement": "UPDATE_APPROVED" + }}, + {{ + "id": "88a4", + "text": "2025年12月22日 UTC 时间6:58，用户提到Mission Terra 来自德国。", + "old_memory": "2025年12月22日 UTC 时间6:52，用户确认 Gladys Liu 是意大利公民。", + "judgement": "INVALID" + }} + ] +}} + +输入operations列表： +{raw_operations} + +安全评估输出： +""" From 1b0e3af54cac679405c08e2e3d120028672ba29e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 23 Dec 2025 16:59:01 +0800 Subject: [PATCH 54/72] add feedback judgement --- src/memos/mem_feedback/feedback.py | 58 ++++++++++++++++-------------- 1 file changed, 31 insertions(+), 27 deletions(-) diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index 35109dbae..0b3fc3846 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -115,7 +115,7 @@ def _retry_db_operation(self, operation): return operation() except Exception as e: logger.error( - f"[Feedback Core: _retry_db_operation] DB operation failed: {e}", exc_info=True + f"[1223 Feedback Core: _retry_db_operation] DB operation failed: {e}", exc_info=True ) raise @@ -129,7 +129,7 @@ def _batch_embed(self, texts: list[str], embed_bs: int = 5): results.extend(self._embed_once(batch)) except Exception as e: logger.error( - f"[Feedback Core: process_feedback_core] Embedding batch failed, Cover with all zeros: {len(batch)} entries: {e}" + f"[1223 Feedback Core: process_feedback_core] Embedding batch failed, Cover with all zeros: {len(batch)} entries: {e}" ) results.extend([[0.0] * dim for _ in range(len(batch))]) return results @@ -145,7 +145,7 @@ def _pure_add(self, user_name: str, feedback_content: str, feedback_time: str, i lambda: self.memory_manager.add(to_add_memories, user_name=user_name, use_batch=False) ) logger.info( - f"[Feedback Core: _pure_add] Pure added {len(added_ids)} memories for user {user_name}." + f"[1223 Feedback Core: _pure_add] Pure added {len(added_ids)} memories for user {user_name}." ) return { "record": { @@ -182,7 +182,7 @@ def _keyword_replace_judgement(self, feedback_content: str) -> dict | None: return judge_res else: logger.warning( - "[Feedback Core: _feedback_judgement] feedback judgement failed, return []" + "[1223 Feedback Core: _feedback_judgement] feedback judgement failed, return []" ) return {} @@ -207,7 +207,7 @@ def _feedback_judgement( return judge_res else: logger.warning( - "[Feedback Core: _feedback_judgement] feedback judgement failed, return []" + "[1223 Feedback Core: _feedback_judgement] feedback judgement failed, return []" ) return [] @@ -334,11 +334,11 @@ def _del_working_binding(self, user_name, mem_items: list[TextualMemoryItem]) -> self.graph_store.delete_node(mid, user_name=user_name) logger.info( - f"[Feedback Core:_del_working_binding] Delete raw/working mem_ids: {delete_ids} for user_name: {user_name}" + f"[1223 Feedback Core:_del_working_binding] Delete raw/working mem_ids: {delete_ids} for user_name: {user_name}" ) except Exception as e: logger.warning( - f"[Feedback Core:_del_working_binding] TreeTextMemory.delete_hard: failed to delete {mid}: {e}" + f"[1223 Feedback Core:_del_working_binding] TreeTextMemory.delete_hard: failed to delete {mid}: {e}" ) def semantics_feedback( @@ -407,7 +407,9 @@ def semantics_feedback( ): all_operations.extend(chunk_operations["operations"]) except Exception as e: - logger.error(f"[Feedback Core: semantics_feedback] Operation failed: {e}") + logger.error( + f"[1223 Feedback Core: semantics_feedback] Operation failed: {e}" + ) standard_operations = self.standard_operations(all_operations, current_memories) operations = self.filter_fault_update(standard_operations) @@ -456,7 +458,7 @@ def semantics_feedback( update_results.append(result) except Exception as e: logger.error( - f"[Feedback Core: semantics_feedback] Operation failed for {original_op}: {e}", + f"[1223 Feedback Core: semantics_feedback] Operation failed for {original_op}: {e}", exc_info=True, ) if update_results: @@ -484,7 +486,7 @@ def _feedback_memory( ] if filterd_ids: logger.warning( - f"[Feedback Core: _feedback_memory] Since the tags mode is fast, no modifications are made to the following memory {filterd_ids}." + f"[1223 Feedback Core: _feedback_memory] Since the tags mode is fast, no modifications are made to the following memory {filterd_ids}." ) current_memories = [ @@ -516,7 +518,7 @@ def _feedback_memory( results[i] = node except Exception as e: logger.error( - f"[Feedback Core: _feedback_memory] Error processing memory index {i}: {e}", + f"[1223 Feedback Core: _feedback_memory] Error processing memory index {i}: {e}", exc_info=True, ) mem_res = [r for r in results if r] @@ -575,7 +577,7 @@ def _vec_query(self, new_memories_embedding: list[float], user_name=None): if not retrieved_ids: logger.info( - f"[Feedback Core: _vec_query] No similar memories found for embedding query for user {user_name}." + f"[1223 Feedback Core: _vec_query] No similar memories found for embedding query for user {user_name}." ) filterd_ids = [ @@ -583,7 +585,7 @@ def _vec_query(self, new_memories_embedding: list[float], user_name=None): ] if filterd_ids: logger.warning( - f"[Feedback Core: _vec_query] Since the tags mode is fast, no modifications are made to the following memory {filterd_ids}." + f"[1223 Feedback Core: _vec_query] Since the tags mode is fast, no modifications are made to the following memory {filterd_ids}." ) return [ TextualMemoryItem(**item) @@ -637,9 +639,9 @@ def filter_fault_update(self, operations: list[dict]): ): all_judge.extend(judge_res["operations_judgement"]) except Exception as e: - logger.error(f"[Feedback Core: filter_fault_update] Judgement failed: {e}") + logger.error(f"[1223 Feedback Core: filter_fault_update] Judgement failed: {e}") - logger.info(f"[Feedback Core: filter_fault_update] LLM judgement: {all_judge}") + logger.info(f"[1223 Feedback Core: filter_fault_update] LLM judgement: {all_judge}") id2op = {item["id"]: item for item in updated_operations} valid_updates = [] for judge in all_judge: @@ -650,7 +652,7 @@ def filter_fault_update(self, operations: list[dict]): valid_updates.append(valid_update) logger.info( - f"[Feedback Core: filter_fault_update] {len(updated_operations)} -> {len(valid_updates)}" + f"[1223 Feedback Core: filter_fault_update] {len(updated_operations)} -> {len(valid_updates)}" ) return valid_updates + [item for item in operations if item["operation"] != "UPDATE"] @@ -682,7 +684,7 @@ def correct_item(data): if not should_keep_update(data["text"], data["old_memory"]): logger.warning( - f"[Feedback Core: semantics_feedback] Due to the excessive proportion of changes, skip update: {data}" + f"[1223 Feedback Core: semantics_feedback] Due to the excessive proportion of changes, skip update: {data}" ) return None @@ -702,14 +704,14 @@ def correct_item(data): return data except Exception: logger.error( - f"[Feedback Core: standard_operations] Error processing operation item: {data}", + f"[1223 Feedback Core: standard_operations] Error processing operation item: {data}", exc_info=True, ) return None dehallu_res = [correct_item(item) for item in operations] dehalluded_operations = [item for item in dehallu_res if item] - logger.info(f"[Feedback Core: dehalluded_operations] {dehalluded_operations}") + logger.info(f"[1223 Feedback Core: dehalluded_operations] {dehalluded_operations}") # c add objects add_texts = [] @@ -723,7 +725,7 @@ def correct_item(data): elif item["operation"].lower() == "update": llm_operations.append(item) logger.info( - f"[Feedback Core: deduplicate add] {len(dehalluded_operations)} -> {len(llm_operations)} memories" + f"[1223 Feedback Core: deduplicate add] {len(dehalluded_operations)} -> {len(llm_operations)} memories" ) # Update takes precedence over add @@ -737,7 +739,7 @@ def correct_item(data): ] if filtered_items: logger.info( - f"[Feedback Core: semantics_feedback] Due to have update objects, skip add: {filtered_items}" + f"[1223 Feedback Core: semantics_feedback] Due to have update objects, skip add: {filtered_items}" ) return update_items else: @@ -785,7 +787,7 @@ def _doc_filter(self, doc_scope: str, memories: list[TextualMemoryItem]): memid for inscope_file in inscope_docs for memid in filename2_memid[inscope_file] ] logger.info( - f"[Feedback Core: process_keyword_replace] These docs are in scope : {inscope_docs}, relared memids: {inscope_ids}" + f"[1223 Feedback Core: process_keyword_replace] These docs are in scope : {inscope_docs}, relared memids: {inscope_ids}" ) filter_memories = [mem for mem in memories if mem.id in inscope_ids] return filter_memories @@ -839,7 +841,7 @@ def process_keyword_replace( retrieved_memories = self._doc_filter(doc_scope, retrieved_memories) logger.info( - f"[Feedback Core: process_keyword_replace] Keywords recalled memory for user {user_name}: {len(retrieved_ids)} memories | After filtering: {len(retrieved_memories)} memories." + f"[1223 Feedback Core: process_keyword_replace] Keywords recalled memory for user {user_name}: {len(retrieved_ids)} memories | After filtering: {len(retrieved_memories)} memories." ) if not retrieved_memories: @@ -924,7 +926,7 @@ def check_validity(item): info.update({"user_id": user_id, "user_name": user_name, "session_id": session_id}) logger.info( - f"[Feedback Core: process_feedback_core] Starting memory feedback process for user {user_name}" + f"[1223 Feedback Core: process_feedback_core] Starting memory feedback process for user {user_name}" ) # feedback keywords update kwp_judge = self._keyword_replace_judgement(feedback_content) @@ -957,7 +959,7 @@ def check_validity(item): if not valid_feedback: logger.warning( - f"[Feedback Core: process_feedback_core] No valid judgements for user {user_name}: {raw_judge}." + f"[1223 Feedback Core: process_feedback_core] No valid judgements for user {user_name}: {raw_judge}." ) return {"record": {"add": [], "update": []}} @@ -1005,12 +1007,14 @@ def check_validity(item): add_memories = mem_record["record"]["add"] update_memories = mem_record["record"]["update"] logger.info( - f"[Feedback Core: process_feedback_core] Processed {len(feedback_memories)} feedback | add {len(add_memories)} memories | update {len(update_memories)} memories for user {user_name}." + f"[1223 Feedback Core: process_feedback_core] Processed {len(feedback_memories)} feedback | add {len(add_memories)} memories | update {len(update_memories)} memories for user {user_name}." ) return mem_record except Exception as e: - logger.error(f"[Feedback Core: process_feedback_core] Error for user {user_name}: {e}") + logger.error( + f"[1223 Feedback Core: process_feedback_core] Error for user {user_name}: {e}" + ) return {"record": {"add": [], "update": []}} def process_feedback( From f6d8f77be4509361c23b8d82fe8697ec79b89fb9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 24 Dec 2025 20:33:00 +0800 Subject: [PATCH 55/72] add pref feedback --- src/memos/api/handlers/component_init.py | 1 + src/memos/mem_feedback/feedback.py | 157 +++++++++++++----- src/memos/mem_feedback/simple_feedback.py | 3 + src/memos/mem_feedback/utils.py | 5 +- .../init_components_for_scheduler.py | 1 + src/memos/memories/textual/preference.py | 3 + 6 files changed, 132 insertions(+), 38 deletions(-) diff --git a/src/memos/api/handlers/component_init.py b/src/memos/api/handlers/component_init.py index 9c1212fe0..5109e866e 100644 --- a/src/memos/api/handlers/component_init.py +++ b/src/memos/api/handlers/component_init.py @@ -309,6 +309,7 @@ def init_server() -> dict[str, Any]: mem_reader=mem_reader, searcher=searcher, reranker=feedback_reranker, + pref_mem=pref_mem, ) # Initialize Scheduler diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index 0b3fc3846..fad15a7cd 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -2,6 +2,7 @@ import difflib import json import re +import uuid from datetime import datetime from typing import TYPE_CHECKING, Any @@ -33,6 +34,7 @@ if TYPE_CHECKING: + from memos.memories.textual.simple_preference import SimplePreferenceTextMemory from memos.memories.textual.tree_text_memory.retrieve.searcher import Searcher from memos.templates.mem_feedback_prompts import ( FEEDBACK_ANSWER_PROMPT, @@ -90,6 +92,7 @@ def __init__(self, config: MemFeedbackConfig): self.stopword_manager = StopwordManager self.searcher: Searcher = None self.reranker = None + self.pref_mem: SimplePreferenceTextMemory = None self.DB_IDX_READY = False @require_python_package( @@ -115,7 +118,7 @@ def _retry_db_operation(self, operation): return operation() except Exception as e: logger.error( - f"[1223 Feedback Core: _retry_db_operation] DB operation failed: {e}", exc_info=True + f"[1224 Feedback Core: _retry_db_operation] DB operation failed: {e}", exc_info=True ) raise @@ -129,7 +132,7 @@ def _batch_embed(self, texts: list[str], embed_bs: int = 5): results.extend(self._embed_once(batch)) except Exception as e: logger.error( - f"[1223 Feedback Core: process_feedback_core] Embedding batch failed, Cover with all zeros: {len(batch)} entries: {e}" + f"[1224 Feedback Core: process_feedback_core] Embedding batch failed, Cover with all zeros: {len(batch)} entries: {e}" ) results.extend([[0.0] * dim for _ in range(len(batch))]) return results @@ -145,7 +148,7 @@ def _pure_add(self, user_name: str, feedback_content: str, feedback_time: str, i lambda: self.memory_manager.add(to_add_memories, user_name=user_name, use_batch=False) ) logger.info( - f"[1223 Feedback Core: _pure_add] Pure added {len(added_ids)} memories for user {user_name}." + f"[1224 Feedback Core: _pure_add] Pure added {len(added_ids)} memories for user {user_name}." ) return { "record": { @@ -182,7 +185,7 @@ def _keyword_replace_judgement(self, feedback_content: str) -> dict | None: return judge_res else: logger.warning( - "[1223 Feedback Core: _feedback_judgement] feedback judgement failed, return []" + "[1224 Feedback Core: _feedback_judgement] feedback judgement failed, return []" ) return {} @@ -207,7 +210,7 @@ def _feedback_judgement( return judge_res else: logger.warning( - "[1223 Feedback Core: _feedback_judgement] feedback judgement failed, return []" + "[1224 Feedback Core: _feedback_judgement] feedback judgement failed, return []" ) return [] @@ -271,6 +274,14 @@ def _single_update_operation( """ Individual update operations """ + if "preference" in old_memory_item.metadata.__dict__: + logger.info( + f"[1224 Feedback Core: _single_update_operation] pref_memory: {old_memory_item.id}" + ) + return self._single_update_pref( + old_memory_item, new_memory_item, user_id, user_name, operation + ) + memory_type = old_memory_item.metadata.memory_type source_doc_id = ( old_memory_item.metadata.file_ids[0] @@ -281,6 +292,7 @@ def _single_update_operation( ) if operation and "text" in operation and operation["text"]: new_memory_item.memory = operation["text"] + new_memory_item.metadata.embedding = self._batch_embed([operation["text"]])[0] if memory_type == "WorkingMemory": fields = { @@ -317,6 +329,68 @@ def _single_update_operation( "origin_memory": old_memory_item.memory, } + def _single_update_pref( + self, + old_memory_item: TextualMemoryItem, + new_memory_item: TextualMemoryItem, + user_id: str, + user_name: str, + operation: dict, + ): + """update preference memory""" + + feedback_context = new_memory_item.memory + if operation and "text" in operation and operation["text"]: + new_memory_item.memory = operation["text"] + new_memory_item.metadata.embedding = self._batch_embed([operation["text"]])[0] + + to_add_memory = old_memory_item.model_copy(deep=True) + to_add_memory.metadata.key = new_memory_item.metadata.key + to_add_memory.metadata.tags = new_memory_item.metadata.tags + to_add_memory.memory = new_memory_item.memory + to_add_memory.metadata.preference = new_memory_item.memory + to_add_memory.metadata.embedding = new_memory_item.metadata.embedding + + to_add_memory.metadata.user_id = new_memory_item.metadata.user_id + to_add_memory.metadata.original_text = old_memory_item.memory + to_add_memory.metadata.covered_history = old_memory_item.id + + to_add_memory.metadata.created_at = to_add_memory.metadata.updated_at = ( + datetime.now().isoformat() + ) + to_add_memory.metadata.context_summary = ( + old_memory_item.metadata.context_summary + " \n" + feedback_context + ) + + # add new memory + to_add_memory.id = str(uuid.uuid4()) + added_ids = self._retry_db_operation(lambda: self.pref_mem.add([to_add_memory])) + # delete + deleted_id = old_memory_item.id + collection_name = old_memory_item.metadata.preference_type + self._retry_db_operation( + lambda: self.pref_mem.delete_with_collection_name(collection_name, [deleted_id]) + ) + # add archived + old_memory_item.metadata.status = "archived" + old_memory_item.metadata.original_text = "archived" + old_memory_item.metadata.embedding = [0.0] * 1024 + + archived_ids = self._retry_db_operation(lambda: self.pref_mem.add([old_memory_item])) + + logger.info( + f"[Memory Feedback UPDATE Pref] New Add:{added_ids!s} | Set archived:{archived_ids!s}" + ) + + return { + "id": to_add_memory.id, + "text": new_memory_item.memory, + "source_doc_id": "", + "archived_id": old_memory_item.id, + "origin_memory": old_memory_item.memory, + "type": "preference", + } + def _del_working_binding(self, user_name, mem_items: list[TextualMemoryItem]) -> set[str]: """Delete working memory bindings""" bindings_to_delete = extract_working_binding_ids(mem_items) @@ -334,11 +408,11 @@ def _del_working_binding(self, user_name, mem_items: list[TextualMemoryItem]) -> self.graph_store.delete_node(mid, user_name=user_name) logger.info( - f"[1223 Feedback Core:_del_working_binding] Delete raw/working mem_ids: {delete_ids} for user_name: {user_name}" + f"[1224 Feedback Core:_del_working_binding] Delete raw/working mem_ids: {delete_ids} for user_name: {user_name}" ) except Exception as e: logger.warning( - f"[1223 Feedback Core:_del_working_binding] TreeTextMemory.delete_hard: failed to delete {mid}: {e}" + f"[1224 Feedback Core:_del_working_binding] TreeTextMemory.delete_hard: failed to delete {mid}: {e}" ) def semantics_feedback( @@ -355,13 +429,12 @@ def semantics_feedback( lang = detect_lang("".join(memory_item.memory)) template = FEEDBACK_PROMPT_DICT["compare"][lang] if current_memories == []: - # retrieve feedback - feedback_retrieved = self._retrieve(memory_item.memory, info=info, user_name=user_name) - - # retrieve question + # retrieve last_user_index = max(i for i, d in enumerate(chat_history_list) if d["role"] == "user") last_qa = " ".join([item["content"] for item in chat_history_list[last_user_index:]]) supplementary_retrieved = self._retrieve(last_qa, info=info, user_name=user_name) + feedback_retrieved = self._retrieve(memory_item.memory, info=info, user_name=user_name) + ids = [] for item in feedback_retrieved + supplementary_retrieved: if item.id not in ids: @@ -385,9 +458,14 @@ def semantics_feedback( with ContextThreadPoolExecutor(max_workers=10) as executor: future_to_chunk_idx = {} for chunk in memory_chunks: - current_memories_str = "\n".join( - [f"{item.id}: {item.memory}" for item in chunk] - ) + chunk_list = [] + for item in chunk: + if "preference" in item.metadata.__dict__: + chunk_list.append(f"{item.id}: {item.metadata.preference}") + else: + chunk_list.append(f"{item.id}: {item.memory}") + current_memories_str = "\n".join(chunk_list) + prompt = template.format( now_time=now_time, current_memories=current_memories_str, @@ -408,7 +486,7 @@ def semantics_feedback( all_operations.extend(chunk_operations["operations"]) except Exception as e: logger.error( - f"[1223 Feedback Core: semantics_feedback] Operation failed: {e}" + f"[1224 Feedback Core: semantics_feedback] Operation failed: {e}" ) standard_operations = self.standard_operations(all_operations, current_memories) @@ -458,7 +536,7 @@ def semantics_feedback( update_results.append(result) except Exception as e: logger.error( - f"[1223 Feedback Core: semantics_feedback] Operation failed for {original_op}: {e}", + f"[1224 Feedback Core: semantics_feedback] Operation failed for {original_op}: {e}", exc_info=True, ) if update_results: @@ -486,7 +564,7 @@ def _feedback_memory( ] if filterd_ids: logger.warning( - f"[1223 Feedback Core: _feedback_memory] Since the tags mode is fast, no modifications are made to the following memory {filterd_ids}." + f"[1224 Feedback Core: _feedback_memory] Since the tags mode is fast, no modifications are made to the following memory {filterd_ids}." ) current_memories = [ @@ -518,7 +596,7 @@ def _feedback_memory( results[i] = node except Exception as e: logger.error( - f"[1223 Feedback Core: _feedback_memory] Error processing memory index {i}: {e}", + f"[1224 Feedback Core: _feedback_memory] Error processing memory index {i}: {e}", exc_info=True, ) mem_res = [r for r in results if r] @@ -542,13 +620,18 @@ def _info_comparison(self, memory: TextualMemoryItem, _info: dict, include_keys: record.append(info_v == mem_v) return all(record) - def _retrieve(self, query: str, info=None, top_k=100, user_name=None): + def _retrieve(self, query: str, info=None, top_k=20, user_name=None): """Retrieve memory items""" retrieved_mems = self.searcher.search( query, info=info, user_name=user_name, top_k=top_k, full_recall=True ) retrieved_mems = [item[0] for item in retrieved_mems if float(item[1]) > 0.01] - return retrieved_mems + + pref_info = {} + if "user_id" in info: + pref_info = {"user_id": info["user_id"]} + retrieved_prefs = self.pref_mem.search(query, top_k, pref_info) + return retrieved_mems + retrieved_prefs def _vec_query(self, new_memories_embedding: list[float], user_name=None): """Vector retrieval query""" @@ -577,7 +660,7 @@ def _vec_query(self, new_memories_embedding: list[float], user_name=None): if not retrieved_ids: logger.info( - f"[1223 Feedback Core: _vec_query] No similar memories found for embedding query for user {user_name}." + f"[1224 Feedback Core: _vec_query] No similar memories found for embedding query for user {user_name}." ) filterd_ids = [ @@ -585,7 +668,7 @@ def _vec_query(self, new_memories_embedding: list[float], user_name=None): ] if filterd_ids: logger.warning( - f"[1223 Feedback Core: _vec_query] Since the tags mode is fast, no modifications are made to the following memory {filterd_ids}." + f"[1224 Feedback Core: _vec_query] Since the tags mode is fast, no modifications are made to the following memory {filterd_ids}." ) return [ TextualMemoryItem(**item) @@ -639,9 +722,9 @@ def filter_fault_update(self, operations: list[dict]): ): all_judge.extend(judge_res["operations_judgement"]) except Exception as e: - logger.error(f"[1223 Feedback Core: filter_fault_update] Judgement failed: {e}") + logger.error(f"[1224 Feedback Core: filter_fault_update] Judgement failed: {e}") - logger.info(f"[1223 Feedback Core: filter_fault_update] LLM judgement: {all_judge}") + logger.info(f"[1224 Feedback Core: filter_fault_update] LLM judgement: {all_judge}") id2op = {item["id"]: item for item in updated_operations} valid_updates = [] for judge in all_judge: @@ -652,7 +735,7 @@ def filter_fault_update(self, operations: list[dict]): valid_updates.append(valid_update) logger.info( - f"[1223 Feedback Core: filter_fault_update] {len(updated_operations)} -> {len(valid_updates)}" + f"[1224 Feedback Core: filter_fault_update] {len(updated_operations)} -> {len(valid_updates)}" ) return valid_updates + [item for item in operations if item["operation"] != "UPDATE"] @@ -680,11 +763,11 @@ def correct_item(data): and "text" in data and "old_memory" in data and data["operation"].lower() == "update" - ) + ), "Invalid operation item" if not should_keep_update(data["text"], data["old_memory"]): logger.warning( - f"[1223 Feedback Core: semantics_feedback] Due to the excessive proportion of changes, skip update: {data}" + f"[1224 Feedback Core: correct_item] Due to the excessive proportion of changes, skip update: {data}" ) return None @@ -704,14 +787,14 @@ def correct_item(data): return data except Exception: logger.error( - f"[1223 Feedback Core: standard_operations] Error processing operation item: {data}", + f"[1224 Feedback Core: standard_operations] Error processing operation item: {data}", exc_info=True, ) return None dehallu_res = [correct_item(item) for item in operations] dehalluded_operations = [item for item in dehallu_res if item] - logger.info(f"[1223 Feedback Core: dehalluded_operations] {dehalluded_operations}") + logger.info(f"[1224 Feedback Core: dehalluded_operations] {dehalluded_operations}") # c add objects add_texts = [] @@ -725,7 +808,7 @@ def correct_item(data): elif item["operation"].lower() == "update": llm_operations.append(item) logger.info( - f"[1223 Feedback Core: deduplicate add] {len(dehalluded_operations)} -> {len(llm_operations)} memories" + f"[1224 Feedback Core: deduplicate add] {len(dehalluded_operations)} -> {len(llm_operations)} memories" ) # Update takes precedence over add @@ -739,7 +822,7 @@ def correct_item(data): ] if filtered_items: logger.info( - f"[1223 Feedback Core: semantics_feedback] Due to have update objects, skip add: {filtered_items}" + f"[1224 Feedback Core: semantics_feedback] Due to have update objects, skip add: {filtered_items}" ) return update_items else: @@ -787,7 +870,7 @@ def _doc_filter(self, doc_scope: str, memories: list[TextualMemoryItem]): memid for inscope_file in inscope_docs for memid in filename2_memid[inscope_file] ] logger.info( - f"[1223 Feedback Core: process_keyword_replace] These docs are in scope : {inscope_docs}, relared memids: {inscope_ids}" + f"[1224 Feedback Core: process_keyword_replace] These docs are in scope : {inscope_docs}, relared memids: {inscope_ids}" ) filter_memories = [mem for mem in memories if mem.id in inscope_ids] return filter_memories @@ -841,7 +924,7 @@ def process_keyword_replace( retrieved_memories = self._doc_filter(doc_scope, retrieved_memories) logger.info( - f"[1223 Feedback Core: process_keyword_replace] Keywords recalled memory for user {user_name}: {len(retrieved_ids)} memories | After filtering: {len(retrieved_memories)} memories." + f"[1224 Feedback Core: process_keyword_replace] Keywords recalled memory for user {user_name}: {len(retrieved_ids)} memories | After filtering: {len(retrieved_memories)} memories." ) if not retrieved_memories: @@ -926,7 +1009,7 @@ def check_validity(item): info.update({"user_id": user_id, "user_name": user_name, "session_id": session_id}) logger.info( - f"[1223 Feedback Core: process_feedback_core] Starting memory feedback process for user {user_name}" + f"[1224 Feedback Core: process_feedback_core] Starting memory feedback process for user {user_name}" ) # feedback keywords update kwp_judge = self._keyword_replace_judgement(feedback_content) @@ -959,7 +1042,7 @@ def check_validity(item): if not valid_feedback: logger.warning( - f"[1223 Feedback Core: process_feedback_core] No valid judgements for user {user_name}: {raw_judge}." + f"[1224 Feedback Core: process_feedback_core] No valid judgements for user {user_name}: {raw_judge}." ) return {"record": {"add": [], "update": []}} @@ -1007,13 +1090,13 @@ def check_validity(item): add_memories = mem_record["record"]["add"] update_memories = mem_record["record"]["update"] logger.info( - f"[1223 Feedback Core: process_feedback_core] Processed {len(feedback_memories)} feedback | add {len(add_memories)} memories | update {len(update_memories)} memories for user {user_name}." + f"[1224 Feedback Core: process_feedback_core] Processed {len(feedback_memories)} feedback | add {len(add_memories)} memories | update {len(update_memories)} memories for user {user_name}." ) return mem_record except Exception as e: logger.error( - f"[1223 Feedback Core: process_feedback_core] Error for user {user_name}: {e}" + f"[1224 Feedback Core: process_feedback_core] Error for user {user_name}: {e}" ) return {"record": {"add": [], "update": []}} diff --git a/src/memos/mem_feedback/simple_feedback.py b/src/memos/mem_feedback/simple_feedback.py index 429c2ea20..e32f939c7 100644 --- a/src/memos/mem_feedback/simple_feedback.py +++ b/src/memos/mem_feedback/simple_feedback.py @@ -4,6 +4,7 @@ from memos.llms.factory import AzureLLM, OllamaLLM, OpenAILLM from memos.mem_feedback.feedback import MemFeedback from memos.mem_reader.simple_struct import SimpleStructMemReader +from memos.memories.textual.simple_preference import SimplePreferenceTextMemory from memos.memories.textual.tree_text_memory.organize.manager import MemoryManager from memos.memories.textual.tree_text_memory.retrieve.retrieve_utils import StopwordManager from memos.memories.textual.tree_text_memory.retrieve.searcher import Searcher @@ -23,6 +24,7 @@ def __init__( mem_reader: SimpleStructMemReader, searcher: Searcher, reranker: BaseReranker, + pref_mem: SimplePreferenceTextMemory, ): self.llm = llm self.embedder = embedder @@ -31,5 +33,6 @@ def __init__( self.mem_reader = mem_reader self.searcher = searcher self.stopword_manager = StopwordManager + self.pref_mem = pref_mem self.reranker = reranker self.DB_IDX_READY = False diff --git a/src/memos/mem_feedback/utils.py b/src/memos/mem_feedback/utils.py index c32c12328..8cb7f97a3 100644 --- a/src/memos/mem_feedback/utils.py +++ b/src/memos/mem_feedback/utils.py @@ -48,8 +48,11 @@ def calculate_similarity(text1: str, text2: str) -> float: similarity = calculate_similarity(old_text, new_text) change_ratio = 1 - similarity + if change_ratio == float(0): + return False + if old_len < 200: - return change_ratio < 0.5 + return change_ratio < 0.7 else: return change_ratio < 0.2 diff --git a/src/memos/mem_scheduler/general_modules/init_components_for_scheduler.py b/src/memos/mem_scheduler/general_modules/init_components_for_scheduler.py index ba7b558fd..8fd60153d 100644 --- a/src/memos/mem_scheduler/general_modules/init_components_for_scheduler.py +++ b/src/memos/mem_scheduler/general_modules/init_components_for_scheduler.py @@ -418,6 +418,7 @@ def init_components() -> dict[str, Any]: mem_reader=mem_reader, searcher=searcher, reranker=feedback_reranker, + pref_mem=pref_mem, ) # Return all components as a dictionary for easy access and extension return {"naive_mem_cube": naive_mem_cube, "feedback_server": feedback_server} diff --git a/src/memos/memories/textual/preference.py b/src/memos/memories/textual/preference.py index e1bc0e72b..9e521158d 100644 --- a/src/memos/memories/textual/preference.py +++ b/src/memos/memories/textual/preference.py @@ -87,6 +87,9 @@ def search( Returns: list[TextualMemoryItem]: List of matching memories. """ + if not isinstance(search_filter, dict): + search_filter = {} + search_filter.update({"status": "activated"}) logger.info(f"search_filter for preference memory: {search_filter}") return self.retriever.retrieve(query, top_k, info, search_filter) From ed34f0cb71e38d026e0a8d10cd4434e7378cfe5c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 24 Dec 2025 20:45:16 +0800 Subject: [PATCH 56/72] add pref feedback --- src/memos/memories/textual/simple_preference.py | 3 +++ 1 file changed, 3 insertions(+) diff --git a/src/memos/memories/textual/simple_preference.py b/src/memos/memories/textual/simple_preference.py index 1f02132bb..ee37d638c 100644 --- a/src/memos/memories/textual/simple_preference.py +++ b/src/memos/memories/textual/simple_preference.py @@ -61,6 +61,9 @@ def search( Returns: list[TextualMemoryItem]: List of matching memories. """ + if not isinstance(search_filter, dict): + search_filter = {} + search_filter.update({"status": "activated"}) return self.retriever.retrieve(query, top_k, info, search_filter) def add(self, memories: list[TextualMemoryItem | dict[str, Any]]) -> list[str]: From 827a503fc4db354be31c47e29a89a0c58c25f502 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Mon, 29 Dec 2025 15:05:05 +0800 Subject: [PATCH 57/72] patch: get_memory func filter user id and make page chunk --- src/memos/api/handlers/memory_handler.py | 7 ++++++- src/memos/api/product_models.py | 7 +++++++ src/memos/graph_dbs/nebular.py | 2 +- src/memos/graph_dbs/polardb.py | 1 + src/memos/memories/textual/tree.py | 12 ++++++++++-- 5 files changed, 25 insertions(+), 4 deletions(-) diff --git a/src/memos/api/handlers/memory_handler.py b/src/memos/api/handlers/memory_handler.py index 5cfa98160..2a99d912c 100644 --- a/src/memos/api/handlers/memory_handler.py +++ b/src/memos/api/handlers/memory_handler.py @@ -180,7 +180,12 @@ def handle_get_memories( get_mem_req: GetMemoryRequest, naive_mem_cube: NaiveMemCube ) -> GetMemoryResponse: # TODO: Implement get memory with filter - memories = naive_mem_cube.text_mem.get_all(user_name=get_mem_req.mem_cube_id)["nodes"] + memories = naive_mem_cube.text_mem.get_all( + user_name=get_mem_req.mem_cube_id, + user_id=get_mem_req.user_id, + page=get_mem_req.page, + page_size=get_mem_req.page_size, + )["nodes"] preferences: list[TextualMemoryItem] = [] if get_mem_req.include_preference and naive_mem_cube.pref_mem is not None: filter_params: dict[str, Any] = {} diff --git a/src/memos/api/product_models.py b/src/memos/api/product_models.py index 120da8b55..25e0d809d 100644 --- a/src/memos/api/product_models.py +++ b/src/memos/api/product_models.py @@ -772,6 +772,13 @@ class GetMemoryRequest(BaseRequest): mem_cube_id: str = Field(..., description="Cube ID") user_id: str | None = Field(None, description="User ID") include_preference: bool = Field(True, description="Whether to handle preference memory") + page: int | None = Field( + None, + description="Page number (starts from 1). If None, exports all data without pagination.", + ) + page_size: int | None = Field( + None, description="Number of items per page. If None, exports all data without pagination." + ) class DeleteMemoryRequest(BaseRequest): diff --git a/src/memos/graph_dbs/nebular.py b/src/memos/graph_dbs/nebular.py index 89b58f417..428d6d09e 100644 --- a/src/memos/graph_dbs/nebular.py +++ b/src/memos/graph_dbs/nebular.py @@ -1207,7 +1207,7 @@ def clear(self, user_name: str | None = None) -> None: @timed def export_graph( - self, include_embedding: bool = False, user_name: str | None = None + self, include_embedding: bool = False, user_name: str | None = None, **kwargs ) -> dict[str, Any]: """ Export all graph nodes and edges in a structured form. diff --git a/src/memos/graph_dbs/polardb.py b/src/memos/graph_dbs/polardb.py index fcb7e0caa..4799542bf 100644 --- a/src/memos/graph_dbs/polardb.py +++ b/src/memos/graph_dbs/polardb.py @@ -2507,6 +2507,7 @@ def export_graph( user_name: str | None = None, page: int | None = None, page_size: int | None = None, + **kwargs, ) -> dict[str, Any]: """ Export all graph nodes and edges in a structured form. diff --git a/src/memos/memories/textual/tree.py b/src/memos/memories/textual/tree.py index fb33a2d03..764ceee67 100644 --- a/src/memos/memories/textual/tree.py +++ b/src/memos/memories/textual/tree.py @@ -321,12 +321,20 @@ def get_by_ids( ) -> list[TextualMemoryItem]: raise NotImplementedError - def get_all(self, user_name: str | None = None) -> dict: + def get_all( + self, + user_name: str, + user_id: str | None = None, + page: int | None = None, + page_size: int | None = None, + ) -> dict: """Get all memories. Returns: list[TextualMemoryItem]: List of all memories. """ - all_items = self.graph_store.export_graph(user_name=user_name) + all_items = self.graph_store.export_graph( + user_name=user_name, user_id=user_id, page=page, page_size=page_size + ) return all_items def delete(self, memory_ids: list[str], user_name: str | None = None) -> None: From 4fab2611ece679cfaf3942cd9b8d78c225af6d7a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 30 Dec 2025 10:34:20 +0800 Subject: [PATCH 58/72] add total num --- src/memos/api/handlers/memory_handler.py | 22 +++++++++++++++++++--- src/memos/memories/textual/tree.py | 4 ++-- 2 files changed, 21 insertions(+), 5 deletions(-) diff --git a/src/memos/api/handlers/memory_handler.py b/src/memos/api/handlers/memory_handler.py index 2a99d912c..de0f7d0c5 100644 --- a/src/memos/api/handlers/memory_handler.py +++ b/src/memos/api/handlers/memory_handler.py @@ -185,7 +185,10 @@ def handle_get_memories( user_id=get_mem_req.user_id, page=get_mem_req.page, page_size=get_mem_req.page_size, - )["nodes"] + ) + memory_nodes = memories["nodes"] + memory_edges = memories["edges"] + preferences: list[TextualMemoryItem] = [] if get_mem_req.include_preference and naive_mem_cube.pref_mem is not None: filter_params: dict[str, Any] = {} @@ -198,8 +201,21 @@ def handle_get_memories( return GetMemoryResponse( message="Memories retrieved successfully", data={ - "text_mem": [{"cube_id": get_mem_req.mem_cube_id, "memories": memories}], - "pref_mem": [{"cube_id": get_mem_req.mem_cube_id, "memories": preferences}], + "text_mem": [ + { + "cube_id": get_mem_req.mem_cube_id, + "memories": memories, + "total_nodes": len(memory_nodes), + "total_edges": len(memory_edges), + } + ], + "pref_mem": [ + { + "cube_id": get_mem_req.mem_cube_id, + "memories": preferences, + "total_nodes": len(preferences), + } + ], }, ) diff --git a/src/memos/memories/textual/tree.py b/src/memos/memories/textual/tree.py index 764ceee67..e576c0ea9 100644 --- a/src/memos/memories/textual/tree.py +++ b/src/memos/memories/textual/tree.py @@ -332,10 +332,10 @@ def get_all( Returns: list[TextualMemoryItem]: List of all memories. """ - all_items = self.graph_store.export_graph( + graph_output = self.graph_store.export_graph( user_name=user_name, user_id=user_id, page=page, page_size=page_size ) - return all_items + return graph_output def delete(self, memory_ids: list[str], user_name: str | None = None) -> None: """Hard delete: permanently remove nodes and their edges from the graph.""" From b5f71a54b1c2719fb28018717e40e8a22537ac6a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 30 Dec 2025 10:48:54 +0800 Subject: [PATCH 59/72] add total num --- src/memos/api/handlers/memory_handler.py | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) diff --git a/src/memos/api/handlers/memory_handler.py b/src/memos/api/handlers/memory_handler.py index de0f7d0c5..d05da19db 100644 --- a/src/memos/api/handlers/memory_handler.py +++ b/src/memos/api/handlers/memory_handler.py @@ -186,8 +186,10 @@ def handle_get_memories( page=get_mem_req.page, page_size=get_mem_req.page_size, ) - memory_nodes = memories["nodes"] - memory_edges = memories["edges"] + total_nodes = memories["total_nodes"] + total_edges = memories["total_edges"] + del memories["total_nodes"] + del memories["total_edges"] preferences: list[TextualMemoryItem] = [] if get_mem_req.include_preference and naive_mem_cube.pref_mem is not None: @@ -198,6 +200,7 @@ def handle_get_memories( filter_params["mem_cube_id"] = get_mem_req.mem_cube_id preferences = naive_mem_cube.pref_mem.get_memory_by_filter(filter_params) preferences = [format_memory_item(mem) for mem in preferences] + return GetMemoryResponse( message="Memories retrieved successfully", data={ @@ -205,8 +208,8 @@ def handle_get_memories( { "cube_id": get_mem_req.mem_cube_id, "memories": memories, - "total_nodes": len(memory_nodes), - "total_edges": len(memory_edges), + "total_nodes": total_nodes, + "total_edges": total_edges, } ], "pref_mem": [ From 098a83031d1a061b4021cfbe3174aada32826c12 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 30 Dec 2025 14:51:23 +0800 Subject: [PATCH 60/72] add milvus pagination --- src/memos/api/handlers/memory_handler.py | 19 +++++- src/memos/memories/textual/preference.py | 38 ++++++++--- src/memos/vec_dbs/milvus.py | 86 +++++++++++++++++------- 3 files changed, 103 insertions(+), 40 deletions(-) diff --git a/src/memos/api/handlers/memory_handler.py b/src/memos/api/handlers/memory_handler.py index d05da19db..941b59106 100644 --- a/src/memos/api/handlers/memory_handler.py +++ b/src/memos/api/handlers/memory_handler.py @@ -192,14 +192,26 @@ def handle_get_memories( del memories["total_edges"] preferences: list[TextualMemoryItem] = [] + total_explicit_nodes, total_implicit_nodes = 0, 0 if get_mem_req.include_preference and naive_mem_cube.pref_mem is not None: filter_params: dict[str, Any] = {} if get_mem_req.user_id is not None: filter_params["user_id"] = get_mem_req.user_id if get_mem_req.mem_cube_id is not None: filter_params["mem_cube_id"] = get_mem_req.mem_cube_id - preferences = naive_mem_cube.pref_mem.get_memory_by_filter(filter_params) - preferences = [format_memory_item(mem) for mem in preferences] + preferences = naive_mem_cube.pref_mem.get_memory_by_filter( + filter_params, page=get_mem_req.page, page_size=get_mem_req.page_size + ) + + for key, value_list in preferences.items(): + if key in ["explicit_preference", "implicit_preference"]: + formatted_list = [format_memory_item(item) for item in value_list] + preferences[key] = formatted_list + + total_explicit_nodes = preferences["total_explicit_nodes"] + total_implicit_nodes = preferences["total_implicit_nodes"] + del preferences["total_explicit_nodes"] + del preferences["total_implicit_nodes"] return GetMemoryResponse( message="Memories retrieved successfully", @@ -216,7 +228,8 @@ def handle_get_memories( { "cube_id": get_mem_req.mem_cube_id, "memories": preferences, - "total_nodes": len(preferences), + "total_explicit_nodes": total_explicit_nodes, + "total_implicit_nodes": total_implicit_nodes, } ], }, diff --git a/src/memos/memories/textual/preference.py b/src/memos/memories/textual/preference.py index 9e521158d..75d7d2a4c 100644 --- a/src/memos/memories/textual/preference.py +++ b/src/memos/memories/textual/preference.py @@ -261,7 +261,9 @@ def get_all(self) -> list[TextualMemoryItem]: ] return all_memories - def get_memory_by_filter(self, filter: dict[str, Any] | None = None) -> list[TextualMemoryItem]: + def get_memory_by_filter( + self, filter: dict[str, Any] | None = None, **kwargs + ) -> list[TextualMemoryItem]: """Get memories by filter. Args: filter (dict[str, Any]): Filter criteria. @@ -269,18 +271,32 @@ def get_memory_by_filter(self, filter: dict[str, Any] | None = None) -> list[Tex list[TextualMemoryItem]: List of memories that match the filter. """ collection_list = self.vector_db.config.collection_name - all_db_items = [] + + memories = {} + total_explicit_nodes = 0 + total_implicit_nodes = 0 for collection_name in collection_list: - db_items = self.vector_db.get_by_filter(collection_name=collection_name, filter=filter) - all_db_items.extend(db_items) - memories = [ - TextualMemoryItem( - id=memo.id, - memory=memo.memory, - metadata=PreferenceTextualMemoryMetadata(**memo.payload), + memories[collection_name] = [] + db_items, total_count = self.vector_db.get_by_filter( + collection_name=collection_name, filter=filter, count_total=True, **kwargs ) - for memo in all_db_items - ] + db_items_memory = [ + TextualMemoryItem( + id=memo.id, + memory=memo.memory, + metadata=PreferenceTextualMemoryMetadata(**memo.payload), + ) + for memo in db_items + ] + memories[collection_name].extend(db_items_memory) + + if collection_name == "explicit_preference": + total_explicit_nodes = total_count + if collection_name == "implicit_preference": + total_implicit_nodes = total_count + memories["total_explicit_nodes"] = total_explicit_nodes + memories["total_implicit_nodes"] = total_implicit_nodes + return memories def delete(self, memory_ids: list[str]) -> None: diff --git a/src/memos/vec_dbs/milvus.py b/src/memos/vec_dbs/milvus.py index ecbca5815..b0753b31d 100644 --- a/src/memos/vec_dbs/milvus.py +++ b/src/memos/vec_dbs/milvus.py @@ -493,7 +493,14 @@ def get_by_ids(self, collection_name: str, ids: list[str]) -> list[MilvusVecDBIt return items def get_by_filter( - self, collection_name: str, filter: dict[str, Any], scroll_limit: int = 100 + self, + collection_name: str, + filter: dict[str, Any], + scroll_limit: int = 100, + page: int | None = None, + page_size: int | None = None, + count_total=False, + **kwargs, ) -> list[MilvusVecDBItem]: """ Retrieve all items that match the given filter criteria using query_iterator. @@ -506,47 +513,74 @@ def get_by_filter( List of items including vectors and payload that match the filter """ expr = self._dict_to_expr(filter) if filter else "" - all_items = [] + if count_total: + total_count = 0 + count_iterator = self.client.query_iterator( + collection_name=collection_name, + filter=expr, + batch_size=scroll_limit, + output_fields=["id"], + ) + try: + while True: + batch = count_iterator.next() + if not batch: + break + total_count += len(batch) + finally: + count_iterator.close() + + result = [] + skipped = 0 + needed = page_size - # Use query_iterator for efficient pagination iterator = self.client.query_iterator( collection_name=collection_name, filter=expr, batch_size=scroll_limit, - output_fields=["*"], # Include all fields including payload + output_fields=["*"], ) - # Iterate through all batches try: - while True: - batch_results = iterator.next() - - if not batch_results: + while needed > 0: + batch = iterator.next() + if not batch: break - # Convert batch results to MilvusVecDBItem objects - for entity in batch_results: - # Extract the actual payload from Milvus entity + for entity in batch: + skipped += 1 + + if skipped <= (page - 1) * page_size: + continue + payload = entity.get("payload", {}) - all_items.append( - MilvusVecDBItem( - id=entity["id"], - memory=entity.get("memory"), - original_text=entity.get("original_text"), - vector=entity.get("vector"), - payload=payload, - ) + item = MilvusVecDBItem( + id=entity["id"], + memory=entity.get("memory"), + original_text=entity.get("original_text"), + vector=entity.get("vector"), + payload=payload, ) + result.append(item) + needed -= 1 + + if needed <= 0: + if count_total: + return result, total_count + return result + except Exception as e: - logger.warning( - f"Error during Milvus query iteration: {e}. Returning {len(all_items)} items found so far." - ) + logger.warning(f"Error during iteration: {e}") finally: - # Close the iterator iterator.close() - logger.info(f"Milvus retrieve by filter completed with {len(all_items)} results.") - return all_items + logger.info( + f"Milvus retrieve by filter completed - " + f"page {page}, page_size {page_size}, got {len(result)} items." + ) + if count_total: + return result, total_count + return result def get_all(self, collection_name: str, scroll_limit=100) -> list[MilvusVecDBItem]: """Retrieve all items in the vector database.""" From 3fd1743a6703c84baa4af79fbece3de1130fdfac Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 30 Dec 2025 17:27:35 +0800 Subject: [PATCH 61/72] fix merge implicit explicit pref --- src/memos/api/handlers/memory_handler.py | 20 ++---- src/memos/memories/textual/preference.py | 42 +++++++----- src/memos/vec_dbs/milvus.py | 86 +++++++----------------- 3 files changed, 55 insertions(+), 93 deletions(-) diff --git a/src/memos/api/handlers/memory_handler.py b/src/memos/api/handlers/memory_handler.py index 941b59106..d46c83e94 100644 --- a/src/memos/api/handlers/memory_handler.py +++ b/src/memos/api/handlers/memory_handler.py @@ -192,26 +192,17 @@ def handle_get_memories( del memories["total_edges"] preferences: list[TextualMemoryItem] = [] - total_explicit_nodes, total_implicit_nodes = 0, 0 if get_mem_req.include_preference and naive_mem_cube.pref_mem is not None: filter_params: dict[str, Any] = {} if get_mem_req.user_id is not None: filter_params["user_id"] = get_mem_req.user_id if get_mem_req.mem_cube_id is not None: filter_params["mem_cube_id"] = get_mem_req.mem_cube_id - preferences = naive_mem_cube.pref_mem.get_memory_by_filter( + + preferences, total_pref = naive_mem_cube.pref_mem.get_memory_by_filter( filter_params, page=get_mem_req.page, page_size=get_mem_req.page_size ) - - for key, value_list in preferences.items(): - if key in ["explicit_preference", "implicit_preference"]: - formatted_list = [format_memory_item(item) for item in value_list] - preferences[key] = formatted_list - - total_explicit_nodes = preferences["total_explicit_nodes"] - total_implicit_nodes = preferences["total_implicit_nodes"] - del preferences["total_explicit_nodes"] - del preferences["total_implicit_nodes"] + format_preferences = [format_memory_item(item) for item in preferences] return GetMemoryResponse( message="Memories retrieved successfully", @@ -227,9 +218,8 @@ def handle_get_memories( "pref_mem": [ { "cube_id": get_mem_req.mem_cube_id, - "memories": preferences, - "total_explicit_nodes": total_explicit_nodes, - "total_implicit_nodes": total_implicit_nodes, + "memories": format_preferences, + "total_nodes": total_pref, } ], }, diff --git a/src/memos/memories/textual/preference.py b/src/memos/memories/textual/preference.py index 75d7d2a4c..cb4f00735 100644 --- a/src/memos/memories/textual/preference.py +++ b/src/memos/memories/textual/preference.py @@ -1,6 +1,7 @@ import json import os +from datetime import datetime from typing import Any from memos.configs.memory import PreferenceTextMemoryConfig @@ -262,8 +263,11 @@ def get_all(self) -> list[TextualMemoryItem]: return all_memories def get_memory_by_filter( - self, filter: dict[str, Any] | None = None, **kwargs - ) -> list[TextualMemoryItem]: + self, + filter: dict[str, Any] | None = None, + page: int | None = None, + page_size: int | None = None, + ): """Get memories by filter. Args: filter (dict[str, Any]): Filter criteria. @@ -272,14 +276,9 @@ def get_memory_by_filter( """ collection_list = self.vector_db.config.collection_name - memories = {} - total_explicit_nodes = 0 - total_implicit_nodes = 0 + memories = [] for collection_name in collection_list: - memories[collection_name] = [] - db_items, total_count = self.vector_db.get_by_filter( - collection_name=collection_name, filter=filter, count_total=True, **kwargs - ) + db_items = self.vector_db.get_by_filter(collection_name=collection_name, filter=filter) db_items_memory = [ TextualMemoryItem( id=memo.id, @@ -288,16 +287,23 @@ def get_memory_by_filter( ) for memo in db_items ] - memories[collection_name].extend(db_items_memory) + memories.extend(db_items_memory) - if collection_name == "explicit_preference": - total_explicit_nodes = total_count - if collection_name == "implicit_preference": - total_implicit_nodes = total_count - memories["total_explicit_nodes"] = total_explicit_nodes - memories["total_implicit_nodes"] = total_implicit_nodes - - return memories + # sort + sorted_memories = sorted( + memories, + key=lambda item: datetime.fromisoformat(item.metadata.created_at), + reverse=True, + ) + if page and page_size: + if page < 1: + page = 1 + if page_size < 1: + page_size = 10 + pick_memories = sorted_memories[(page - 1) * page_size : page * page_size] + return pick_memories, len(sorted_memories) + + return sorted_memories, len(sorted_memories) def delete(self, memory_ids: list[str]) -> None: """Delete memories. diff --git a/src/memos/vec_dbs/milvus.py b/src/memos/vec_dbs/milvus.py index b0753b31d..ecbca5815 100644 --- a/src/memos/vec_dbs/milvus.py +++ b/src/memos/vec_dbs/milvus.py @@ -493,14 +493,7 @@ def get_by_ids(self, collection_name: str, ids: list[str]) -> list[MilvusVecDBIt return items def get_by_filter( - self, - collection_name: str, - filter: dict[str, Any], - scroll_limit: int = 100, - page: int | None = None, - page_size: int | None = None, - count_total=False, - **kwargs, + self, collection_name: str, filter: dict[str, Any], scroll_limit: int = 100 ) -> list[MilvusVecDBItem]: """ Retrieve all items that match the given filter criteria using query_iterator. @@ -513,74 +506,47 @@ def get_by_filter( List of items including vectors and payload that match the filter """ expr = self._dict_to_expr(filter) if filter else "" - if count_total: - total_count = 0 - count_iterator = self.client.query_iterator( - collection_name=collection_name, - filter=expr, - batch_size=scroll_limit, - output_fields=["id"], - ) - try: - while True: - batch = count_iterator.next() - if not batch: - break - total_count += len(batch) - finally: - count_iterator.close() - - result = [] - skipped = 0 - needed = page_size + all_items = [] + # Use query_iterator for efficient pagination iterator = self.client.query_iterator( collection_name=collection_name, filter=expr, batch_size=scroll_limit, - output_fields=["*"], + output_fields=["*"], # Include all fields including payload ) + # Iterate through all batches try: - while needed > 0: - batch = iterator.next() - if not batch: - break - - for entity in batch: - skipped += 1 + while True: + batch_results = iterator.next() - if skipped <= (page - 1) * page_size: - continue + if not batch_results: + break + # Convert batch results to MilvusVecDBItem objects + for entity in batch_results: + # Extract the actual payload from Milvus entity payload = entity.get("payload", {}) - item = MilvusVecDBItem( - id=entity["id"], - memory=entity.get("memory"), - original_text=entity.get("original_text"), - vector=entity.get("vector"), - payload=payload, + all_items.append( + MilvusVecDBItem( + id=entity["id"], + memory=entity.get("memory"), + original_text=entity.get("original_text"), + vector=entity.get("vector"), + payload=payload, + ) ) - result.append(item) - needed -= 1 - - if needed <= 0: - if count_total: - return result, total_count - return result - except Exception as e: - logger.warning(f"Error during iteration: {e}") + logger.warning( + f"Error during Milvus query iteration: {e}. Returning {len(all_items)} items found so far." + ) finally: + # Close the iterator iterator.close() - logger.info( - f"Milvus retrieve by filter completed - " - f"page {page}, page_size {page_size}, got {len(result)} items." - ) - if count_total: - return result, total_count - return result + logger.info(f"Milvus retrieve by filter completed with {len(all_items)} results.") + return all_items def get_all(self, collection_name: str, scroll_limit=100) -> list[MilvusVecDBItem]: """Retrieve all items in the vector database.""" From af610fea6e1e793cc8cada1463e01a4c02ab60d3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 30 Dec 2025 17:30:57 +0800 Subject: [PATCH 62/72] fix merge implicit explicit pref --- src/memos/vec_dbs/milvus.py | 86 +++++++++++-------------------------- 1 file changed, 26 insertions(+), 60 deletions(-) diff --git a/src/memos/vec_dbs/milvus.py b/src/memos/vec_dbs/milvus.py index b0753b31d..ecbca5815 100644 --- a/src/memos/vec_dbs/milvus.py +++ b/src/memos/vec_dbs/milvus.py @@ -493,14 +493,7 @@ def get_by_ids(self, collection_name: str, ids: list[str]) -> list[MilvusVecDBIt return items def get_by_filter( - self, - collection_name: str, - filter: dict[str, Any], - scroll_limit: int = 100, - page: int | None = None, - page_size: int | None = None, - count_total=False, - **kwargs, + self, collection_name: str, filter: dict[str, Any], scroll_limit: int = 100 ) -> list[MilvusVecDBItem]: """ Retrieve all items that match the given filter criteria using query_iterator. @@ -513,74 +506,47 @@ def get_by_filter( List of items including vectors and payload that match the filter """ expr = self._dict_to_expr(filter) if filter else "" - if count_total: - total_count = 0 - count_iterator = self.client.query_iterator( - collection_name=collection_name, - filter=expr, - batch_size=scroll_limit, - output_fields=["id"], - ) - try: - while True: - batch = count_iterator.next() - if not batch: - break - total_count += len(batch) - finally: - count_iterator.close() - - result = [] - skipped = 0 - needed = page_size + all_items = [] + # Use query_iterator for efficient pagination iterator = self.client.query_iterator( collection_name=collection_name, filter=expr, batch_size=scroll_limit, - output_fields=["*"], + output_fields=["*"], # Include all fields including payload ) + # Iterate through all batches try: - while needed > 0: - batch = iterator.next() - if not batch: - break - - for entity in batch: - skipped += 1 + while True: + batch_results = iterator.next() - if skipped <= (page - 1) * page_size: - continue + if not batch_results: + break + # Convert batch results to MilvusVecDBItem objects + for entity in batch_results: + # Extract the actual payload from Milvus entity payload = entity.get("payload", {}) - item = MilvusVecDBItem( - id=entity["id"], - memory=entity.get("memory"), - original_text=entity.get("original_text"), - vector=entity.get("vector"), - payload=payload, + all_items.append( + MilvusVecDBItem( + id=entity["id"], + memory=entity.get("memory"), + original_text=entity.get("original_text"), + vector=entity.get("vector"), + payload=payload, + ) ) - result.append(item) - needed -= 1 - - if needed <= 0: - if count_total: - return result, total_count - return result - except Exception as e: - logger.warning(f"Error during iteration: {e}") + logger.warning( + f"Error during Milvus query iteration: {e}. Returning {len(all_items)} items found so far." + ) finally: + # Close the iterator iterator.close() - logger.info( - f"Milvus retrieve by filter completed - " - f"page {page}, page_size {page_size}, got {len(result)} items." - ) - if count_total: - return result, total_count - return result + logger.info(f"Milvus retrieve by filter completed with {len(all_items)} results.") + return all_items def get_all(self, collection_name: str, scroll_limit=100) -> list[MilvusVecDBItem]: """Retrieve all items in the vector database.""" From eab230369f64442530e2c11ea537ebe820a3bb2d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 30 Dec 2025 17:45:02 +0800 Subject: [PATCH 63/72] fix merge implicit explicit pref --- src/memos/api/handlers/memory_handler.py | 1 + 1 file changed, 1 insertion(+) diff --git a/src/memos/api/handlers/memory_handler.py b/src/memos/api/handlers/memory_handler.py index cdacbea5a..a744e16e2 100644 --- a/src/memos/api/handlers/memory_handler.py +++ b/src/memos/api/handlers/memory_handler.py @@ -192,6 +192,7 @@ def handle_get_memories( del memories["total_edges"] preferences: list[TextualMemoryItem] = [] + total_pref = 0 if get_mem_req.include_preference and naive_mem_cube.pref_mem is not None: filter_params: dict[str, Any] = {} From 83061d12d20523c7415fcda8d68049418a29c839 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Wed, 7 Jan 2026 17:12:51 +0800 Subject: [PATCH 64/72] fix json load bug --- src/memos/mem_feedback/feedback.py | 99 ++++++++++++++++++------------ src/memos/mem_feedback/utils.py | 81 ++++++++++++++++++++++++ 2 files changed, 141 insertions(+), 39 deletions(-) diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index fad15a7cd..15d7c336a 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -5,7 +5,7 @@ import uuid from datetime import datetime -from typing import TYPE_CHECKING, Any +from typing import TYPE_CHECKING, Any, Literal from tenacity import retry, stop_after_attempt, wait_random_exponential @@ -18,6 +18,8 @@ from memos.log import get_logger from memos.mem_feedback.base import BaseMemFeedback from memos.mem_feedback.utils import ( + extract_bracket_content, + extract_square_brackets_content, general_split_into_chunks, make_mem_item, should_keep_update, @@ -118,7 +120,7 @@ def _retry_db_operation(self, operation): return operation() except Exception as e: logger.error( - f"[1224 Feedback Core: _retry_db_operation] DB operation failed: {e}", exc_info=True + f"[0107 Feedback Core: _retry_db_operation] DB operation failed: {e}", exc_info=True ) raise @@ -132,7 +134,7 @@ def _batch_embed(self, texts: list[str], embed_bs: int = 5): results.extend(self._embed_once(batch)) except Exception as e: logger.error( - f"[1224 Feedback Core: process_feedback_core] Embedding batch failed, Cover with all zeros: {len(batch)} entries: {e}" + f"[0107 Feedback Core: process_feedback_core] Embedding batch failed, Cover with all zeros: {len(batch)} entries: {e}" ) results.extend([[0.0] * dim for _ in range(len(batch))]) return results @@ -148,7 +150,7 @@ def _pure_add(self, user_name: str, feedback_content: str, feedback_time: str, i lambda: self.memory_manager.add(to_add_memories, user_name=user_name, use_batch=False) ) logger.info( - f"[1224 Feedback Core: _pure_add] Pure added {len(added_ids)} memories for user {user_name}." + f"[0107 Feedback Core: _pure_add] Pure added {len(added_ids)} memories for user {user_name}." ) return { "record": { @@ -180,12 +182,12 @@ def _keyword_replace_judgement(self, feedback_content: str) -> dict | None: user_feedback=feedback_content, ) - judge_res = self._get_llm_response(prompt) + judge_res = self._get_llm_response(prompt, load_type="bracket") if judge_res: return judge_res else: logger.warning( - "[1224 Feedback Core: _feedback_judgement] feedback judgement failed, return []" + "[0107 Feedback Core: _feedback_judgement] feedback judgement failed, return []" ) return {} @@ -205,12 +207,12 @@ def _feedback_judgement( feedback_time=feedback_time, ) - judge_res = self._get_llm_response(prompt) + judge_res = self._get_llm_response(prompt, load_type="square_bracket") if judge_res: return judge_res else: logger.warning( - "[1224 Feedback Core: _feedback_judgement] feedback judgement failed, return []" + "[0107 Feedback Core: _feedback_judgement] feedback judgement failed, return []" ) return [] @@ -276,7 +278,7 @@ def _single_update_operation( """ if "preference" in old_memory_item.metadata.__dict__: logger.info( - f"[1224 Feedback Core: _single_update_operation] pref_memory: {old_memory_item.id}" + f"[0107 Feedback Core: _single_update_operation] pref_memory: {old_memory_item.id}" ) return self._single_update_pref( old_memory_item, new_memory_item, user_id, user_name, operation @@ -408,11 +410,11 @@ def _del_working_binding(self, user_name, mem_items: list[TextualMemoryItem]) -> self.graph_store.delete_node(mid, user_name=user_name) logger.info( - f"[1224 Feedback Core:_del_working_binding] Delete raw/working mem_ids: {delete_ids} for user_name: {user_name}" + f"[0107 Feedback Core:_del_working_binding] Delete raw/working mem_ids: {delete_ids} for user_name: {user_name}" ) except Exception as e: logger.warning( - f"[1224 Feedback Core:_del_working_binding] TreeTextMemory.delete_hard: failed to delete {mid}: {e}" + f"[0107 Feedback Core:_del_working_binding] TreeTextMemory.delete_hard: failed to delete {mid}: {e}" ) def semantics_feedback( @@ -473,7 +475,7 @@ def semantics_feedback( chat_history=history_str, ) - future = executor.submit(self._get_llm_response, prompt) + future = executor.submit(self._get_llm_response, prompt, load_type="bracket") future_to_chunk_idx[future] = chunk for future in concurrent.futures.as_completed(future_to_chunk_idx): try: @@ -486,7 +488,7 @@ def semantics_feedback( all_operations.extend(chunk_operations["operations"]) except Exception as e: logger.error( - f"[1224 Feedback Core: semantics_feedback] Operation failed: {e}" + f"[0107 Feedback Core: semantics_feedback] Operation failed: {e}" ) standard_operations = self.standard_operations(all_operations, current_memories) @@ -536,7 +538,7 @@ def semantics_feedback( update_results.append(result) except Exception as e: logger.error( - f"[1224 Feedback Core: semantics_feedback] Operation failed for {original_op}: {e}", + f"[0107 Feedback Core: semantics_feedback] Operation failed for {original_op}: {e}", exc_info=True, ) if update_results: @@ -564,7 +566,7 @@ def _feedback_memory( ] if filterd_ids: logger.warning( - f"[1224 Feedback Core: _feedback_memory] Since the tags mode is fast, no modifications are made to the following memory {filterd_ids}." + f"[0107 Feedback Core: _feedback_memory] Since the tags mode is fast, no modifications are made to the following memory {filterd_ids}." ) current_memories = [ @@ -596,7 +598,7 @@ def _feedback_memory( results[i] = node except Exception as e: logger.error( - f"[1224 Feedback Core: _feedback_memory] Error processing memory index {i}: {e}", + f"[0107 Feedback Core: _feedback_memory] Error processing memory index {i}: {e}", exc_info=True, ) mem_res = [r for r in results if r] @@ -660,7 +662,7 @@ def _vec_query(self, new_memories_embedding: list[float], user_name=None): if not retrieved_ids: logger.info( - f"[1224 Feedback Core: _vec_query] No similar memories found for embedding query for user {user_name}." + f"[0107 Feedback Core: _vec_query] No similar memories found for embedding query for user {user_name}." ) filterd_ids = [ @@ -668,7 +670,7 @@ def _vec_query(self, new_memories_embedding: list[float], user_name=None): ] if filterd_ids: logger.warning( - f"[1224 Feedback Core: _vec_query] Since the tags mode is fast, no modifications are made to the following memory {filterd_ids}." + f"[0107 Feedback Core: _vec_query] Since the tags mode is fast, no modifications are made to the following memory {filterd_ids}." ) return [ TextualMemoryItem(**item) @@ -676,22 +678,41 @@ def _vec_query(self, new_memories_embedding: list[float], user_name=None): if "mode:fast" not in item["metadata"]["tags"] ] - def _get_llm_response(self, prompt: str, dsl: bool = True) -> dict: + def _get_llm_response( + self, + prompt: str, + dsl: bool = True, + load_type: Literal["bracket", "square_bracket"] | None = None, + ) -> dict: messages = [{"role": "user", "content": prompt}] + response_text = "" try: response_text = self.llm.generate(messages, temperature=0.3, timeout=60) - if dsl: + if not dsl: + return response_text + try: response_text = response_text.replace("```", "").replace("json", "") cleaned_text = re.sub(r"[\x00-\x08\x0B\x0C\x0E-\x1F\x7F]", "", response_text) response_json = json.loads(cleaned_text) - else: - return response_text + return response_json + except (json.JSONDecodeError, ValueError) as e: + if load_type == "bracket": + response_json = extract_bracket_content(response_text) + return response_json + elif load_type == "square_bracket": + response_json = extract_square_brackets_content(response_text) + return response_json + else: + logger.error( + f"[Feedback Core LLM Error] Exception during chat generation: {e} | response_text： {response_text}" + ) + return None + except Exception as e: logger.error( f"[Feedback Core LLM Error] Exception during chat generation: {e} | response_text： {response_text}" ) - response_json = None - return response_json + return None def filter_fault_update(self, operations: list[dict]): """To address the randomness of large model outputs, it is necessary to conduct validity evaluation on the texts used for memory override operations.""" @@ -710,7 +731,7 @@ def filter_fault_update(self, operations: list[dict]): raw_operations_str = {"operations": chunk} prompt = template.format(raw_operations=str(raw_operations_str)) - future = executor.submit(self._get_llm_response, prompt) + future = executor.submit(self._get_llm_response, prompt, load_type="bracket") future_to_chunk_idx[future] = chunk for future in concurrent.futures.as_completed(future_to_chunk_idx): try: @@ -722,9 +743,9 @@ def filter_fault_update(self, operations: list[dict]): ): all_judge.extend(judge_res["operations_judgement"]) except Exception as e: - logger.error(f"[1224 Feedback Core: filter_fault_update] Judgement failed: {e}") + logger.error(f"[0107 Feedback Core: filter_fault_update] Judgement failed: {e}") - logger.info(f"[1224 Feedback Core: filter_fault_update] LLM judgement: {all_judge}") + logger.info(f"[0107 Feedback Core: filter_fault_update] LLM judgement: {all_judge}") id2op = {item["id"]: item for item in updated_operations} valid_updates = [] for judge in all_judge: @@ -735,7 +756,7 @@ def filter_fault_update(self, operations: list[dict]): valid_updates.append(valid_update) logger.info( - f"[1224 Feedback Core: filter_fault_update] {len(updated_operations)} -> {len(valid_updates)}" + f"[0107 Feedback Core: filter_fault_update] {len(updated_operations)} -> {len(valid_updates)}" ) return valid_updates + [item for item in operations if item["operation"] != "UPDATE"] @@ -767,7 +788,7 @@ def correct_item(data): if not should_keep_update(data["text"], data["old_memory"]): logger.warning( - f"[1224 Feedback Core: correct_item] Due to the excessive proportion of changes, skip update: {data}" + f"[0107 Feedback Core: correct_item] Due to the excessive proportion of changes, skip update: {data}" ) return None @@ -787,14 +808,14 @@ def correct_item(data): return data except Exception: logger.error( - f"[1224 Feedback Core: standard_operations] Error processing operation item: {data}", + f"[0107 Feedback Core: standard_operations] Error processing operation item: {data}", exc_info=True, ) return None dehallu_res = [correct_item(item) for item in operations] dehalluded_operations = [item for item in dehallu_res if item] - logger.info(f"[1224 Feedback Core: dehalluded_operations] {dehalluded_operations}") + logger.info(f"[0107 Feedback Core: dehalluded_operations] {dehalluded_operations}") # c add objects add_texts = [] @@ -808,7 +829,7 @@ def correct_item(data): elif item["operation"].lower() == "update": llm_operations.append(item) logger.info( - f"[1224 Feedback Core: deduplicate add] {len(dehalluded_operations)} -> {len(llm_operations)} memories" + f"[0107 Feedback Core: deduplicate add] {len(dehalluded_operations)} -> {len(llm_operations)} memories" ) # Update takes precedence over add @@ -822,7 +843,7 @@ def correct_item(data): ] if filtered_items: logger.info( - f"[1224 Feedback Core: semantics_feedback] Due to have update objects, skip add: {filtered_items}" + f"[0107 Feedback Core: semantics_feedback] Due to have update objects, skip add: {filtered_items}" ) return update_items else: @@ -870,7 +891,7 @@ def _doc_filter(self, doc_scope: str, memories: list[TextualMemoryItem]): memid for inscope_file in inscope_docs for memid in filename2_memid[inscope_file] ] logger.info( - f"[1224 Feedback Core: process_keyword_replace] These docs are in scope : {inscope_docs}, relared memids: {inscope_ids}" + f"[0107 Feedback Core: process_keyword_replace] These docs are in scope : {inscope_docs}, relared memids: {inscope_ids}" ) filter_memories = [mem for mem in memories if mem.id in inscope_ids] return filter_memories @@ -924,7 +945,7 @@ def process_keyword_replace( retrieved_memories = self._doc_filter(doc_scope, retrieved_memories) logger.info( - f"[1224 Feedback Core: process_keyword_replace] Keywords recalled memory for user {user_name}: {len(retrieved_ids)} memories | After filtering: {len(retrieved_memories)} memories." + f"[0107 Feedback Core: process_keyword_replace] Keywords recalled memory for user {user_name}: {len(retrieved_ids)} memories | After filtering: {len(retrieved_memories)} memories." ) if not retrieved_memories: @@ -1009,7 +1030,7 @@ def check_validity(item): info.update({"user_id": user_id, "user_name": user_name, "session_id": session_id}) logger.info( - f"[1224 Feedback Core: process_feedback_core] Starting memory feedback process for user {user_name}" + f"[0107 Feedback Core: process_feedback_core] Starting memory feedback process for user {user_name}" ) # feedback keywords update kwp_judge = self._keyword_replace_judgement(feedback_content) @@ -1042,7 +1063,7 @@ def check_validity(item): if not valid_feedback: logger.warning( - f"[1224 Feedback Core: process_feedback_core] No valid judgements for user {user_name}: {raw_judge}." + f"[0107 Feedback Core: process_feedback_core] No valid judgements for user {user_name}: {raw_judge}." ) return {"record": {"add": [], "update": []}} @@ -1090,13 +1111,13 @@ def check_validity(item): add_memories = mem_record["record"]["add"] update_memories = mem_record["record"]["update"] logger.info( - f"[1224 Feedback Core: process_feedback_core] Processed {len(feedback_memories)} feedback | add {len(add_memories)} memories | update {len(update_memories)} memories for user {user_name}." + f"[0107 Feedback Core: process_feedback_core] Processed {len(feedback_memories)} feedback | add {len(add_memories)} memories | update {len(update_memories)} memories for user {user_name}." ) return mem_record except Exception as e: logger.error( - f"[1224 Feedback Core: process_feedback_core] Error for user {user_name}: {e}" + f"[0107 Feedback Core: process_feedback_core] Error for user {user_name}: {e}" ) return {"record": {"add": [], "update": []}} diff --git a/src/memos/mem_feedback/utils.py b/src/memos/mem_feedback/utils.py index 8cb7f97a3..8e3b2f34c 100644 --- a/src/memos/mem_feedback/utils.py +++ b/src/memos/mem_feedback/utils.py @@ -1,3 +1,6 @@ +import json +import re + from memos.memories.textual.item import TextualMemoryItem, TreeNodeTextualMemoryMetadata @@ -147,3 +150,81 @@ def make_mem_item(text: str, **kwargs) -> TextualMemoryItem: info=info_, ), ) + + +def extract_bracket_content(text): + """ + Extract and parse JSON content enclosed in curly braces {} from text. + """ + # Strategy 1: Greedy match to capture the outermost complete brace pair + greedy_match = re.search(r"\{.*\}", text, re.DOTALL) + if greedy_match is None: + error_msg = f"No curly brace content found in text: {text}" + raise ValueError(error_msg) + + greedy_content = greedy_match.group(0) + + # Strategy 2: Non-greedy match to find all brace pairs, use the last one + non_greedy_matches = re.findall(r"\{.*?\}", text, re.DOTALL) + if not non_greedy_matches: + error_msg = f"No curly brace content found in text: {text}" + raise ValueError(error_msg) + + non_greedy_content = non_greedy_matches[-1] + + for content in [greedy_content, non_greedy_content]: + try: + parsed_data = json.loads(content) + return parsed_data + except json.JSONDecodeError: + continue + + for content in [greedy_content, non_greedy_content]: + try: + fixed_content = content.replace("{{", "{").replace("}}", "}") + parsed_data = json.loads(fixed_content) + return parsed_data + except json.JSONDecodeError: + continue + + error_msg = f"Failed to parse JSON content from curly braces. Text preview: {text}" + raise ValueError(error_msg) + + +def extract_square_brackets_content(text): + """ + Extract and parse JSON content enclosed in square brackets [] from text. + """ + # Strategy 1: Greedy match to capture the outermost complete bracket pair + greedy_match = re.search(r"\[.*\]", text, re.DOTALL) + if greedy_match is None: + error_msg = f"No square bracket content found in text: {text}" + raise ValueError(error_msg) + + greedy_content = greedy_match.group(0) + + # Strategy 2: Non-greedy match to find all bracket pairs, use the last one + non_greedy_matches = re.findall(r"\[.*?\]", text, re.DOTALL) + if not non_greedy_matches: + error_msg = f"No square bracket content found in text: {text}" + raise ValueError(error_msg) + + non_greedy_content = non_greedy_matches[-1] + + for content in [greedy_content, non_greedy_content]: + try: + parsed_data = json.loads(content) + return parsed_data + except json.JSONDecodeError: + continue + + for content in [greedy_content, non_greedy_content]: + try: + fixed_content = content.replace("{{", "{").replace("}}", "}") + parsed_data = json.loads(fixed_content) + return parsed_data + except json.JSONDecodeError: + continue + + error_msg = f"Failed to parse JSON content from square brackets. Text preview: {text}" + raise ValueError(error_msg) From 48d4ff1c51c5ab6eaf2353544ae47538f9608079 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Thu, 8 Jan 2026 14:34:14 +0800 Subject: [PATCH 65/72] knowledge raw_text replace memory --- src/memos/api/handlers/formatters_handler.py | 104 ++++++++++++++++++- src/memos/api/handlers/memory_handler.py | 2 +- src/memos/api/handlers/search_handler.py | 15 +++ src/memos/reranker/concat.py | 12 ++- src/memos/reranker/http_bge.py | 13 +-- 5 files changed, 135 insertions(+), 11 deletions(-) diff --git a/src/memos/api/handlers/formatters_handler.py b/src/memos/api/handlers/formatters_handler.py index 94988295b..ca87d95d2 100644 --- a/src/memos/api/handlers/formatters_handler.py +++ b/src/memos/api/handlers/formatters_handler.py @@ -7,9 +7,13 @@ from typing import Any +from memos.log import get_logger from memos.templates.instruction_completion import instruct_completion +logger = get_logger(__name__) + + def to_iter(running: Any) -> list[Any]: """ Normalize running tasks to a list of task objects. @@ -29,7 +33,9 @@ def to_iter(running: Any) -> list[Any]: return list(running) if running else [] -def format_memory_item(memory_data: Any, include_embedding: bool = False) -> dict[str, Any]: +def format_memory_item( + memory_data: Any, include_embedding: bool = False, save_sources: bool = True +) -> dict[str, Any]: """ Format a single memory item for API response. @@ -49,7 +55,8 @@ def format_memory_item(memory_data: Any, include_embedding: bool = False) -> dic memory["ref_id"] = ref_id if not include_embedding: memory["metadata"]["embedding"] = [] - memory["metadata"]["sources"] = [] + if not save_sources: + memory["metadata"]["sources"] = [] memory["metadata"]["usage"] = [] memory["metadata"]["ref_id"] = ref_id memory["metadata"]["id"] = memory_id @@ -125,3 +132,96 @@ def post_process_textual_mem( } ) return memories_result + + +def separate_knowledge_and_conversation_mem(memories: list[dict[str, Any]]): + """ + Separate knowledge and conversation memories from retrieval results. + """ + knowledge_mem = [] + conversation_mem = [] + for item in memories: + sources = item["metadata"]["sources"] + if ( + len(sources) > 0 + and "type" in sources[0] + and sources[0]["type"] == "file" + and "content" in sources[0] + and sources[0]["content"] != "" + ): # TODO change to memory_type + knowledge_mem.append(item) + else: + conversation_mem.append(item) + + logger.info( + f"Retrieval results number of knowledge_mem: {len(knowledge_mem)}, conversation_mem: {len(conversation_mem)}" + ) + return knowledge_mem, conversation_mem + + +def rerank_knowledge_mem( + reranker: Any, + query: str, + text_mem: list[dict[str, Any]], + top_k: int, + file_mem_proportion: float = 0.5, +) -> list[dict[str, Any]]: + """ + Rerank knowledge memories and keep conversation memories. + """ + memid2cubeid = {} + memories_list = [] + for memory_group in text_mem: + cube_id = memory_group["cube_id"] + memories = memory_group["memories"] + memories_list.extend(memories) + for memory in memories: + memid2cubeid[memory["id"]] = cube_id + + knowledge_mem, conversation_mem = separate_knowledge_and_conversation_mem(memories_list) + knowledge_mem_top_k = max(int(top_k * file_mem_proportion), int(top_k - len(conversation_mem))) + reranked_knowledge_mem = reranker.rerank(query, knowledge_mem, top_k=len(knowledge_mem)) + reranked_knowledge_mem = [item[0] for item in reranked_knowledge_mem] + + # TODO revoke sources replace memory value + for item in reranked_knowledge_mem: + item["memory"] = item["metadata"]["sources"][0]["content"] + item["metadata"]["sources"] = [] + + for item in conversation_mem: + item["metadata"]["sources"] = [] + + # deduplicate: remove items with duplicate memory content + original_count = len(reranked_knowledge_mem) + seen_memories = set[Any]() + deduplicated_knowledge_mem = [] + for item in reranked_knowledge_mem: + memory_content = item.get("memory", "") + if memory_content and memory_content not in seen_memories: + seen_memories.add(memory_content) + deduplicated_knowledge_mem.append(item) + deduplicated_count = len(deduplicated_knowledge_mem) + logger.info( + f"After filtering duplicate knowledge base text from sources, count changed from {original_count} to {deduplicated_count}" + ) + + reranked_knowledge_mem = deduplicated_knowledge_mem[:knowledge_mem_top_k] + conversation_mem_top_k = top_k - len(reranked_knowledge_mem) + cubeid2memories = {} + text_mem_res = [] + + for memory in reranked_knowledge_mem + conversation_mem[:conversation_mem_top_k]: + cube_id = memid2cubeid[memory["id"]] + if cube_id not in cubeid2memories: + cubeid2memories[cube_id] = [] + cubeid2memories[cube_id].append(memory) + + for cube_id, memories in cubeid2memories.items(): + text_mem_res.append( + { + "cube_id": cube_id, + "memories": memories, + } + ) + + return text_mem_res diff --git a/src/memos/api/handlers/memory_handler.py b/src/memos/api/handlers/memory_handler.py index ef829d757..14bb8eec5 100644 --- a/src/memos/api/handlers/memory_handler.py +++ b/src/memos/api/handlers/memory_handler.py @@ -204,7 +204,7 @@ def handle_get_memories( preferences, total_pref = naive_mem_cube.pref_mem.get_memory_by_filter( filter_params, page=get_mem_req.page, page_size=get_mem_req.page_size ) - format_preferences = [format_memory_item(item) for item in preferences] + format_preferences = [format_memory_item(item, save_sources=False) for item in preferences] return GetMemoryResponse( message="Memories retrieved successfully", diff --git a/src/memos/api/handlers/search_handler.py b/src/memos/api/handlers/search_handler.py index 3774410dc..32a970b22 100644 --- a/src/memos/api/handlers/search_handler.py +++ b/src/memos/api/handlers/search_handler.py @@ -5,9 +5,12 @@ using dependency injection for better modularity and testability. """ +import time + from typing import Any from memos.api.handlers.base_handler import BaseHandler, HandlerDependencies +from memos.api.handlers.formatters_handler import rerank_knowledge_mem from memos.api.product_models import APISearchRequest, SearchResponse from memos.log import get_logger from memos.memories.textual.tree_text_memory.retrieve.retrieve_utils import ( @@ -69,6 +72,18 @@ def handle_search_memories(self, search_req: APISearchRequest) -> SearchResponse # Restore original top_k for downstream logic or response metadata search_req.top_k = original_top_k + start_time = time.time() + text_mem = results["text_mem"] + results["text_mem"] = rerank_knowledge_mem( + self.reranker, + query=search_req.query, + text_mem=text_mem, + top_k=original_top_k, + file_mem_proportion=0.5, + ) + rerank_time = time.time() - start_time + self.logger.info(f"[Knowledge_replace_memory_time] Rerank time: {rerank_time} seconds") + self.logger.info( f"[SearchHandler] Final search results: count={len(results)} results={results}" ) diff --git a/src/memos/reranker/concat.py b/src/memos/reranker/concat.py index 502af18b6..b39496a1c 100644 --- a/src/memos/reranker/concat.py +++ b/src/memos/reranker/concat.py @@ -83,10 +83,18 @@ def concat_original_source( merge_field = ["sources"] if rerank_source is None else rerank_source.split(",") documents = [] for item in graph_results: - memory = _TAG1.sub("", m) if isinstance((m := getattr(item, "memory", None)), str) else m + m = item.get("memory") if isinstance(item, dict) else getattr(item, "memory", None) + + memory = _TAG1.sub("", m) if isinstance(m, str) else m + sources = [] for field in merge_field: - source = getattr(item.metadata, field, None) + if isinstance(item, dict): + metadata = item.get("metadata", {}) + source = metadata.get(field) if isinstance(metadata, dict) else None + else: + source = getattr(item.metadata, field, None) if hasattr(item, "metadata") else None + if source is None: continue sources.append((memory, source)) diff --git a/src/memos/reranker/http_bge.py b/src/memos/reranker/http_bge.py index 4e9054f1e..dba0006b6 100644 --- a/src/memos/reranker/http_bge.py +++ b/src/memos/reranker/http_bge.py @@ -129,7 +129,7 @@ def __init__( def rerank( self, query: str, - graph_results: list[TextualMemoryItem], + graph_results: list[TextualMemoryItem] | list[dict[str, Any]], top_k: int, search_priority: dict | None = None, **kwargs, @@ -164,11 +164,12 @@ def rerank( if self.rerank_source: documents = concat_original_source(graph_results, self.rerank_source) else: - documents = [ - (_TAG1.sub("", m) if isinstance((m := getattr(item, "memory", None)), str) else m) - for item in graph_results - ] - documents = [d for d in documents if isinstance(d, str) and d] + documents = [] + for item in graph_results: + m = item.get("memory") if isinstance(item, dict) else getattr(item, "memory", None) + + if isinstance(m, str) and m: + documents.append(_TAG1.sub("", m)) logger.info(f"[HTTPBGERerankerSample] query: {query} , documents: {documents[:5]}...") From 21c12cd1ce67d097c836561fcd6df5253088d77e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Thu, 8 Jan 2026 14:48:54 +0800 Subject: [PATCH 66/72] knowledge raw_text replace memory --- 1 | 6 ++++++ src/memos/reranker/http_bge.py | 3 +++ 2 files changed, 9 insertions(+) create mode 100644 1 diff --git a/1 b/1 new file mode 100644 index 000000000..92688c950 --- /dev/null +++ b/1 @@ -0,0 +1,6 @@ +Merge branch 'dev' of github.com:MemTensor/MemOS into dev +# Please enter a commit message to explain why this merge is necessary, +# especially if it merges an updated upstream into a topic branch. +# +# Lines starting with '#' will be ignored, and an empty message aborts +# the commit. diff --git a/src/memos/reranker/http_bge.py b/src/memos/reranker/http_bge.py index dba0006b6..32034cf6d 100644 --- a/src/memos/reranker/http_bge.py +++ b/src/memos/reranker/http_bge.py @@ -165,11 +165,14 @@ def rerank( documents = concat_original_source(graph_results, self.rerank_source) else: documents = [] + filtered_graph_results = [] for item in graph_results: m = item.get("memory") if isinstance(item, dict) else getattr(item, "memory", None) if isinstance(m, str) and m: documents.append(_TAG1.sub("", m)) + filtered_graph_results.append(item) + graph_results = filtered_graph_results logger.info(f"[HTTPBGERerankerSample] query: {query} , documents: {documents[:5]}...") From 7164b715d0f5e53166dd0e939f5ffff0633272fc Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Thu, 8 Jan 2026 15:42:37 +0800 Subject: [PATCH 67/72] knowledge raw_text replace memory --- 1 | 6 ------ 1 file changed, 6 deletions(-) delete mode 100644 1 diff --git a/1 b/1 deleted file mode 100644 index 92688c950..000000000 --- a/1 +++ /dev/null @@ -1,6 +0,0 @@ -Merge branch 'dev' of github.com:MemTensor/MemOS into dev -# Please enter a commit message to explain why this merge is necessary, -# especially if it merges an updated upstream into a topic branch. -# -# Lines starting with '#' will be ignored, and an empty message aborts -# the commit. From 217465f71d65119c48cce21bd2f7a811f2bf3344 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Mon, 19 Jan 2026 20:55:47 +0800 Subject: [PATCH 68/72] fix: when rerank query is too long --- src/memos/reranker/http_bge.py | 3 ++- src/memos/reranker/http_bge_strategy.py | 2 ++ 2 files changed, 4 insertions(+), 1 deletion(-) diff --git a/src/memos/reranker/http_bge.py b/src/memos/reranker/http_bge.py index 32034cf6d..c11644dc6 100644 --- a/src/memos/reranker/http_bge.py +++ b/src/memos/reranker/http_bge.py @@ -154,7 +154,8 @@ def rerank( list[tuple[TextualMemoryItem, float]] Re-ranked items with scores, sorted descending by score. """ - + if len(query) > 8000: + query = query[:500] + "\n" + query[-500:] if not graph_results: return [] diff --git a/src/memos/reranker/http_bge_strategy.py b/src/memos/reranker/http_bge_strategy.py index b0567698c..951893ef5 100644 --- a/src/memos/reranker/http_bge_strategy.py +++ b/src/memos/reranker/http_bge_strategy.py @@ -149,6 +149,8 @@ def rerank( list[tuple[TextualMemoryItem, float]] Re-ranked items with scores, sorted descending by score. """ + if len(query) > 8000: + query = query[:500] + "\n" + query[-500:] if not graph_results: return [] From 6bf44994e71386675f4bab91793879cf3be95f0d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 20 Jan 2026 10:37:44 +0800 Subject: [PATCH 69/72] fix too long rerank --- src/memos/reranker/http_bge.py | 8 ++++++-- src/memos/reranker/http_bge_strategy.py | 8 ++++++-- 2 files changed, 12 insertions(+), 4 deletions(-) diff --git a/src/memos/reranker/http_bge.py b/src/memos/reranker/http_bge.py index c11644dc6..4a773d188 100644 --- a/src/memos/reranker/http_bge.py +++ b/src/memos/reranker/http_bge.py @@ -80,6 +80,8 @@ def __init__( token: str = "", model: str = "bge-reranker-v2-m3", timeout: int = 10, + max_token_thredhold: int = 8000, + concate_len: int = 500, headers_extra: dict | None = None, rerank_source: str | None = None, boost_weights: dict[str, float] | None = None, @@ -109,6 +111,8 @@ def __init__( self.timeout = timeout self.headers_extra = headers_extra or {} self.rerank_source = rerank_source + self.max_token_thredhold = max_token_thredhold + self.concate_len = concate_len self.boost_weights = ( DEFAULT_BOOST_WEIGHTS.copy() @@ -154,8 +158,8 @@ def rerank( list[tuple[TextualMemoryItem, float]] Re-ranked items with scores, sorted descending by score. """ - if len(query) > 8000: - query = query[:500] + "\n" + query[-500:] + if len(query) > self.max_token_thredhold: + query = query[: self.concate_len] + "\n" + query[-self.concate_len :] if not graph_results: return [] diff --git a/src/memos/reranker/http_bge_strategy.py b/src/memos/reranker/http_bge_strategy.py index 951893ef5..7dbee5f01 100644 --- a/src/memos/reranker/http_bge_strategy.py +++ b/src/memos/reranker/http_bge_strategy.py @@ -80,6 +80,8 @@ def __init__( token: str = "", model: str = "bge-reranker-v2-m3", timeout: int = 10, + max_token_thredhold: int = 8000, + concate_len: int = 500, headers_extra: dict | None = None, rerank_source: str | None = None, boost_weights: dict[str, float] | None = None, @@ -109,6 +111,8 @@ def __init__( self.model = model self.timeout = timeout self.headers_extra = headers_extra or {} + self.max_token_thredhold = max_token_thredhold + self.concate_len = concate_len self.boost_weights = ( DEFAULT_BOOST_WEIGHTS.copy() @@ -149,8 +153,8 @@ def rerank( list[tuple[TextualMemoryItem, float]] Re-ranked items with scores, sorted descending by score. """ - if len(query) > 8000: - query = query[:500] + "\n" + query[-500:] + if len(query) > self.max_token_thredhold: + query = query[: self.concate_len] + "\n" + query[-self.concate_len :] if not graph_results: return [] From b5c286ecfbdcbf4658827f0ec0f30a1dd1f602d1 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 20 Jan 2026 10:39:39 +0800 Subject: [PATCH 70/72] fix too long rerank --- src/memos/reranker/http_bge.py | 4 ++-- src/memos/reranker/http_bge_strategy.py | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-) diff --git a/src/memos/reranker/http_bge.py b/src/memos/reranker/http_bge.py index 4a773d188..2e176f0ab 100644 --- a/src/memos/reranker/http_bge.py +++ b/src/memos/reranker/http_bge.py @@ -111,8 +111,8 @@ def __init__( self.timeout = timeout self.headers_extra = headers_extra or {} self.rerank_source = rerank_source - self.max_token_thredhold = max_token_thredhold - self.concate_len = concate_len + self.max_token_thredhold = min(max_token_thredhold, 8000) + self.concate_len = min(concate_len, 500) self.boost_weights = ( DEFAULT_BOOST_WEIGHTS.copy() diff --git a/src/memos/reranker/http_bge_strategy.py b/src/memos/reranker/http_bge_strategy.py index 7dbee5f01..7a58aec3f 100644 --- a/src/memos/reranker/http_bge_strategy.py +++ b/src/memos/reranker/http_bge_strategy.py @@ -111,8 +111,8 @@ def __init__( self.model = model self.timeout = timeout self.headers_extra = headers_extra or {} - self.max_token_thredhold = max_token_thredhold - self.concate_len = concate_len + self.max_token_thredhold = min(max_token_thredhold, 8000) + self.concate_len = min(concate_len, 500) self.boost_weights = ( DEFAULT_BOOST_WEIGHTS.copy() From 4f37c361b37d458dc63520048b9b8b9a3aedfda0 Mon Sep 17 00:00:00 2001 From: CaralHsi Date: Tue, 20 Jan 2026 11:29:51 +0800 Subject: [PATCH 71/72] feat: refactor & reorganize examples with unified structure and updated demos (#903) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit * fix bugs: try to fix bugs in _submit_web_logs * fix bugs: try to address bugs * fix bugs * refactor: modify examples * revise add operation and fix an unbelievable bug * address the bug issues * the doc file has a format problem which has been fixed in this commit * add a range of new feats for the add operation * address the incompatible issue of local scheduler * feat(scheduler): optimize redis queue consumer group management - Proactively ensure consumer groups exist in '_refresh_stream_keys' for newly discovered streams. - Remove redundant consumer group checks in '_read_new_messages_batch' to improve read performance. - Clean up 'seen_streams' cache when streams are deleted to ensure correct group recreation. - This change reduces unnecessary Redis calls during high-frequency polling. * fix(tests): resolve AttributeError in SimpleStructMemReader tests - Import 'parse_json_result' from 'memos.mem_reader.utils' instead of accessing it as an instance attribute. - Fixes 'AttributeError: 'SimpleStructMemReader' object has no attribute 'parse_json_result'' in 'test_parse_json_result_success' and 'test_parse_json_result_failure'. - Remove incorrect mock assignment of 'parse_json_result' in 'test_process_chat_data'. * fix(mem_reader): pass info dict to add_before_search for correct user_id usage - Update 'add_before_search' signature in 'SimpleStructMemReader' to accept 'info' dict. - Pass 'info' (containing 'user_id' and 'session_id') to 'self.searcher.search' instead of using empty strings. - Add 'test_add_before_search' to 'TestSimpleStructMemReader' to verify the fix and ensure 'searcher.search' receives the correct 'info'. - This ensures that memory searches are scoped to the correct user and session. * refactor add_before_search from mem_reader to SingleCubeView * address bugs * fix: fix the qsize bug of task queue, and accept change from hotfix/scheduler * fix: address some issues to run old scheduler example and kv cache example * fix: address the issue of Top-level import of unavailable module 'torch' * fix: resolve linting errors and make optional dependencies lazy loaded - Fix ambiguous characters and commented-out code in examples/mem_scheduler/quick_start_examples.py - Fix nested if statements in src/memos/mem_os/core.py - Move torch and transformers imports to method scope in src/memos/llms/hf.py to support optional dependencies - Update tests/llms/test_hf.py to patch transformers module directly * refactor: revise the rewrite prompt to make it better * refactor: update examples * refactor: update examples for scheduler * fix bugs: address the unsupported xautoclaim command when redis version larger than 6.2.0 via adding a new feature of manul auto claim with the combination of xpending + xclaim * refactor: review settings * refactor: adjust examples to make it run better for code debugging * refactor: review slow add apis to get a better performance on Halumen * fix bugs: address the issue when set user_redis_queue to false, the status_tracker is still using * refactor: allow the code to run without rabbitmq * refactor: create a _parse_pending_entry for redis queue * refactor: add a try/catch for status_tracker * fix: revise examples of textual memories to run normally * feat: improve MemCube examples with View architecture (#859) * feat: improve MemCube examples with View architecture - Add load_cube.py: demonstrate loading MemCube and using SingleCubeView - Add dump_cube.py: demonstrate persisting MemCube to disk - Deprecate old examples: move load_from_folder.py, load_from_remote.py, load_lazily.py to _deprecated/ - Add _deprecated/README.md with migration guide - All code in English with ruff check and format passed - Use 80-char separator lines consistent with server_router_api.py style * feat: improve MemCube examples with SingleCubeView - dump_cube.py: export cube data via graph_store.export_graph() - load_cube.py: import and display graph data - Add example data in examples/data/mem_cube_tree/ --------- Co-authored-by: glin1993@outlook.com <> * fix:examples update for core memories (#863) * fix bugs: try to fix bugs in _submit_web_logs * fix bugs: try to address bugs * fix bugs * refactor: modify examples * revise add operation and fix an unbelievable bug * address the bug issues * the doc file has a format problem which has been fixed in this commit * add a range of new feats for the add operation * address the incompatible issue of local scheduler * feat(scheduler): optimize redis queue consumer group management - Proactively ensure consumer groups exist in '_refresh_stream_keys' for newly discovered streams. - Remove redundant consumer group checks in '_read_new_messages_batch' to improve read performance. - Clean up 'seen_streams' cache when streams are deleted to ensure correct group recreation. - This change reduces unnecessary Redis calls during high-frequency polling. * fix(tests): resolve AttributeError in SimpleStructMemReader tests - Import 'parse_json_result' from 'memos.mem_reader.utils' instead of accessing it as an instance attribute. - Fixes 'AttributeError: 'SimpleStructMemReader' object has no attribute 'parse_json_result'' in 'test_parse_json_result_success' and 'test_parse_json_result_failure'. - Remove incorrect mock assignment of 'parse_json_result' in 'test_process_chat_data'. * fix(mem_reader): pass info dict to add_before_search for correct user_id usage - Update 'add_before_search' signature in 'SimpleStructMemReader' to accept 'info' dict. - Pass 'info' (containing 'user_id' and 'session_id') to 'self.searcher.search' instead of using empty strings. - Add 'test_add_before_search' to 'TestSimpleStructMemReader' to verify the fix and ensure 'searcher.search' receives the correct 'info'. - This ensures that memory searches are scoped to the correct user and session. * refactor add_before_search from mem_reader to SingleCubeView * address bugs * fix: fix the qsize bug of task queue, and accept change from hotfix/scheduler * fix: address some issues to run old scheduler example and kv cache example * fix: address the issue of Top-level import of unavailable module 'torch' * fix: resolve linting errors and make optional dependencies lazy loaded - Fix ambiguous characters and commented-out code in examples/mem_scheduler/quick_start_examples.py - Fix nested if statements in src/memos/mem_os/core.py - Move torch and transformers imports to method scope in src/memos/llms/hf.py to support optional dependencies - Update tests/llms/test_hf.py to patch transformers module directly * refactor: revise the rewrite prompt to make it better * refactor: update examples * refactor: update examples for scheduler * fix bugs: address the unsupported xautoclaim command when redis version larger than 6.2.0 via adding a new feature of manul auto claim with the combination of xpending + xclaim * refactor: review settings * refactor: adjust examples to make it run better for code debugging * refactor: review slow add apis to get a better performance on Halumen * fix bugs: address the issue when set user_redis_queue to false, the status_tracker is still using * refactor: allow the code to run without rabbitmq * refactor: create a _parse_pending_entry for redis queue * refactor: add a try/catch for status_tracker * fix: revise examples of textual memories to run normally * fix:address the issues in existing examples of core memories * refactor: remove deprecated scheduler examples, and revise half examples * refactor: address all issues in examples * fix:feedback mute pref (#876) * feedback mute pref * mute pref feedback --------- Co-authored-by: 黑布林 <11641432+heiheiyouyou@user.noreply.gitee.com> * examples update from tree_textual_memories and preference_textual_mem… (#872) examples update from tree_textual_memories and preference_textual_memories * feat(examples): update basic modules examples and configs (#856) * feat(examples): update basic modules examples and configs * feat: update MCP examples and internet search * docs: api reference --------- Co-authored-by: nova Co-authored-by: CaralHsi * Examples update (#887) * neo4j_example.py * neo4j_example.py * neo4j_example.py * Examples memfeedback (#885) * update example of feedback * update example of mem_feedback --------- Co-authored-by: 周函数 * fix:examples update for mem_reader (#866) * fix:examples update for mem_reader * update readme of the examples of mem_reader --------- Co-authored-by: 周函数 * update example of mem_chat (#892) Co-authored-by: 周函数 * fix: example config bug * docs: Delete deprecated examples. * docs: Adding clear instructions as comments for examples in basic_modules. * fix: fixing bugs relevant to examples in basic_module. * doc: Adding clear instructions as comments for examples in mem_scheduler. * fix: Fix example bug caused by using the default SchedulerLocalQueue(without .show_task_status method). * revert: "doc: Examples update and related minor bug fix" (#899) Revert "doc: Examples update and related minor bug fix" * doc: Examples update and related minor bug fix (#900) Revert "revert: "doc: Examples update and related minor bug fix" (#899)" The test failed because of other reasons than commits in this pr, so we revert this revert. This reverts commit d0f3bab577786252edc206f17707487d05e0b898. Co-authored-by: CaralHsi * feat: delete useless example (#904) * fix: fix api examples and add search/chat examples. (#906) * feat: delete useless file; fix: chat model list bug (#907) * feat: delete useless example * fix: chat_model_list bug * feat: delete useless file * feat: delete useless test file (#908) * feat: delete useless example * fix: chat_model_list bug * feat: delete useless file * feat: delete useless test * feat: add mem_cube_2 because tests need them (#909) * feat: delete useless example * fix: chat_model_list bug * feat: delete useless file * feat: delete useless test * feat: add mem_cube_2 because tests need them * feat: add back mem_cube because tests use them (#910) * feat: delete useless example * fix: chat_model_list bug * feat: delete useless file * feat: delete useless test * feat: add mem_cube_2 because tests need them * feat: add mem_cube_2 because tests need them --------- Co-authored-by: chentang Co-authored-by: Zehao Lin Co-authored-by: Dubberman <48425266+whipser030@users.noreply.github.com> Co-authored-by: 黑布林 <11641432+heiheiyouyou@user.noreply.gitee.com> Co-authored-by: Wenqiang Wei <46308778+endxxxx@users.noreply.github.com> Co-authored-by: novaleolin <15062791135@163.com> Co-authored-by: nova Co-authored-by: caocuilong <39516215+caocuilong@users.noreply.github.com> Co-authored-by: HanshuZhou <118414084+zhouhanshu@users.noreply.github.com> Co-authored-by: 周函数 Co-authored-by: Wenqiang Wei Co-authored-by: bittergreen --- docs/openapi.json | 3596 +++++++++++--- examples/api/product_api.py | 170 - examples/api/server_router_api.py | 276 +- examples/basic_modules/embedder.py | 14 +- examples/basic_modules/llm.py | 8 + examples/basic_modules/nebular_example.py | 361 -- examples/basic_modules/neo4j_example.py | 73 + examples/basic_modules/parser.py | 15 - examples/basic_modules/qdrant_vec_db.py | 143 - examples/basic_modules/reranker.py | 17 + .../textual_memory_internet_search_example.py | 307 +- .../tree_textual_memory_reasoner.py | 167 - .../tree_textual_memory_recall.py | 203 +- ...textual_memory_relation_reason_detector.py | 272 +- .../tree_textual_memory_task_goal_parser.py | 205 +- .../core_memories/general_textual_memory.py | 61 +- examples/core_memories/kv_cache_memory.py | 186 +- .../core_memories/naive_textual_memory.py | 79 +- ...ternet_memoy.py => pref_textual_memory.py} | 111 +- examples/core_memories/tree_textual_memory.py | 28 + .../core_memories/vllm_kv_cache_memory.py | 2 +- examples/data/config/preference_config.json | 53 + .../config/simple_struct_reader_config.json | 22 +- .../config/tree_config_shared_database.json | 59 +- .../data/mem_cube_tree/textual_memory.json | 4219 +++++++++++++++++ examples/data/mem_scheduler/scene_data.json | 118 - examples/mem_api/pipeline_test.py | 178 - ...t_w_generated_cube_explicit_memory_only.py | 231 +- examples/mem_chat/simple_mem_chat.py | 33 - examples/mem_chunk/markdown_chunk.py | 33 - examples/mem_cube/_deprecated/README.md | 81 + .../{ => _deprecated}/load_from_folder.py | 0 .../{ => _deprecated}/load_from_remote.py | 0 .../mem_cube/{ => _deprecated}/load_lazily.py | 0 examples/mem_cube/dump_cube.py | 137 + examples/mem_cube/load_cube.py | 137 + examples/mem_feedback/example_feedback.py | 248 + examples/mem_mcp/simple_fastmcp_client.py | 69 +- examples/mem_mcp/simple_fastmcp_serve.py | 77 +- .../chat_w_generated_cube_explicit_memory.py | 180 - examples/mem_os/chat_w_scheduler.py | 74 - examples/mem_os/cot_usage_example.py | 342 -- examples/mem_os/easy_memos.py | 21 - .../mem_os/locomo_shared_database_memos.py | 203 - examples/mem_os/multi_user_memos_example.py | 162 - examples/mem_os/naive_memos.py | 30 - examples/mem_os/persistent_memos_example.py | 192 - examples/mem_os/simple_memos.py | 42 - examples/mem_os/simple_openapi_memos.py | 300 -- .../simple_openapi_memos_neo4j_community.py | 315 -- examples/mem_os/simple_prefs_memos_product.py | 399 -- examples/mem_os/simple_treekvcache_memos.py | 46 - examples/mem_os/simple_vllm_memos.py | 61 - examples/mem_reader/README.md | 90 + examples/mem_reader/builders.py | 78 + .../compare_simple_vs_multimodal.py | 461 -- .../mem_reader/multimodal_struct_reader.py | 1042 ---- examples/mem_reader/parser/__init__.py | 1 - examples/mem_reader/parser/config_utils.py | 132 - .../parser/example_assistant_parser.py | 94 - .../parser/example_file_content_parser.py | 132 - .../parser/example_multi_modal_parser.py | 400 -- .../parser/example_string_parser.py | 66 - .../parser/example_system_parser.py | 158 - .../parser/example_text_content_parser.py | 72 - .../mem_reader/parser/example_tool_parser.py | 101 - .../mem_reader/parser/example_user_parser.py | 135 - examples/mem_reader/parser/print_utils.py | 11 - examples/mem_reader/parser_demos/__init__.py | 0 examples/mem_reader/parser_demos/_base.py | 76 + .../mem_reader/parser_demos/demo_assistant.py | 44 + .../parser_demos/demo_file_content.py | 68 + .../mem_reader/parser_demos/demo_image.py | 87 + .../parser_demos/demo_multi_modal.py | 346 ++ .../mem_reader/parser_demos/demo_string.py | 30 + .../mem_reader/parser_demos/demo_system.py | 41 + .../parser_demos/demo_text_content.py | 40 + examples/mem_reader/parser_demos/demo_tool.py | 28 + examples/mem_reader/parser_demos/demo_user.py | 43 + examples/mem_reader/runners/__init__.py | 0 examples/mem_reader/runners/run_multimodal.py | 139 + examples/mem_reader/runners/run_simple.py | 113 + examples/mem_reader/samples.py | 773 +++ examples/mem_reader/settings.py | 107 + examples/mem_reader/simple_struct_reader.py | 831 ---- examples/mem_reader/utils.py | 84 + examples/mem_scheduler/api_w_scheduler.py | 163 +- .../mem_scheduler/debug_text_mem_replace.py | 109 - examples/mem_scheduler/memos_w_scheduler.py | 265 +- .../memos_w_scheduler_for_test.py | 88 - .../mem_scheduler/quick_start_examples.py | 312 -- examples/mem_scheduler/rabbitmq_example.py | 60 - examples/mem_scheduler/redis_example.py | 12 +- ..._for_async_tasks.py => run_async_tasks.py} | 9 + examples/mem_scheduler/show_redis_status.py | 52 +- examples/mem_scheduler/task_fair_schedule.py | 87 - .../mem_user/user_manager_factory_example.py | 111 - src/memos/api/config.py | 2 +- src/memos/api/handlers/config_builders.py | 2 +- src/memos/api/mcp_serve.py | 80 +- src/memos/api/start_api.py | 2 +- src/memos/cli.py | 2 +- src/memos/graph_dbs/neo4j.py | 2 +- src/memos/llms/hf.py | 2 +- src/memos/mem_feedback/feedback.py | 14 +- src/memos/mem_feedback/simple_feedback.py | 2 + .../init_components_for_scheduler.py | 2 +- .../task_schedule_modules/redis_queue.py | 35 +- tests/test_openapi_sync.py | 79 - 109 files changed, 11885 insertions(+), 9786 deletions(-) delete mode 100644 examples/api/product_api.py delete mode 100644 examples/basic_modules/nebular_example.py delete mode 100644 examples/basic_modules/parser.py delete mode 100644 examples/basic_modules/qdrant_vec_db.py delete mode 100644 examples/basic_modules/tree_textual_memory_reasoner.py rename examples/core_memories/{textual_internet_memoy.py => pref_textual_memory.py} (76%) create mode 100644 examples/data/config/preference_config.json create mode 100644 examples/data/mem_cube_tree/textual_memory.json delete mode 100644 examples/data/mem_scheduler/scene_data.json delete mode 100644 examples/mem_api/pipeline_test.py delete mode 100644 examples/mem_chat/simple_mem_chat.py delete mode 100644 examples/mem_chunk/markdown_chunk.py create mode 100644 examples/mem_cube/_deprecated/README.md rename examples/mem_cube/{ => _deprecated}/load_from_folder.py (100%) rename examples/mem_cube/{ => _deprecated}/load_from_remote.py (100%) rename examples/mem_cube/{ => _deprecated}/load_lazily.py (100%) create mode 100644 examples/mem_cube/dump_cube.py create mode 100644 examples/mem_cube/load_cube.py create mode 100644 examples/mem_feedback/example_feedback.py delete mode 100644 examples/mem_os/chat_w_generated_cube_explicit_memory.py delete mode 100644 examples/mem_os/chat_w_scheduler.py delete mode 100644 examples/mem_os/cot_usage_example.py delete mode 100644 examples/mem_os/easy_memos.py delete mode 100644 examples/mem_os/locomo_shared_database_memos.py delete mode 100644 examples/mem_os/multi_user_memos_example.py delete mode 100644 examples/mem_os/naive_memos.py delete mode 100644 examples/mem_os/persistent_memos_example.py delete mode 100644 examples/mem_os/simple_memos.py delete mode 100644 examples/mem_os/simple_openapi_memos.py delete mode 100644 examples/mem_os/simple_openapi_memos_neo4j_community.py delete mode 100644 examples/mem_os/simple_prefs_memos_product.py delete mode 100644 examples/mem_os/simple_treekvcache_memos.py delete mode 100644 examples/mem_os/simple_vllm_memos.py create mode 100644 examples/mem_reader/README.md create mode 100644 examples/mem_reader/builders.py delete mode 100644 examples/mem_reader/compare_simple_vs_multimodal.py delete mode 100644 examples/mem_reader/multimodal_struct_reader.py delete mode 100644 examples/mem_reader/parser/__init__.py delete mode 100644 examples/mem_reader/parser/config_utils.py delete mode 100644 examples/mem_reader/parser/example_assistant_parser.py delete mode 100644 examples/mem_reader/parser/example_file_content_parser.py delete mode 100644 examples/mem_reader/parser/example_multi_modal_parser.py delete mode 100644 examples/mem_reader/parser/example_string_parser.py delete mode 100644 examples/mem_reader/parser/example_system_parser.py delete mode 100644 examples/mem_reader/parser/example_text_content_parser.py delete mode 100644 examples/mem_reader/parser/example_tool_parser.py delete mode 100644 examples/mem_reader/parser/example_user_parser.py delete mode 100644 examples/mem_reader/parser/print_utils.py create mode 100644 examples/mem_reader/parser_demos/__init__.py create mode 100644 examples/mem_reader/parser_demos/_base.py create mode 100644 examples/mem_reader/parser_demos/demo_assistant.py create mode 100644 examples/mem_reader/parser_demos/demo_file_content.py create mode 100644 examples/mem_reader/parser_demos/demo_image.py create mode 100644 examples/mem_reader/parser_demos/demo_multi_modal.py create mode 100644 examples/mem_reader/parser_demos/demo_string.py create mode 100644 examples/mem_reader/parser_demos/demo_system.py create mode 100644 examples/mem_reader/parser_demos/demo_text_content.py create mode 100644 examples/mem_reader/parser_demos/demo_tool.py create mode 100644 examples/mem_reader/parser_demos/demo_user.py create mode 100644 examples/mem_reader/runners/__init__.py create mode 100644 examples/mem_reader/runners/run_multimodal.py create mode 100644 examples/mem_reader/runners/run_simple.py create mode 100644 examples/mem_reader/samples.py create mode 100644 examples/mem_reader/settings.py delete mode 100644 examples/mem_reader/simple_struct_reader.py create mode 100644 examples/mem_reader/utils.py delete mode 100644 examples/mem_scheduler/debug_text_mem_replace.py delete mode 100644 examples/mem_scheduler/memos_w_scheduler_for_test.py delete mode 100644 examples/mem_scheduler/quick_start_examples.py delete mode 100644 examples/mem_scheduler/rabbitmq_example.py rename examples/mem_scheduler/{scheduler_for_async_tasks.py => run_async_tasks.py} (90%) delete mode 100644 examples/mem_scheduler/task_fair_schedule.py delete mode 100644 examples/mem_user/user_manager_factory_example.py delete mode 100644 tests/test_openapi_sync.py diff --git a/docs/openapi.json b/docs/openapi.json index 46d715147..d9ef710b5 100644 --- a/docs/openapi.json +++ b/docs/openapi.json @@ -1,21 +1,24 @@ { "openapi": "3.1.0", "info": { - "title": "MemOS REST APIs", - "description": "A REST API for managing and searching memories using MemOS.", - "version": "1.0.0" + "title": "MemOS Server REST APIs", + "description": "A REST API for managing multiple users with MemOS Server.", + "version": "1.0.1" }, "paths": { - "/configure": { + "/product/search": { "post": { - "summary": "Configure MemOS", - "description": "Set MemOS configuration.", - "operationId": "set_config_configure_post", + "tags": [ + "Server API" + ], + "summary": "Search memories", + "description": "Search memories for a specific user.\n\nThis endpoint uses the class-based SearchHandler for better code organization.", + "operationId": "search_memories_product_search_post", "requestBody": { "content": { "application/json": { "schema": { - "$ref": "#/components/schemas/MOSConfig" + "$ref": "#/components/schemas/APISearchRequest" } } }, @@ -27,7 +30,7 @@ "content": { "application/json": { "schema": { - "$ref": "#/components/schemas/ConfigResponse" + "$ref": "#/components/schemas/SearchResponse" } } } @@ -45,33 +48,19 @@ } } }, - "/users": { - "get": { - "summary": "List all users", - "description": "List all active users.", - "operationId": "list_users_users_get", - "responses": { - "200": { - "description": "Successful Response", - "content": { - "application/json": { - "schema": { - "$ref": "#/components/schemas/UserListResponse" - } - } - } - } - } - }, + "/product/add": { "post": { - "summary": "Create a new user", - "description": "Create a new user.", - "operationId": "create_user_users_post", + "tags": [ + "Server API" + ], + "summary": "Add memories", + "description": "Add memories for a specific user.\n\nThis endpoint uses the class-based AddHandler for better code organization.", + "operationId": "add_memories_product_add_post", "requestBody": { "content": { "application/json": { "schema": { - "$ref": "#/components/schemas/UserCreate" + "$ref": "#/components/schemas/APIADDRequest" } } }, @@ -83,7 +72,7 @@ "content": { "application/json": { "schema": { - "$ref": "#/components/schemas/UserResponse" + "$ref": "#/components/schemas/MemoryResponse" } } } @@ -101,57 +90,21 @@ } } }, - "/users/me": { + "/product/scheduler/allstatus": { "get": { - "summary": "Get current user info", - "description": "Get current user information including accessible cubes.", - "operationId": "get_user_info_users_me_get", - "responses": { - "200": { - "description": "Successful Response", - "content": { - "application/json": { - "schema": { - "$ref": "#/components/schemas/UserResponse" - } - } - } - } - } - } - }, - "/mem_cubes": { - "post": { - "summary": "Register a MemCube", - "description": "Register a new MemCube.", - "operationId": "register_mem_cube_mem_cubes_post", - "requestBody": { - "content": { - "application/json": { - "schema": { - "$ref": "#/components/schemas/MemCubeRegister" - } - } - }, - "required": true - }, + "tags": [ + "Server API" + ], + "summary": "Get detailed scheduler status", + "description": "Get detailed scheduler status including running tasks and queue metrics.", + "operationId": "scheduler_allstatus_product_scheduler_allstatus_get", "responses": { "200": { "description": "Successful Response", "content": { "application/json": { "schema": { - "$ref": "#/components/schemas/SimpleResponse" - } - } - } - }, - "422": { - "description": "Validation Error", - "content": { - "application/json": { - "schema": { - "$ref": "#/components/schemas/HTTPValidationError" + "$ref": "#/components/schemas/AllStatusResponse" } } } @@ -159,23 +112,28 @@ } } }, - "/mem_cubes/{mem_cube_id}": { - "delete": { - "summary": "Unregister a MemCube", - "description": "Unregister a MemCube.", - "operationId": "unregister_mem_cube_mem_cubes__mem_cube_id__delete", + "/product/scheduler/status": { + "get": { + "tags": [ + "Server API" + ], + "summary": "Get scheduler running status", + "description": "Get scheduler running status.", + "operationId": "scheduler_status_product_scheduler_status_get", "parameters": [ { - "name": "mem_cube_id", - "in": "path", + "name": "user_id", + "in": "query", "required": true, "schema": { "type": "string", - "title": "Mem Cube Id" - } + "description": "User ID", + "title": "User Id" + }, + "description": "User ID" }, { - "name": "user_id", + "name": "task_id", "in": "query", "required": false, "schema": { @@ -187,8 +145,10 @@ "type": "null" } ], - "title": "User Id" - } + "description": "Optional Task ID to query a specific task", + "title": "Task Id" + }, + "description": "Optional Task ID to query a specific task" } ], "responses": { @@ -197,7 +157,7 @@ "content": { "application/json": { "schema": { - "$ref": "#/components/schemas/SimpleResponse" + "$ref": "#/components/schemas/StatusResponse" } } } @@ -215,39 +175,34 @@ } } }, - "/mem_cubes/{cube_id}/share": { - "post": { - "summary": "Share a cube with another user", - "description": "Share a cube with another user.", - "operationId": "share_cube_mem_cubes__cube_id__share_post", + "/product/scheduler/task_queue_status": { + "get": { + "tags": [ + "Server API" + ], + "summary": "Get scheduler task queue status", + "description": "Get scheduler task queue backlog/pending status for a user.", + "operationId": "scheduler_task_queue_status_product_scheduler_task_queue_status_get", "parameters": [ { - "name": "cube_id", - "in": "path", + "name": "user_id", + "in": "query", "required": true, "schema": { "type": "string", - "title": "Cube Id" - } + "description": "User ID whose queue status is requested", + "title": "User Id" + }, + "description": "User ID whose queue status is requested" } ], - "requestBody": { - "required": true, - "content": { - "application/json": { - "schema": { - "$ref": "#/components/schemas/CubeShare" - } - } - } - }, "responses": { "200": { "description": "Successful Response", "content": { "application/json": { "schema": { - "$ref": "#/components/schemas/SimpleResponse" + "$ref": "#/components/schemas/TaskQueueResponse" } } } @@ -265,79 +220,42 @@ } } }, - "/memories": { + "/product/scheduler/wait": { "post": { - "summary": "Create memories", - "description": "Store new memories in a MemCube.", - "operationId": "add_memory_memories_post", - "requestBody": { - "required": true, - "content": { - "application/json": { - "schema": { - "$ref": "#/components/schemas/MemoryCreate" - } - } - } - }, - "responses": { - "200": { - "description": "Successful Response", - "content": { - "application/json": { - "schema": { - "$ref": "#/components/schemas/SimpleResponse" - } - } + "tags": [ + "Server API" + ], + "summary": "Wait until scheduler is idle for a specific user", + "description": "Wait until scheduler is idle for a specific user.", + "operationId": "scheduler_wait_product_scheduler_wait_post", + "parameters": [ + { + "name": "user_name", + "in": "query", + "required": true, + "schema": { + "type": "string", + "title": "User Name" } }, - "422": { - "description": "Validation Error", - "content": { - "application/json": { - "schema": { - "$ref": "#/components/schemas/HTTPValidationError" - } - } - } - } - } - }, - "get": { - "summary": "Get all memories", - "description": "Retrieve all memories from a MemCube.", - "operationId": "get_all_memories_memories_get", - "parameters": [ { - "name": "mem_cube_id", + "name": "timeout_seconds", "in": "query", "required": false, "schema": { - "anyOf": [ - { - "type": "string" - }, - { - "type": "null" - } - ], - "title": "Mem Cube Id" + "type": "number", + "default": 120.0, + "title": "Timeout Seconds" } }, { - "name": "user_id", + "name": "poll_interval", "in": "query", "required": false, "schema": { - "anyOf": [ - { - "type": "string" - }, - { - "type": "null" - } - ], - "title": "User Id" + "type": "number", + "default": 0.5, + "title": "Poll Interval" } } ], @@ -346,9 +264,7 @@ "description": "Successful Response", "content": { "application/json": { - "schema": { - "$ref": "#/components/schemas/MemoryResponse" - } + "schema": {} } } }, @@ -365,44 +281,42 @@ } } }, - "/memories/{mem_cube_id}/{memory_id}": { + "/product/scheduler/wait/stream": { "get": { - "summary": "Get a memory", - "description": "Retrieve a specific memory by ID from a MemCube.", - "operationId": "get_memory_memories__mem_cube_id___memory_id__get", + "tags": [ + "Server API" + ], + "summary": "Stream scheduler progress for a user", + "description": "Stream scheduler progress via Server-Sent Events (SSE).", + "operationId": "scheduler_wait_stream_product_scheduler_wait_stream_get", "parameters": [ { - "name": "mem_cube_id", - "in": "path", + "name": "user_name", + "in": "query", "required": true, "schema": { "type": "string", - "title": "Mem Cube Id" + "title": "User Name" } }, { - "name": "memory_id", - "in": "path", - "required": true, + "name": "timeout_seconds", + "in": "query", + "required": false, "schema": { - "type": "string", - "title": "Memory Id" + "type": "number", + "default": 120.0, + "title": "Timeout Seconds" } }, { - "name": "user_id", + "name": "poll_interval", "in": "query", "required": false, "schema": { - "anyOf": [ - { - "type": "string" - }, - { - "type": "null" - } - ], - "title": "User Id" + "type": "number", + "default": 0.5, + "title": "Poll Interval" } } ], @@ -411,9 +325,7 @@ "description": "Successful Response", "content": { "application/json": { - "schema": { - "$ref": "#/components/schemas/MemoryResponse" - } + "schema": {} } } }, @@ -428,67 +340,32 @@ } } } - }, - "put": { - "summary": "Update a memory", - "description": "Update an existing memory in a MemCube.", - "operationId": "update_memory_memories__mem_cube_id___memory_id__put", - "parameters": [ - { - "name": "mem_cube_id", - "in": "path", - "required": true, - "schema": { - "type": "string", - "title": "Mem Cube Id" - } - }, - { - "name": "memory_id", - "in": "path", - "required": true, - "schema": { - "type": "string", - "title": "Memory Id" - } - }, - { - "name": "user_id", - "in": "query", - "required": false, - "schema": { - "anyOf": [ - { - "type": "string" - }, - { - "type": "null" - } - ], - "title": "User Id" - } - } + } + }, + "/product/chat/complete": { + "post": { + "tags": [ + "Server API" ], + "summary": "Chat with MemOS (Complete Response)", + "description": "Chat with MemOS for a specific user. Returns complete response (non-streaming).\n\nThis endpoint uses the class-based ChatHandler.", + "operationId": "chat_complete_product_chat_complete_post", "requestBody": { - "required": true, "content": { "application/json": { "schema": { - "type": "object", - "additionalProperties": true, - "title": "Updated Memory" + "$ref": "#/components/schemas/APIChatCompleteRequest" } } - } + }, + "required": true }, "responses": { "200": { "description": "Successful Response", "content": { "application/json": { - "schema": { - "$ref": "#/components/schemas/SimpleResponse" - } + "schema": {} } } }, @@ -503,55 +380,32 @@ } } } - }, - "delete": { - "summary": "Delete a memory", - "description": "Delete a specific memory from a MemCube.", - "operationId": "delete_memory_memories__mem_cube_id___memory_id__delete", - "parameters": [ - { - "name": "mem_cube_id", - "in": "path", - "required": true, - "schema": { - "type": "string", - "title": "Mem Cube Id" - } - }, - { - "name": "memory_id", - "in": "path", - "required": true, - "schema": { - "type": "string", - "title": "Memory Id" + } + }, + "/product/chat/stream": { + "post": { + "tags": [ + "Server API" + ], + "summary": "Chat with MemOS", + "description": "Chat with MemOS for a specific user. Returns SSE stream.\n\nThis endpoint uses the class-based ChatHandler which internally\ncomposes SearchHandler and AddHandler for a clean architecture.", + "operationId": "chat_stream_product_chat_stream_post", + "requestBody": { + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/ChatRequest" + } } }, - { - "name": "user_id", - "in": "query", - "required": false, - "schema": { - "anyOf": [ - { - "type": "string" - }, - { - "type": "null" - } - ], - "title": "User Id" - } - } - ], + "required": true + }, "responses": { "200": { "description": "Successful Response", "content": { "application/json": { - "schema": { - "$ref": "#/components/schemas/SimpleResponse" - } + "schema": {} } } }, @@ -568,16 +422,19 @@ } } }, - "/search": { + "/product/chat/stream/playground": { "post": { - "summary": "Search memories", - "description": "Search for memories across MemCubes.", - "operationId": "search_memories_search_post", + "tags": [ + "Server API" + ], + "summary": "Chat with MemOS playground", + "description": "Chat with MemOS for a specific user. Returns SSE stream.\n\nThis endpoint uses the class-based ChatHandler which internally\ncomposes SearchHandler and AddHandler for a clean architecture.", + "operationId": "chat_stream_playground_product_chat_stream_playground_post", "requestBody": { "content": { "application/json": { "schema": { - "$ref": "#/components/schemas/SearchRequest" + "$ref": "#/components/schemas/ChatPlaygroundRequest" } } }, @@ -588,8 +445,131 @@ "description": "Successful Response", "content": { "application/json": { - "schema": { - "$ref": "#/components/schemas/SearchResponse" + "schema": {} + } + } + }, + "422": { + "description": "Validation Error", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/HTTPValidationError" + } + } + } + } + } + } + }, + "/product/suggestions": { + "post": { + "tags": [ + "Server API" + ], + "summary": "Get suggestion queries", + "description": "Get suggestion queries for a specific user with language preference.", + "operationId": "get_suggestion_queries_product_suggestions_post", + "requestBody": { + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/SuggestionRequest" + } + } + }, + "required": true + }, + "responses": { + "200": { + "description": "Successful Response", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/SuggestionResponse" + } + } + } + }, + "422": { + "description": "Validation Error", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/HTTPValidationError" + } + } + } + } + } + } + }, + "/product/get_all": { + "post": { + "tags": [ + "Server API" + ], + "summary": "Get all memories for user", + "description": "Get all memories or subgraph for a specific user.\n\nIf search_query is provided, returns a subgraph based on the query.\nOtherwise, returns all memories of the specified type.", + "operationId": "get_all_memories_product_get_all_post", + "requestBody": { + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/GetMemoryPlaygroundRequest" + } + } + }, + "required": true + }, + "responses": { + "200": { + "description": "Successful Response", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/MemoryResponse" + } + } + } + }, + "422": { + "description": "Validation Error", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/HTTPValidationError" + } + } + } + } + } + } + }, + "/product/get_memory": { + "post": { + "tags": [ + "Server API" + ], + "summary": "Get memories for user", + "operationId": "get_memories_product_get_memory_post", + "requestBody": { + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/GetMemoryRequest" + } + } + }, + "required": true + }, + "responses": { + "200": { + "description": "Successful Response", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/GetMemoryResponse" } } } @@ -607,107 +587,2013 @@ } } }, - "/memories/{mem_cube_id}": { - "delete": { - "summary": "Delete all memories", - "description": "Delete all memories from a MemCube.", - "operationId": "delete_all_memories_memories__mem_cube_id__delete", + "/product/get_memory/{memory_id}": { + "get": { + "tags": [ + "Server API" + ], + "summary": "Get memory by id", + "operationId": "get_memory_by_id_product_get_memory__memory_id__get", "parameters": [ { - "name": "mem_cube_id", + "name": "memory_id", "in": "path", "required": true, "schema": { "type": "string", - "title": "Mem Cube Id" + "title": "Memory Id" + } + } + ], + "responses": { + "200": { + "description": "Successful Response", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/GetMemoryResponse" + } + } } }, - { - "name": "user_id", - "in": "query", - "required": false, - "schema": { - "anyOf": [ - { - "type": "string" - }, - { - "type": "null" + "422": { + "description": "Validation Error", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/HTTPValidationError" } - ], - "title": "User Id" + } + } + } + } + } + }, + "/product/delete_memory": { + "post": { + "tags": [ + "Server API" + ], + "summary": "Delete memories for user", + "operationId": "delete_memories_product_delete_memory_post", + "requestBody": { + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/DeleteMemoryRequest" + } + } + }, + "required": true + }, + "responses": { + "200": { + "description": "Successful Response", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/DeleteMemoryResponse" + } + } + } + }, + "422": { + "description": "Validation Error", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/HTTPValidationError" + } + } + } + } + } + } + }, + "/product/feedback": { + "post": { + "tags": [ + "Server API" + ], + "summary": "Feedback memories", + "description": "Feedback memories for a specific user.\n\nThis endpoint uses the class-based FeedbackHandler for better code organization.", + "operationId": "feedback_memories_product_feedback_post", + "requestBody": { + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/APIFeedbackRequest" + } + } + }, + "required": true + }, + "responses": { + "200": { + "description": "Successful Response", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/MemoryResponse" + } + } + } + }, + "422": { + "description": "Validation Error", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/HTTPValidationError" + } + } + } + } + } + } + }, + "/product/get_user_names_by_memory_ids": { + "post": { + "tags": [ + "Server API" + ], + "summary": "Get user names by memory ids", + "description": "Get user names by memory ids.", + "operationId": "get_user_names_by_memory_ids_product_get_user_names_by_memory_ids_post", + "requestBody": { + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/GetUserNamesByMemoryIdsRequest" + } + } + }, + "required": true + }, + "responses": { + "200": { + "description": "Successful Response", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/GetUserNamesByMemoryIdsResponse" + } + } + } + }, + "422": { + "description": "Validation Error", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/HTTPValidationError" + } + } } } + } + } + }, + "/product/exist_mem_cube_id": { + "post": { + "tags": [ + "Server API" ], + "summary": "Check if mem cube id exists", + "description": "Check if mem cube id exists.", + "operationId": "exist_mem_cube_id_product_exist_mem_cube_id_post", + "requestBody": { + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/ExistMemCubeIdRequest" + } + } + }, + "required": true + }, "responses": { "200": { "description": "Successful Response", "content": { "application/json": { "schema": { - "$ref": "#/components/schemas/SimpleResponse" + "$ref": "#/components/schemas/ExistMemCubeIdResponse" } } - } + } + }, + "422": { + "description": "Validation Error", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/HTTPValidationError" + } + } + } + } + } + } + } + }, + "components": { + "schemas": { + "APIADDRequest": { + "properties": { + "user_id": { + "type": "string", + "title": "User Id", + "description": "User ID" + }, + "session_id": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Session Id", + "description": "Session ID. If not provided, a default session will be used." + }, + "task_id": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Task Id", + "description": "Task ID for monitering async tasks" + }, + "writable_cube_ids": { + "anyOf": [ + { + "items": { + "type": "string" + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "Writable Cube Ids", + "description": "List of cube IDs user can write for multi-cube add" + }, + "async_mode": { + "type": "string", + "enum": [ + "async", + "sync" + ], + "title": "Async Mode", + "description": "Whether to add memory in async mode. Use 'async' to enqueue background add (non-blocking), or 'sync' to add memories in the current call. Default: 'async'.", + "default": "async" + }, + "mode": { + "anyOf": [ + { + "type": "string", + "enum": [ + "fast", + "fine" + ] + }, + { + "type": "null" + } + ], + "title": "Mode", + "description": "(Internal) Add mode used only when async_mode='sync'. If set to 'fast', the handler will use a fast add pipeline. Ignored when async_mode='async'." + }, + "custom_tags": { + "anyOf": [ + { + "items": { + "type": "string" + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "Custom Tags", + "description": "Custom tags for this add request, e.g. ['Travel', 'family']. These tags can be used as filters in search." + }, + "info": { + "anyOf": [ + { + "additionalProperties": true, + "type": "object" + }, + { + "type": "null" + } + ], + "title": "Info", + "description": "Additional metadata for the add request. All keys can be used as filters in search. Example: {'agent_id': 'xxxxxx', 'app_id': 'xxxx', 'source_type': 'web', 'source_url': 'https://www.baidu.com', 'source_content': '西湖是杭州最著名的景点'}." + }, + "messages": { + "anyOf": [ + { + "type": "string" + }, + { + "items": { + "anyOf": [ + { + "$ref": "#/components/schemas/ChatCompletionSystemMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionUserMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionAssistantMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionToolMessageParam" + } + ] + }, + "type": "array" + }, + { + "items": { + "anyOf": [ + { + "$ref": "#/components/schemas/ChatCompletionContentPartTextParam" + }, + { + "$ref": "#/components/schemas/File" + } + ] + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "Messages", + "description": "List of messages to store. Supports: - system / user / assistant messages with 'content' and 'chat_time'; - tool messages including: * tool_description (name, description, parameters), * tool_input (call_id, name, argument), * raw tool messages where content is str or list[str], * tool_output with structured output items (input_text / input_image / input_file, etc.). Also supports pure input items when there is no dialog." + }, + "chat_history": { + "anyOf": [ + { + "items": { + "anyOf": [ + { + "$ref": "#/components/schemas/ChatCompletionSystemMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionUserMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionAssistantMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionToolMessageParam" + } + ] + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "Chat History", + "description": "Historical chat messages used internally by algorithms. If None, internal stored history will be used; if provided (even an empty list), this value will be used as-is." + }, + "is_feedback": { + "type": "boolean", + "title": "Is Feedback", + "description": "Whether this request represents user feedback. Default: False.", + "default": false + }, + "mem_cube_id": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Mem Cube Id", + "description": "(Deprecated) Target cube ID for this add request (optional for developer API)." + }, + "memory_content": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Memory Content", + "description": "(Deprecated) Plain memory content to store. Prefer using `messages`." + }, + "doc_path": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Doc Path", + "description": "(Deprecated / internal) Path to document to store." + }, + "source": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Source", + "description": "(Deprecated) Simple source tag of the memory. Prefer using `info.source_type` / `info.source_url`." + }, + "operation": { + "anyOf": [ + { + "items": { + "$ref": "#/components/schemas/PermissionDict" + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "Operation", + "description": "(Internal) Operation definitions for multi-cube write permissions." + } + }, + "type": "object", + "title": "APIADDRequest", + "description": "Request model for creating memories." + }, + "APIChatCompleteRequest": { + "properties": { + "user_id": { + "type": "string", + "title": "User Id", + "description": "User ID" + }, + "query": { + "type": "string", + "title": "Query", + "description": "Chat query message" + }, + "readable_cube_ids": { + "anyOf": [ + { + "items": { + "type": "string" + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "Readable Cube Ids", + "description": "List of cube IDs user can read for multi-cube chat" + }, + "writable_cube_ids": { + "anyOf": [ + { + "items": { + "type": "string" + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "Writable Cube Ids", + "description": "List of cube IDs user can write for multi-cube chat" + }, + "history": { + "anyOf": [ + { + "items": { + "anyOf": [ + { + "$ref": "#/components/schemas/ChatCompletionSystemMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionUserMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionAssistantMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionToolMessageParam" + } + ] + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "History", + "description": "Chat history" + }, + "mode": { + "$ref": "#/components/schemas/SearchMode", + "description": "search mode: fast, fine, or mixture", + "default": "fast" + }, + "system_prompt": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "System Prompt", + "description": "Base system prompt to use for chat" + }, + "top_k": { + "type": "integer", + "title": "Top K", + "description": "Number of results to return", + "default": 10 + }, + "session_id": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Session Id", + "description": "Session ID for soft-filtering memories" + }, + "include_preference": { + "type": "boolean", + "title": "Include Preference", + "description": "Whether to handle preference memory", + "default": true + }, + "pref_top_k": { + "type": "integer", + "title": "Pref Top K", + "description": "Number of preference results to return", + "default": 6 + }, + "model_name_or_path": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Model Name Or Path", + "description": "Model name to use for chat" + }, + "max_tokens": { + "anyOf": [ + { + "type": "integer" + }, + { + "type": "null" + } + ], + "title": "Max Tokens", + "description": "Max tokens to generate" + }, + "temperature": { + "anyOf": [ + { + "type": "number" + }, + { + "type": "null" + } + ], + "title": "Temperature", + "description": "Temperature for sampling" + }, + "top_p": { + "anyOf": [ + { + "type": "number" + }, + { + "type": "null" + } + ], + "title": "Top P", + "description": "Top-p (nucleus) sampling parameter" + }, + "add_message_on_answer": { + "type": "boolean", + "title": "Add Message On Answer", + "description": "Add dialogs to memory after chat", + "default": true + }, + "filter": { + "anyOf": [ + { + "additionalProperties": true, + "type": "object" + }, + { + "type": "null" + } + ], + "title": "Filter", + "description": "\n Filter for the memory, example:\n {\n \"`and` or `or`\": [\n {\"id\": \"uuid-xxx\"},\n {\"created_at\": {\"gt\": \"2024-01-01\"}},\n ]\n }\n " + }, + "internet_search": { + "type": "boolean", + "title": "Internet Search", + "description": "Whether to use internet search", + "default": false + }, + "threshold": { + "type": "number", + "title": "Threshold", + "description": "Threshold for filtering references", + "default": 0.5 + }, + "mem_cube_id": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Mem Cube Id", + "description": "Cube ID to use for chat" + }, + "moscube": { + "type": "boolean", + "title": "Moscube", + "description": "(Deprecated) Whether to use legacy MemOSCube pipeline", + "default": false + } + }, + "type": "object", + "required": [ + "user_id", + "query" + ], + "title": "APIChatCompleteRequest", + "description": "Request model for chat operations." + }, + "APIFeedbackRequest": { + "properties": { + "user_id": { + "type": "string", + "title": "User Id", + "description": "User ID" + }, + "session_id": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Session Id", + "description": "Session ID for soft-filtering memories", + "default": "default_session" + }, + "task_id": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Task Id", + "description": "Task ID for monitering async tasks" + }, + "history": { + "anyOf": [ + { + "items": { + "anyOf": [ + { + "$ref": "#/components/schemas/ChatCompletionSystemMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionUserMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionAssistantMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionToolMessageParam" + } + ] + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "History", + "description": "Chat history" + }, + "retrieved_memory_ids": { + "anyOf": [ + { + "items": { + "type": "string" + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "Retrieved Memory Ids", + "description": "Retrieved memory ids at last turn" + }, + "feedback_content": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Feedback Content", + "description": "Feedback content to process" + }, + "feedback_time": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Feedback Time", + "description": "Feedback time" + }, + "writable_cube_ids": { + "anyOf": [ + { + "items": { + "type": "string" + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "Writable Cube Ids", + "description": "List of cube IDs user can write for multi-cube add" + }, + "async_mode": { + "type": "string", + "enum": [ + "sync", + "async" + ], + "title": "Async Mode", + "description": "feedback mode: sync or async", + "default": "async" + }, + "corrected_answer": { + "type": "boolean", + "title": "Corrected Answer", + "description": "Whether need return corrected answer", + "default": false + }, + "info": { + "anyOf": [ + { + "additionalProperties": true, + "type": "object" + }, + { + "type": "null" + } + ], + "title": "Info", + "description": "Additional metadata for the add request. All keys can be used as filters in search. Example: {'agent_id': 'xxxxxx', 'app_id': 'xxxx', 'source_type': 'web', 'source_url': 'https://www.baidu.com', 'source_content': 'West Lake is the most famous scenic spot in Hangzhou'}." + }, + "mem_cube_id": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Mem Cube Id", + "description": "(Deprecated) Single cube ID to search in. Prefer `readable_cube_ids` for multi-cube search." + } + }, + "type": "object", + "required": [ + "user_id", + "history", + "feedback_content" + ], + "title": "APIFeedbackRequest", + "description": "Request model for processing feedback info." + }, + "APISearchRequest": { + "properties": { + "query": { + "type": "string", + "title": "Query", + "description": "User search query" + }, + "user_id": { + "type": "string", + "title": "User Id", + "description": "User ID" + }, + "readable_cube_ids": { + "anyOf": [ + { + "items": { + "type": "string" + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "Readable Cube Ids", + "description": "List of cube IDs that are readable for this request. Required for algorithm-facing API; optional for developer-facing API." + }, + "mode": { + "$ref": "#/components/schemas/SearchMode", + "description": "Search mode: fast, fine, or mixture.", + "default": "fast" + }, + "session_id": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Session Id", + "description": "Session ID used as a soft signal to prioritize more relevant memories. Only used for weighting, not as a hard filter." + }, + "top_k": { + "type": "integer", + "minimum": 1.0, + "title": "Top K", + "description": "Number of textual memories to retrieve (top-K). Default: 10.", + "default": 10 + }, + "dedup": { + "anyOf": [ + { + "type": "string", + "enum": [ + "no", + "sim" + ] + }, + { + "type": "null" + } + ], + "title": "Dedup", + "description": "Optional dedup option for textual memories. Use 'no' for no dedup, 'sim' for similarity dedup. If None, default exact-text dedup is applied." + }, + "pref_top_k": { + "type": "integer", + "minimum": 0.0, + "title": "Pref Top K", + "description": "Number of preference memories to retrieve (top-K). Default: 6.", + "default": 6 + }, + "include_preference": { + "type": "boolean", + "title": "Include Preference", + "description": "Whether to retrieve preference memories along with general memories. If enabled, the system will automatically recall user preferences relevant to the query. Default: True.", + "default": true + }, + "search_tool_memory": { + "type": "boolean", + "title": "Search Tool Memory", + "description": "Whether to retrieve tool memories along with general memories. If enabled, the system will automatically recall tool memories relevant to the query. Default: True.", + "default": true + }, + "tool_mem_top_k": { + "type": "integer", + "minimum": 0.0, + "title": "Tool Mem Top K", + "description": "Number of tool memories to retrieve (top-K). Default: 6.", + "default": 6 + }, + "filter": { + "anyOf": [ + { + "additionalProperties": true, + "type": "object" + }, + { + "type": "null" + } + ], + "title": "Filter", + "description": "\n Filter for the memory, example:\n {\n \"`and` or `or`\": [\n {\"id\": \"uuid-xxx\"},\n {\"created_at\": {\"gt\": \"2024-01-01\"}},\n ]\n }\n " + }, + "internet_search": { + "type": "boolean", + "title": "Internet Search", + "description": "Whether to enable internet search in addition to memory search. Primarily used by internal algorithms. Default: False.", + "default": false + }, + "threshold": { + "anyOf": [ + { + "type": "number" + }, + { + "type": "null" + } + ], + "title": "Threshold", + "description": "Internal similarity threshold for searching plaintext memories. If None, default thresholds will be applied." + }, + "search_memory_type": { + "type": "string", + "title": "Search Memory Type", + "description": "Type of memory to search: All, WorkingMemory, LongTermMemory, UserMemory, OuterMemory, ToolSchemaMemory, ToolTrajectoryMemory", + "default": "All" + }, + "chat_history": { + "anyOf": [ + { + "items": { + "anyOf": [ + { + "$ref": "#/components/schemas/ChatCompletionSystemMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionUserMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionAssistantMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionToolMessageParam" + } + ] + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "Chat History", + "description": "Historical chat messages used internally by algorithms. If None, internal stored history may be used; if provided (even an empty list), this value will be used as-is." + }, + "mem_cube_id": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Mem Cube Id", + "description": "(Deprecated) Single cube ID to search in. Prefer `readable_cube_ids` for multi-cube search." + }, + "moscube": { + "type": "boolean", + "title": "Moscube", + "description": "(Deprecated / internal) Whether to use legacy MemOSCube path.", + "default": false + }, + "operation": { + "anyOf": [ + { + "items": { + "$ref": "#/components/schemas/PermissionDict" + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "Operation", + "description": "(Internal) Operation definitions for multi-cube read permissions." + }, + "source": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Source", + "description": "Source of the search query [plugin will router diff search]" + } + }, + "type": "object", + "required": [ + "query", + "user_id" + ], + "title": "APISearchRequest", + "description": "Request model for searching memories." + }, + "AllStatusResponse": { + "properties": { + "code": { + "type": "integer", + "title": "Code", + "description": "Response status code", + "default": 200 + }, + "message": { + "type": "string", + "title": "Message", + "default": "Scheduler status summary retrieved successfully" + }, + "data": { + "anyOf": [ + { + "$ref": "#/components/schemas/AllStatusResponseData" + }, + { + "type": "null" + } + ], + "description": "Response data" + } + }, + "type": "object", + "title": "AllStatusResponse", + "description": "Response model for full scheduler status operations." + }, + "AllStatusResponseData": { + "properties": { + "scheduler_summary": { + "$ref": "#/components/schemas/TaskSummary", + "description": "Aggregated status for scheduler-managed tasks" + }, + "all_tasks_summary": { + "$ref": "#/components/schemas/TaskSummary", + "description": "Aggregated status for all tracked tasks" + } + }, + "type": "object", + "required": [ + "scheduler_summary", + "all_tasks_summary" + ], + "title": "AllStatusResponseData", + "description": "Aggregated scheduler status metrics." + }, + "Audio": { + "properties": { + "id": { + "type": "string", + "title": "Id" + } + }, + "type": "object", + "required": [ + "id" + ], + "title": "Audio" + }, + "ChatCompletionAssistantMessageParam": { + "properties": { + "role": { + "type": "string", + "const": "assistant", + "title": "Role" + }, + "audio": { + "anyOf": [ + { + "$ref": "#/components/schemas/Audio" + }, + { + "type": "null" + } + ] + }, + "content": { + "anyOf": [ + { + "type": "string" + }, + { + "items": { + "anyOf": [ + { + "$ref": "#/components/schemas/ChatCompletionContentPartTextParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionContentPartRefusalParam" + } + ] + }, + "type": "array" + }, + { + "$ref": "#/components/schemas/ChatCompletionContentPartTextParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionContentPartRefusalParam" + }, + { + "type": "null" + } + ], + "title": "Content" + }, + "refusal": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Refusal" + }, + "tool_calls": { + "anyOf": [ + { + "items": { + "anyOf": [ + { + "$ref": "#/components/schemas/ChatCompletionMessageFunctionToolCallParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionMessageCustomToolCallParam" + } + ] + }, + "type": "array" + }, + { + "$ref": "#/components/schemas/ChatCompletionMessageFunctionToolCallParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionMessageCustomToolCallParam" + } + ], + "title": "Tool Calls" + }, + "chat_time": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Chat Time" + }, + "message_id": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Message Id" + } + }, + "type": "object", + "required": [ + "role" + ], + "title": "ChatCompletionAssistantMessageParam" + }, + "ChatCompletionContentPartImageParam": { + "properties": { + "image_url": { + "$ref": "#/components/schemas/ImageURL" + }, + "type": { + "type": "string", + "const": "image_url", + "title": "Type" + } + }, + "type": "object", + "required": [ + "image_url", + "type" + ], + "title": "ChatCompletionContentPartImageParam" + }, + "ChatCompletionContentPartInputAudioParam": { + "properties": { + "input_audio": { + "$ref": "#/components/schemas/InputAudio" + }, + "type": { + "type": "string", + "const": "input_audio", + "title": "Type" + } + }, + "type": "object", + "required": [ + "input_audio", + "type" + ], + "title": "ChatCompletionContentPartInputAudioParam" + }, + "ChatCompletionContentPartRefusalParam": { + "properties": { + "refusal": { + "type": "string", + "title": "Refusal" + }, + "type": { + "type": "string", + "const": "refusal", + "title": "Type" + } + }, + "type": "object", + "required": [ + "refusal", + "type" + ], + "title": "ChatCompletionContentPartRefusalParam" + }, + "ChatCompletionContentPartTextParam": { + "properties": { + "text": { + "type": "string", + "title": "Text" + }, + "type": { + "type": "string", + "const": "text", + "title": "Type" + } + }, + "type": "object", + "required": [ + "text", + "type" + ], + "title": "ChatCompletionContentPartTextParam" + }, + "ChatCompletionMessageCustomToolCallParam": { + "properties": { + "id": { + "type": "string", + "title": "Id" + }, + "custom": { + "$ref": "#/components/schemas/Custom" + }, + "type": { + "type": "string", + "const": "custom", + "title": "Type" + } + }, + "type": "object", + "required": [ + "id", + "custom", + "type" + ], + "title": "ChatCompletionMessageCustomToolCallParam" + }, + "ChatCompletionMessageFunctionToolCallParam": { + "properties": { + "id": { + "type": "string", + "title": "Id" + }, + "function": { + "$ref": "#/components/schemas/Function" + }, + "type": { + "type": "string", + "const": "function", + "title": "Type" + } + }, + "type": "object", + "required": [ + "id", + "function", + "type" + ], + "title": "ChatCompletionMessageFunctionToolCallParam" + }, + "ChatCompletionSystemMessageParam": { + "properties": { + "content": { + "anyOf": [ + { + "type": "string" + }, + { + "items": { + "$ref": "#/components/schemas/ChatCompletionContentPartTextParam" + }, + "type": "array" + }, + { + "$ref": "#/components/schemas/ChatCompletionContentPartTextParam" + } + ], + "title": "Content" + }, + "role": { + "type": "string", + "const": "system", + "title": "Role" + }, + "name": { + "type": "string", + "title": "Name" + }, + "chat_time": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Chat Time" + }, + "message_id": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Message Id" + } + }, + "type": "object", + "required": [ + "content", + "role" + ], + "title": "ChatCompletionSystemMessageParam" + }, + "ChatCompletionToolMessageParam": { + "properties": { + "content": { + "anyOf": [ + { + "type": "string" + }, + { + "items": { + "anyOf": [ + { + "$ref": "#/components/schemas/ChatCompletionContentPartTextParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionContentPartImageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionContentPartInputAudioParam" + }, + { + "$ref": "#/components/schemas/File" + } + ] + }, + "type": "array" + }, + { + "$ref": "#/components/schemas/ChatCompletionContentPartTextParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionContentPartImageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionContentPartInputAudioParam" + }, + { + "$ref": "#/components/schemas/File" + } + ], + "title": "Content" + }, + "role": { + "type": "string", + "const": "tool", + "title": "Role" + }, + "tool_call_id": { + "type": "string", + "title": "Tool Call Id" + }, + "chat_time": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Chat Time" + }, + "message_id": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Message Id" + } + }, + "type": "object", + "required": [ + "content", + "role", + "tool_call_id" + ], + "title": "ChatCompletionToolMessageParam" + }, + "ChatCompletionUserMessageParam": { + "properties": { + "content": { + "anyOf": [ + { + "type": "string" + }, + { + "items": { + "anyOf": [ + { + "$ref": "#/components/schemas/ChatCompletionContentPartTextParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionContentPartImageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionContentPartInputAudioParam" + }, + { + "$ref": "#/components/schemas/File" + } + ] + }, + "type": "array" + }, + { + "$ref": "#/components/schemas/ChatCompletionContentPartTextParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionContentPartImageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionContentPartInputAudioParam" + }, + { + "$ref": "#/components/schemas/File" + } + ], + "title": "Content" + }, + "role": { + "type": "string", + "const": "user", + "title": "Role" + }, + "name": { + "type": "string", + "title": "Name" + }, + "chat_time": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Chat Time" + }, + "message_id": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Message Id" + } + }, + "type": "object", + "required": [ + "content", + "role" + ], + "title": "ChatCompletionUserMessageParam" + }, + "ChatPlaygroundRequest": { + "properties": { + "user_id": { + "type": "string", + "title": "User Id", + "description": "User ID" + }, + "query": { + "type": "string", + "title": "Query", + "description": "Chat query message" + }, + "readable_cube_ids": { + "anyOf": [ + { + "items": { + "type": "string" + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "Readable Cube Ids", + "description": "List of cube IDs user can read for multi-cube chat" + }, + "writable_cube_ids": { + "anyOf": [ + { + "items": { + "type": "string" + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "Writable Cube Ids", + "description": "List of cube IDs user can write for multi-cube chat" + }, + "history": { + "anyOf": [ + { + "items": { + "anyOf": [ + { + "$ref": "#/components/schemas/ChatCompletionSystemMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionUserMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionAssistantMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionToolMessageParam" + } + ] + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "History", + "description": "Chat history" + }, + "mode": { + "$ref": "#/components/schemas/SearchMode", + "description": "search mode: fast, fine, or mixture", + "default": "fast" + }, + "system_prompt": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "System Prompt", + "description": "Base system prompt to use for chat" + }, + "top_k": { + "type": "integer", + "title": "Top K", + "description": "Number of results to return", + "default": 10 + }, + "session_id": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Session Id", + "description": "Session ID for soft-filtering memories" + }, + "include_preference": { + "type": "boolean", + "title": "Include Preference", + "description": "Whether to handle preference memory", + "default": true + }, + "pref_top_k": { + "type": "integer", + "title": "Pref Top K", + "description": "Number of preference results to return", + "default": 6 + }, + "model_name_or_path": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Model Name Or Path", + "description": "Model name to use for chat" + }, + "max_tokens": { + "anyOf": [ + { + "type": "integer" + }, + { + "type": "null" + } + ], + "title": "Max Tokens", + "description": "Max tokens to generate" + }, + "temperature": { + "anyOf": [ + { + "type": "number" + }, + { + "type": "null" + } + ], + "title": "Temperature", + "description": "Temperature for sampling" + }, + "top_p": { + "anyOf": [ + { + "type": "number" + }, + { + "type": "null" + } + ], + "title": "Top P", + "description": "Top-p (nucleus) sampling parameter" + }, + "add_message_on_answer": { + "type": "boolean", + "title": "Add Message On Answer", + "description": "Add dialogs to memory after chat", + "default": true + }, + "filter": { + "anyOf": [ + { + "additionalProperties": true, + "type": "object" + }, + { + "type": "null" + } + ], + "title": "Filter", + "description": "\n Filter for the memory, example:\n {\n \"`and` or `or`\": [\n {\"id\": \"uuid-xxx\"},\n {\"created_at\": {\"gt\": \"2024-01-01\"}},\n ]\n }\n " + }, + "internet_search": { + "type": "boolean", + "title": "Internet Search", + "description": "Whether to use internet search", + "default": false + }, + "threshold": { + "type": "number", + "title": "Threshold", + "description": "Threshold for filtering references", + "default": 0.5 + }, + "moscube": { + "type": "boolean", + "title": "Moscube", + "description": "(Deprecated) Whether to use legacy MemOSCube pipeline.", + "default": false + }, + "mem_cube_id": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Mem Cube Id", + "description": "(Deprecated) Single cube ID to use for chat. Prefer `readable_cube_ids` / `writable_cube_ids` for multi-cube chat." + }, + "beginner_guide_step": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Beginner Guide Step", + "description": "Whether to use beginner guide, option: [first, second]" + } + }, + "type": "object", + "required": [ + "user_id", + "query" + ], + "title": "ChatPlaygroundRequest", + "description": "Request model for chat operations in playground." + }, + "ChatRequest": { + "properties": { + "user_id": { + "type": "string", + "title": "User Id", + "description": "User ID" + }, + "query": { + "type": "string", + "title": "Query", + "description": "Chat query message" + }, + "readable_cube_ids": { + "anyOf": [ + { + "items": { + "type": "string" + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "Readable Cube Ids", + "description": "List of cube IDs user can read for multi-cube chat" + }, + "writable_cube_ids": { + "anyOf": [ + { + "items": { + "type": "string" + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "Writable Cube Ids", + "description": "List of cube IDs user can write for multi-cube chat" + }, + "history": { + "anyOf": [ + { + "items": { + "anyOf": [ + { + "$ref": "#/components/schemas/ChatCompletionSystemMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionUserMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionAssistantMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionToolMessageParam" + } + ] + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "History", + "description": "Chat history" + }, + "mode": { + "$ref": "#/components/schemas/SearchMode", + "description": "search mode: fast, fine, or mixture", + "default": "fast" + }, + "system_prompt": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "System Prompt", + "description": "Base system prompt to use for chat" + }, + "top_k": { + "type": "integer", + "title": "Top K", + "description": "Number of results to return", + "default": 10 + }, + "session_id": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Session Id", + "description": "Session ID for soft-filtering memories" + }, + "include_preference": { + "type": "boolean", + "title": "Include Preference", + "description": "Whether to handle preference memory", + "default": true + }, + "pref_top_k": { + "type": "integer", + "title": "Pref Top K", + "description": "Number of preference results to return", + "default": 6 + }, + "model_name_or_path": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Model Name Or Path", + "description": "Model name to use for chat" + }, + "max_tokens": { + "anyOf": [ + { + "type": "integer" + }, + { + "type": "null" + } + ], + "title": "Max Tokens", + "description": "Max tokens to generate" }, - "422": { - "description": "Validation Error", - "content": { - "application/json": { - "schema": { - "$ref": "#/components/schemas/HTTPValidationError" - } - } - } - } - } - } - }, - "/chat": { - "post": { - "summary": "Chat with MemOS", - "description": "Chat with the MemOS system.", - "operationId": "chat_chat_post", - "requestBody": { - "content": { - "application/json": { - "schema": { - "$ref": "#/components/schemas/ChatRequest" + "temperature": { + "anyOf": [ + { + "type": "number" + }, + { + "type": "null" } - } + ], + "title": "Temperature", + "description": "Temperature for sampling" }, - "required": true - }, - "responses": { - "200": { - "description": "Successful Response", - "content": { - "application/json": { - "schema": { - "$ref": "#/components/schemas/ChatResponse" - } + "top_p": { + "anyOf": [ + { + "type": "number" + }, + { + "type": "null" } - } + ], + "title": "Top P", + "description": "Top-p (nucleus) sampling parameter" }, - "422": { - "description": "Validation Error", - "content": { - "application/json": { - "schema": { - "$ref": "#/components/schemas/HTTPValidationError" - } + "add_message_on_answer": { + "type": "boolean", + "title": "Add Message On Answer", + "description": "Add dialogs to memory after chat", + "default": true + }, + "filter": { + "anyOf": [ + { + "additionalProperties": true, + "type": "object" + }, + { + "type": "null" } - } - } - } - } - } - }, - "components": { - "schemas": { - "ChatRequest": { - "properties": { - "user_id": { + ], + "title": "Filter", + "description": "\n Filter for the memory, example:\n {\n \"`and` or `or`\": [\n {\"id\": \"uuid-xxx\"},\n {\"created_at\": {\"gt\": \"2024-01-01\"}},\n ]\n }\n " + }, + "internet_search": { + "type": "boolean", + "title": "Internet Search", + "description": "Whether to use internet search", + "default": false + }, + "threshold": { + "type": "number", + "title": "Threshold", + "description": "Threshold for filtering references", + "default": 0.5 + }, + "moscube": { + "type": "boolean", + "title": "Moscube", + "description": "(Deprecated) Whether to use legacy MemOSCube pipeline.", + "default": false + }, + "mem_cube_id": { "anyOf": [ { "type": "string" @@ -716,42 +2602,112 @@ "type": "null" } ], - "title": "User Id", - "description": "User ID for the request", - "example": "user123" + "title": "Mem Cube Id", + "description": "(Deprecated) Single cube ID to use for chat. Prefer `readable_cube_ids` / `writable_cube_ids` for multi-cube chat." + } + }, + "type": "object", + "required": [ + "user_id", + "query" + ], + "title": "ChatRequest", + "description": "Request model for chat operations.\n\nThis model is used as the algorithm-facing chat interface, while also\nremaining backward compatible with older developer-facing APIs." + }, + "Custom": { + "properties": { + "input": { + "type": "string", + "title": "Input" }, - "query": { + "name": { "type": "string", - "title": "Query", - "description": "Chat query message.", - "example": "What is the latest update?" + "title": "Name" } }, "type": "object", "required": [ - "query" + "input", + "name" ], - "title": "ChatRequest" + "title": "Custom" + }, + "DeleteMemoryRequest": { + "properties": { + "writable_cube_ids": { + "items": { + "type": "string" + }, + "type": "array", + "title": "Writable Cube Ids", + "description": "Writable cube IDs" + }, + "memory_ids": { + "anyOf": [ + { + "items": { + "type": "string" + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "Memory Ids", + "description": "Memory IDs" + }, + "file_ids": { + "anyOf": [ + { + "items": { + "type": "string" + }, + "type": "array" + }, + { + "type": "null" + } + ], + "title": "File Ids", + "description": "File IDs" + }, + "filter": { + "anyOf": [ + { + "additionalProperties": true, + "type": "object" + }, + { + "type": "null" + } + ], + "title": "Filter", + "description": "Filter for the memory" + } + }, + "type": "object", + "title": "DeleteMemoryRequest", + "description": "Request model for deleting memories." }, - "ChatResponse": { + "DeleteMemoryResponse": { "properties": { "code": { "type": "integer", "title": "Code", "description": "Response status code", - "default": 200, - "example": 200 + "default": 200 }, "message": { "type": "string", "title": "Message", - "description": "Response message", - "example": "Operation successful" + "description": "Response message" }, "data": { "anyOf": [ { - "type": "string" + "additionalProperties": true, + "type": "object" }, { "type": "null" @@ -765,26 +2721,49 @@ "required": [ "message" ], - "title": "ChatResponse", - "description": "Response model for chat operations." + "title": "DeleteMemoryResponse", + "description": "Response model for deleting memories." + }, + "ExistMemCubeIdRequest": { + "properties": { + "mem_cube_id": { + "type": "string", + "title": "Mem Cube Id", + "description": "Mem cube ID" + } + }, + "type": "object", + "required": [ + "mem_cube_id" + ], + "title": "ExistMemCubeIdRequest", + "description": "Request model for checking if mem cube id exists." }, - "ConfigResponse": { + "ExistMemCubeIdResponse": { "properties": { "code": { "type": "integer", "title": "Code", "description": "Response status code", - "default": 200, - "example": 200 + "default": 200 }, "message": { "type": "string", "title": "Message", - "description": "Response message", - "example": "Operation successful" + "description": "Response message" }, "data": { - "type": "null", + "anyOf": [ + { + "additionalProperties": { + "type": "boolean" + }, + "type": "object" + }, + { + "type": "null" + } + ], "title": "Data", "description": "Response data" } @@ -793,247 +2772,124 @@ "required": [ "message" ], - "title": "ConfigResponse", - "description": "Response model for configuration endpoint." + "title": "ExistMemCubeIdResponse", + "description": "Response model for checking if mem cube id exists." }, - "CubeShare": { + "File": { "properties": { - "user_id": { - "anyOf": [ - { - "type": "string" - }, - { - "type": "null" - } - ], - "title": "User Id", - "description": "User ID for the request", - "example": "user123" + "file": { + "$ref": "#/components/schemas/FileFile" }, - "target_user_id": { + "type": { "type": "string", - "title": "Target User Id", - "description": "Target user ID to share with", - "example": "user456" + "const": "file", + "title": "Type" } }, "type": "object", "required": [ - "target_user_id" + "file", + "type" ], - "title": "CubeShare" + "title": "File" }, - "HTTPValidationError": { + "FileFile": { "properties": { - "detail": { - "items": { - "$ref": "#/components/schemas/ValidationError" - }, - "type": "array", - "title": "Detail" + "file_data": { + "type": "string", + "title": "File Data" + }, + "file_id": { + "type": "string", + "title": "File Id" + }, + "filename": { + "type": "string", + "title": "Filename" } }, "type": "object", - "title": "HTTPValidationError" + "title": "FileFile" }, - "LLMConfigFactory": { + "Function": { "properties": { - "model_schema": { + "arguments": { "type": "string", - "title": "Model Schema", - "description": "Schema for configuration. This value will be automatically set.", - "default": "NOT_SET" + "title": "Arguments" }, - "backend": { + "name": { "type": "string", - "title": "Backend", - "description": "Backend for LLM" - }, - "config": { - "additionalProperties": true, - "type": "object", - "title": "Config", - "description": "Configuration for the LLM backend" + "title": "Name" } }, - "additionalProperties": false, "type": "object", "required": [ - "backend", - "config" + "arguments", + "name" ], - "title": "LLMConfigFactory", - "description": "Factory class for creating LLM configurations." + "title": "Function" }, - "MOSConfig": { + "GetMemoryPlaygroundRequest": { "properties": { - "model_schema": { - "type": "string", - "title": "Model Schema", - "description": "Schema for configuration. This value will be automatically set.", - "default": "NOT_SET" - }, "user_id": { "type": "string", "title": "User Id", - "description": "User ID for the MOS. This is used to distinguish between different users' memories.", - "default": "root" + "description": "User ID" }, - "session_id": { + "memory_type": { "type": "string", - "title": "Session Id", - "description": "Session ID for the MOS. This is used to distinguish between different dialogue", - "default": "8dcdbd62-c231-4678-a3ae-0946b7d9ce14" - }, - "chat_model": { - "$ref": "#/components/schemas/LLMConfigFactory", - "description": "LLM configuration for the chat model in the MOS" - }, - "mem_reader": { - "$ref": "#/components/schemas/MemReaderConfigFactory", - "description": "MemReader configuration for the MOS" - }, - "mem_scheduler": { - "anyOf": [ - { - "$ref": "#/components/schemas/SchedulerConfigFactory" - }, - { - "type": "null" - } + "enum": [ + "text_mem", + "act_mem", + "param_mem", + "para_mem" ], - "description": "Memory scheduler configuration for managing memory operations" - }, - "user_manager": { - "$ref": "#/components/schemas/UserManagerConfigFactory", - "description": "User manager configuration for database operations" - }, - "max_turns_window": { - "type": "integer", - "title": "Max Turns Window", - "description": "Maximum number of turns to keep in the conversation history", - "default": 15 - }, - "top_k": { - "type": "integer", - "title": "Top K", - "description": "Maximum number of memories to retrieve for each query", - "default": 5 + "title": "Memory Type", + "description": "Memory type" }, - "enable_textual_memory": { - "type": "boolean", - "title": "Enable Textual Memory", - "description": "Enable textual memory for the MemChat", - "default": true - }, - "enable_activation_memory": { - "type": "boolean", - "title": "Enable Activation Memory", - "description": "Enable activation memory for the MemChat", - "default": false - }, - "enable_parametric_memory": { - "type": "boolean", - "title": "Enable Parametric Memory", - "description": "Enable parametric memory for the MemChat", - "default": false - }, - "enable_preference_memory": { - "type": "boolean", - "title": "Enable Preference Memory", - "description": "Enable preference memory for the MemChat", - "default": false - }, - "enable_mem_scheduler": { - "type": "boolean", - "title": "Enable Mem Scheduler", - "description": "Enable memory scheduler for automated memory management", - "default": false - }, - "PRO_MODE": { - "type": "boolean", - "title": "Pro Mode", - "description": "Enable PRO mode for complex query decomposition", - "default": false - } - }, - "additionalProperties": false, - "type": "object", - "title": "MOSConfig" - }, - "MemCubeRegister": { - "properties": { - "user_id": { + "mem_cube_ids": { "anyOf": [ { - "type": "string" + "items": { + "type": "string" + }, + "type": "array" }, { "type": "null" } ], - "title": "User Id", - "description": "User ID for the request", - "example": "user123" - }, - "mem_cube_name_or_path": { - "type": "string", - "title": "Mem Cube Name Or Path", - "description": "Name or path of the MemCube to register.", - "example": "/path/to/cube" + "title": "Mem Cube Ids", + "description": "Cube IDs" }, - "mem_cube_id": { + "search_query": { "anyOf": [ { "type": "string" }, - { - "type": "null" - } - ], - "title": "Mem Cube Id", - "description": "ID for the MemCube", - "example": "cube123" - } - }, - "type": "object", - "required": [ - "mem_cube_name_or_path" - ], - "title": "MemCubeRegister" - }, - "MemReaderConfigFactory": { - "properties": { - "model_schema": { - "type": "string", - "title": "Model Schema", - "description": "Schema for configuration. This value will be automatically set.", - "default": "NOT_SET" - }, - "backend": { - "type": "string", - "title": "Backend", - "description": "Backend for MemReader" - }, - "config": { - "additionalProperties": true, - "type": "object", - "title": "Config", - "description": "Configuration for the MemReader backend" + { + "type": "null" + } + ], + "title": "Search Query", + "description": "Search query" } }, - "additionalProperties": false, "type": "object", "required": [ - "backend", - "config" + "user_id", + "memory_type" ], - "title": "MemReaderConfigFactory", - "description": "Factory class for creating MemReader configurations." + "title": "GetMemoryPlaygroundRequest", + "description": "Request model for getting memories." }, - "MemoryCreate": { + "GetMemoryRequest": { "properties": { + "mem_cube_id": { + "type": "string", + "title": "Mem Cube Id", + "description": "Cube ID" + }, "user_id": { "anyOf": [ { @@ -1044,92 +2900,124 @@ } ], "title": "User Id", - "description": "User ID for the request", - "example": "user123" + "description": "User ID" }, - "messages": { - "anyOf": [ - { - "items": { - "$ref": "#/components/schemas/Message" - }, - "type": "array" - }, - { - "type": "null" - } - ], - "title": "Messages", - "description": "List of messages to store.", - "example": [ - { - "content": "Hello", - "role": "user" - } - ] + "include_preference": { + "type": "boolean", + "title": "Include Preference", + "description": "Whether to handle preference memory", + "default": true }, - "mem_cube_id": { + "page": { "anyOf": [ { - "type": "string" + "type": "integer" }, { "type": "null" } ], - "title": "Mem Cube Id", - "description": "ID of the memory cube", - "example": "cube123" + "title": "Page", + "description": "Page number (starts from 1). If None, exports all data without pagination." }, - "memory_content": { + "page_size": { "anyOf": [ { - "type": "string" + "type": "integer" }, { "type": "null" } ], - "title": "Memory Content", - "description": "Content to store as memory", - "example": "This is a memory content" + "title": "Page Size", + "description": "Number of items per page. If None, exports all data without pagination." + } + }, + "type": "object", + "required": [ + "mem_cube_id" + ], + "title": "GetMemoryRequest", + "description": "Request model for getting memories." + }, + "GetMemoryResponse": { + "properties": { + "code": { + "type": "integer", + "title": "Code", + "description": "Response status code", + "default": 200 }, - "doc_path": { + "message": { + "type": "string", + "title": "Message", + "description": "Response message" + }, + "data": { "anyOf": [ { - "type": "string" + "additionalProperties": true, + "type": "object" }, { "type": "null" } ], - "title": "Doc Path", - "description": "Path to document to store", - "example": "/path/to/document.txt" + "title": "Data", + "description": "Response data" } }, "type": "object", - "title": "MemoryCreate" + "required": [ + "message" + ], + "title": "GetMemoryResponse", + "description": "Response model for getting memories." }, - "MemoryResponse": { + "GetUserNamesByMemoryIdsRequest": { + "properties": { + "memory_ids": { + "items": { + "type": "string" + }, + "type": "array", + "title": "Memory Ids", + "description": "Memory IDs" + } + }, + "type": "object", + "required": [ + "memory_ids" + ], + "title": "GetUserNamesByMemoryIdsRequest", + "description": "Request model for getting user names by memory ids." + }, + "GetUserNamesByMemoryIdsResponse": { "properties": { "code": { "type": "integer", "title": "Code", "description": "Response status code", - "default": 200, - "example": 200 + "default": 200 }, "message": { "type": "string", "title": "Message", - "description": "Response message", - "example": "Operation successful" + "description": "Response message" }, "data": { "anyOf": [ { - "additionalProperties": true, + "additionalProperties": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ] + }, "type": "object" }, { @@ -1144,106 +3032,133 @@ "required": [ "message" ], - "title": "MemoryResponse", - "description": "Response model for memory operations." + "title": "GetUserNamesByMemoryIdsResponse", + "description": "Response model for getting user names by memory ids." }, - "Message": { + "HTTPValidationError": { "properties": { - "role": { + "detail": { + "items": { + "$ref": "#/components/schemas/ValidationError" + }, + "type": "array", + "title": "Detail" + } + }, + "type": "object", + "title": "HTTPValidationError" + }, + "ImageURL": { + "properties": { + "url": { "type": "string", - "title": "Role", - "description": "Role of the message (user or assistant).", - "example": "user" + "title": "Url" }, - "content": { + "detail": { "type": "string", - "title": "Content", - "description": "Message content.", - "example": "Hello, how can I help you?" + "enum": [ + "auto", + "low", + "high" + ], + "title": "Detail" } }, "type": "object", "required": [ - "role", - "content" + "url" ], - "title": "Message" + "title": "ImageURL" }, - "SchedulerConfigFactory": { + "InputAudio": { "properties": { - "model_schema": { + "data": { "type": "string", - "title": "Model Schema", - "description": "Schema for configuration. This value will be automatically set.", - "default": "NOT_SET" + "title": "Data" }, - "backend": { + "format": { "type": "string", - "title": "Backend", - "description": "Backend for scheduler" - }, - "config": { - "additionalProperties": true, - "type": "object", - "title": "Config", - "description": "Configuration for the scheduler backend" + "enum": [ + "wav", + "mp3" + ], + "title": "Format" } }, - "additionalProperties": false, "type": "object", "required": [ - "backend", - "config" + "data", + "format" ], - "title": "SchedulerConfigFactory", - "description": "Factory class for creating scheduler configurations." + "title": "InputAudio" }, - "SearchRequest": { + "MemoryResponse": { "properties": { - "user_id": { - "anyOf": [ - { - "type": "string" - }, - { - "type": "null" - } - ], - "title": "User Id", - "description": "User ID for the request", - "example": "user123" + "code": { + "type": "integer", + "title": "Code", + "description": "Response status code", + "default": 200 }, - "query": { + "message": { "type": "string", - "title": "Query", - "description": "Search query.", - "example": "How to implement a feature?" + "title": "Message", + "description": "Response message" }, - "install_cube_ids": { + "data": { "anyOf": [ { - "items": { - "type": "string" - }, + "items": {}, "type": "array" }, { "type": "null" } ], - "title": "Install Cube Ids", - "description": "List of cube IDs to search in", - "example": [ - "cube123", - "cube456" - ] + "title": "Data", + "description": "Response data" } }, "type": "object", "required": [ - "query" + "message" + ], + "title": "MemoryResponse", + "description": "Response model for memory operations." + }, + "PermissionDict": { + "properties": { + "permissions": { + "items": { + "type": "string", + "enum": [ + "read", + "write", + "delete", + "execute" + ] + }, + "type": "array", + "title": "Permissions" + }, + "mem_cube_id": { + "type": "string", + "title": "Mem Cube Id" + } + }, + "type": "object", + "title": "PermissionDict", + "description": "Typed dictionary for chat message dictionaries." + }, + "SearchMode": { + "type": "string", + "enum": [ + "fast", + "fine", + "mixture" ], - "title": "SearchRequest" + "title": "SearchMode", + "description": "Enumeration for search modes." }, "SearchResponse": { "properties": { @@ -1251,14 +3166,12 @@ "type": "integer", "title": "Code", "description": "Response status code", - "default": 200, - "example": 200 + "default": 200 }, "message": { "type": "string", "title": "Message", - "description": "Response message", - "example": "Operation successful" + "description": "Response message" }, "data": { "anyOf": [ @@ -1281,89 +3194,165 @@ "title": "SearchResponse", "description": "Response model for search operations." }, - "SimpleResponse": { + "StatusResponse": { "properties": { "code": { "type": "integer", "title": "Code", "description": "Response status code", - "default": 200, - "example": 200 + "default": 200 }, "message": { "type": "string", "title": "Message", - "description": "Response message", - "example": "Operation successful" + "default": "Memory get status successfully" }, "data": { - "type": "null", + "anyOf": [ + { + "items": { + "$ref": "#/components/schemas/StatusResponseItem" + }, + "type": "array" + }, + { + "type": "null" + } + ], "title": "Data", "description": "Response data" } }, "type": "object", + "title": "StatusResponse", + "description": "Response model for scheduler status operations." + }, + "StatusResponseItem": { + "properties": { + "task_id": { + "type": "string", + "title": "Task Id", + "description": "The ID of the task" + }, + "status": { + "type": "string", + "enum": [ + "in_progress", + "completed", + "waiting", + "failed", + "cancelled" + ], + "title": "Status", + "description": "The current status of the task" + } + }, + "type": "object", "required": [ - "message" + "task_id", + "status" ], - "title": "SimpleResponse", - "description": "Simple response model for operations without data return." + "title": "StatusResponseItem", + "description": "Individual task status item." }, - "UserCreate": { + "SuggestionRequest": { "properties": { "user_id": { "type": "string", "title": "User Id", - "description": "User ID", - "example": "user123" + "description": "User ID" }, - "user_name": { + "mem_cube_id": { + "type": "string", + "title": "Mem Cube Id", + "description": "Cube ID" + }, + "language": { + "type": "string", + "enum": [ + "zh", + "en" + ], + "title": "Language", + "description": "Language for suggestions", + "default": "zh" + }, + "message": { "anyOf": [ { "type": "string" }, + { + "items": { + "anyOf": [ + { + "$ref": "#/components/schemas/ChatCompletionSystemMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionUserMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionAssistantMessageParam" + }, + { + "$ref": "#/components/schemas/ChatCompletionToolMessageParam" + } + ] + }, + "type": "array" + }, + { + "items": { + "anyOf": [ + { + "$ref": "#/components/schemas/ChatCompletionContentPartTextParam" + }, + { + "$ref": "#/components/schemas/File" + } + ] + }, + "type": "array" + }, { "type": "null" } ], - "title": "User Name", - "description": "Name of the user", - "example": "john_doe" - }, - "role": { - "type": "string", - "title": "Role", - "description": "Role of the user", - "default": "user", - "example": "user" + "title": "Message", + "description": "List of messages to store." } }, "type": "object", "required": [ - "user_id" + "user_id", + "mem_cube_id" ], - "title": "UserCreate" + "title": "SuggestionRequest", + "description": "Request model for getting suggestion queries." }, - "UserListResponse": { + "SuggestionResponse": { "properties": { "code": { "type": "integer", "title": "Code", "description": "Response status code", - "default": 200, - "example": 200 + "default": 200 }, "message": { "type": "string", "title": "Message", - "description": "Response message", - "example": "Operation successful" + "description": "Response message" }, "data": { "anyOf": [ { - "items": {}, - "type": "array" + "additionalProperties": { + "items": { + "type": "string" + }, + "type": "array" + }, + "type": "object" }, { "type": "null" @@ -1377,63 +3366,170 @@ "required": [ "message" ], - "title": "UserListResponse", - "description": "Response model for user list operations." + "title": "SuggestionResponse", + "description": "Response model for suggestion operations." }, - "UserManagerConfigFactory": { + "TaskQueueData": { "properties": { - "backend": { + "user_id": { "type": "string", - "title": "Backend", - "description": "Backend for user manager", - "default": "sqlite" + "title": "User Id", + "description": "User ID the query is scoped to" + }, + "user_name": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "User Name", + "description": "User name if available" + }, + "mem_cube_id": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "null" + } + ], + "title": "Mem Cube Id", + "description": "MemCube ID if a single cube is targeted; otherwise None" + }, + "stream_keys": { + "items": { + "type": "string" + }, + "type": "array", + "title": "Stream Keys", + "description": "Matched Redis stream keys for this user" + }, + "users_count": { + "type": "integer", + "title": "Users Count", + "description": "Distinct users currently present in queue streams" + }, + "pending_tasks_count": { + "type": "integer", + "title": "Pending Tasks Count", + "description": "Count of pending (delivered, not acked) tasks" + }, + "remaining_tasks_count": { + "type": "integer", + "title": "Remaining Tasks Count", + "description": "Count of enqueued tasks (xlen)" + }, + "pending_tasks_detail": { + "items": { + "type": "string" + }, + "type": "array", + "title": "Pending Tasks Detail", + "description": "Per-stream pending counts, formatted as '{stream_key}:{count}'" }, - "config": { - "additionalProperties": true, - "type": "object", - "title": "Config", - "description": "Configuration for the user manager backend" + "remaining_tasks_detail": { + "items": { + "type": "string" + }, + "type": "array", + "title": "Remaining Tasks Detail", + "description": "Per-stream remaining counts, formatted as '{stream_key}:{count}'" } }, "type": "object", - "title": "UserManagerConfigFactory", - "description": "Factory for user manager configurations." + "required": [ + "user_id", + "stream_keys", + "users_count", + "pending_tasks_count", + "remaining_tasks_count", + "pending_tasks_detail", + "remaining_tasks_detail" + ], + "title": "TaskQueueData", + "description": "Queue-level metrics for scheduler tasks." }, - "UserResponse": { + "TaskQueueResponse": { "properties": { "code": { "type": "integer", "title": "Code", "description": "Response status code", - "default": 200, - "example": 200 + "default": 200 }, "message": { "type": "string", "title": "Message", - "description": "Response message", - "example": "Operation successful" + "default": "Scheduler task queue status retrieved successfully" }, "data": { "anyOf": [ { - "additionalProperties": true, - "type": "object" + "$ref": "#/components/schemas/TaskQueueData" }, { "type": "null" } ], - "title": "Data", "description": "Response data" } }, "type": "object", - "required": [ - "message" - ], - "title": "UserResponse", - "description": "Response model for user operations." + "title": "TaskQueueResponse", + "description": "Response model for scheduler task queue status." + }, + "TaskSummary": { + "properties": { + "waiting": { + "type": "integer", + "title": "Waiting", + "description": "Number of tasks waiting to run", + "default": 0 + }, + "in_progress": { + "type": "integer", + "title": "In Progress", + "description": "Number of tasks currently running", + "default": 0 + }, + "pending": { + "type": "integer", + "title": "Pending", + "description": "Number of tasks fetched by workers but not yet acknowledged", + "default": 0 + }, + "completed": { + "type": "integer", + "title": "Completed", + "description": "Number of tasks completed", + "default": 0 + }, + "failed": { + "type": "integer", + "title": "Failed", + "description": "Number of tasks failed", + "default": 0 + }, + "cancelled": { + "type": "integer", + "title": "Cancelled", + "description": "Number of tasks cancelled", + "default": 0 + }, + "total": { + "type": "integer", + "title": "Total", + "description": "Total number of tasks counted", + "default": 0 + } + }, + "type": "object", + "title": "TaskSummary", + "description": "Aggregated counts of tasks by status." }, "ValidationError": { "properties": { diff --git a/examples/api/product_api.py b/examples/api/product_api.py deleted file mode 100644 index e364ce483..000000000 --- a/examples/api/product_api.py +++ /dev/null @@ -1,170 +0,0 @@ -#!/usr/bin/env python3 -""" -Simulate full MemOS Product API workflow: -1. Register user -2. Add memory -3. Search memory -4. Chat (stream) -""" - -import json - -import requests - - -BASE_URL = "http://0.0.0.0:8001/product" -HEADERS = {"Content-Type": "application/json"} - -index = "24" -USER_ID = f"memos_user_id_{index}" -USER_NAME = f"memos_user_alice_{index}" -MEM_CUBE_ID = f"memos_cube_id_{index}" -SESSION_ID = f"memos_session_id_{index}" -SESSION_ID2 = f"memos_session_id_{index}_s2" - - -def register_user(): - url = f"{BASE_URL}/users/register" - data = { - "user_id": USER_ID, - "user_name": USER_NAME, - "interests": "memory,retrieval,test", - "mem_cube_id": MEM_CUBE_ID, - } - print(f"[*] Registering user {USER_ID} ...") - resp = requests.post(url, headers=HEADERS, data=json.dumps(data), timeout=30) - print(resp.status_code, resp.text) - return resp.json() - - -def add_memory(): - url = f"{BASE_URL}/add" - data = { - "user_id": USER_ID, - "memory_content": "今天我在测试 MemOS 的记忆添加与检索流程。", - "messages": [{"role": "user", "content": "我今天在做系统测试"}], - "doc_path": None, - "mem_cube_id": MEM_CUBE_ID, - "source": "test_script", - "user_profile": False, - "session_id": SESSION_ID, - } - print("[*] Adding memory ...") - resp = requests.post(url, headers=HEADERS, data=json.dumps(data), timeout=30) - print(resp.status_code, resp.text) - return resp.json() - - -def search_memory(query="系统测试"): - url = f"{BASE_URL}/search" - data = { - "user_id": USER_ID, - "query": query, - "mem_cube_id": MEM_CUBE_ID, - "top_k": 5, - "session_id": SESSION_ID, - } - print("[*] Searching memory ...") - resp = requests.post(url, headers=HEADERS, data=json.dumps(data), timeout=30) - print(resp.status_code, resp.text) - return resp.json() - - -def chat_stream(query: str, session_id: str, history: list | None = None): - url = f"{BASE_URL}/chat" - data = { - "user_id": USER_ID, - "query": query, - "mem_cube_id": MEM_CUBE_ID, - "history": history, - "internet_search": False, - "moscube": False, - "session_id": session_id, - } - - print("[*] Starting streaming chat ...") - - with requests.post(url, headers=HEADERS, data=json.dumps(data), stream=True) as resp: - for raw_line in resp.iter_lines(): - if not raw_line: - continue - line = raw_line.decode("utf-8", errors="ignore") - - payload = line.removeprefix("data: ").strip() - if payload == "[DONE]": - print("[done]") - break - - try: - msg = json.loads(payload) - msg_type = msg.get("type") - msg_data = msg.get("data") or msg.get("content") - - if msg_type == "text": - print(msg_data, end="", flush=True) - elif msg_type == "reference": - print(f"\n[参考记忆] {msg_data}") - elif msg_type == "status": - pass - elif msg_type == "suggestion": - print(f"\n[建议] {msg_data}") - elif msg_type == "end": - print("\n[✅ Chat End]") - else: - print(f"\n[{msg_type}] {msg_data}") - except Exception: - try: - print(payload.encode("latin-1").decode("utf-8"), end="") - except Exception: - print(payload) - - -def feedback_memory(feedback_content: str, history: list | None = None): - url = f"{BASE_URL}/feedback" - data = { - "user_id": USER_ID, - "writable_cube_ids": [MEM_CUBE_ID], - "history": history, - "feedback_content": feedback_content, - "async_mode": "sync", - "corrected_answer": "false", - } - - print("[*] Feedbacking memory ...") - resp = requests.post(url, headers=HEADERS, data=json.dumps(data), timeout=30) - print(resp.status_code, resp.text) - return resp.json() - - -if __name__ == "__main__": - print("===== STEP 1: Register User =====") - register_user() - - print("\n===== STEP 2: Add Memory =====") - add_memory() - - print("\n===== STEP 3: Search Memory =====") - search_memory() - - print("\n===== STEP 4: Stream Chat =====") - chat_stream("我很开心，我今天吃了好吃的拉面", SESSION_ID, history=[]) - chat_stream( - "我刚和你说什么", - SESSION_ID, - history=[ - {"role": "user", "content": "我很开心，我今天吃了好吃的拉面"}, - {"role": "assistant", "content": "🉑"}, - ], - ) - - print("\n===== STEP 5: Stream Chat =====") - chat_stream("我刚和你说什么了呢", SESSION_ID2, history=[]) - - print("\n===== STEP 6: Feedback Memory =====") - feedback_memory( - feedback_content="错啦，我今天没有吃拉面", - history=[ - {"role": "user", "content": "我刚和你说什么了呢"}, - {"role": "assistant", "content": "你今天吃了好吃的拉面"}, - ], - ) diff --git a/examples/api/server_router_api.py b/examples/api/server_router_api.py index e7c7dc558..ffa83dd2c 100644 --- a/examples/api/server_router_api.py +++ b/examples/api/server_router_api.py @@ -34,7 +34,7 @@ # Global config # --------------------------------------------------------------------------- -BASE_URL = "http://0.0.0.0:8001/product" +BASE_URL = "http://127.0.0.1:8000/product" HEADERS = {"Content-Type": "application/json"} # You can change these identifiers if your backend requires pre-registered users/cubes. @@ -88,7 +88,12 @@ def example_01_string_message_minimal(): payload = { "user_id": USER_ID, "writable_cube_ids": [MEM_CUBE_ID], - "messages": "今天心情不错，喝了咖啡。", + "messages": [ + { + "role": "user", + "content": "今天心情不错，喝了咖啡。", + } + ], } call_add_api("01_string_message_minimal", payload) @@ -108,12 +113,7 @@ def example_02_standard_chat_triplet(): "messages": [ { "role": "system", - "content": [ - { - "type": "text", - "text": "You are a helpful travel assistant.", - } - ], + "content": "You are a helpful travel assistant.", "chat_time": "2025-11-24T10:00:00Z", "message_id": "sys-1", }, @@ -160,17 +160,7 @@ def example_03_assistant_with_tool_calls(): "messages": [ { "role": "assistant", - "content": None, - "tool_calls": [ - { - "id": "tool-call-weather-1", - "type": "function", - "function": { - "name": "get_weather", - "arguments": '{"location": "北京"}', - }, - } - ], + "content": '[Tool Call] get_weather({"location": "北京"})', "chat_time": "2025-11-24T10:12:00Z", "message_id": "assistant-with-call-1", } @@ -195,24 +185,13 @@ def example_03b_tool_message_with_result(): "messages": [ { "role": "assistant", - "content": None, - "tool_calls": [ - { - "id": "tool-call-weather-1", - "type": "function", - "function": { - "name": "get_weather", - "arguments": '{"location": "北京"}', - }, - } - ], + "content": '[Tool Call] get_weather({"location": "北京"})', "chat_time": "2025-11-24T10:12:00Z", "message_id": "assistant-with-call-1", }, { - "role": "tool", - "content": "北京今天天气晴朗，温度25°C，湿度60%。", - "tool_call_id": "tool-call-weather-1", + "role": "system", + "content": "[Tool Result] 北京今天天气晴朗，温度25°C，湿度60%。", "chat_time": "2025-11-24T10:12:05Z", "message_id": "tool-result-1", }, @@ -232,31 +211,44 @@ def example_03c_tool_description_input_output(): - `tool_output`: the result/output from the tool execution. - These are alternative formats for representing tool interactions. """ + tool_desc = { + "type": "tool_description", + "name": "get_weather", + "description": "获取指定地点的当前天气信息", + "parameters": { + "type": "object", + "properties": {"location": {"type": "string", "description": "城市名称"}}, + "required": ["location"], + }, + } + tool_input = { + "type": "tool_input", + "call_id": "call_123", + "name": "get_weather", + "argument": {"location": "北京"}, + } + tool_output = { + "type": "tool_output", + "call_id": "call_123", + "name": "get_weather", + "output": {"weather": "晴朗", "temperature": 25, "humidity": 60}, + } + payload = { "user_id": USER_ID, "writable_cube_ids": [MEM_CUBE_ID], "messages": [ { - "type": "tool_description", - "name": "get_weather", - "description": "获取指定地点的当前天气信息", - "parameters": { - "type": "object", - "properties": {"location": {"type": "string", "description": "城市名称"}}, - "required": ["location"], - }, + "role": "system", + "content": f"[Tool Description] {json.dumps(tool_desc, ensure_ascii=False)}", }, { - "type": "tool_input", - "call_id": "call_123", - "name": "get_weather", - "argument": {"location": "北京"}, + "role": "user", + "content": f"[Tool Input] {json.dumps(tool_input, ensure_ascii=False)}", }, { - "type": "tool_output", - "call_id": "call_123", - "name": "get_weather", - "output": {"weather": "晴朗", "temperature": 25, "humidity": 60}, + "role": "system", + "content": f"[Tool Output] {json.dumps(tool_output, ensure_ascii=False)}", }, ], "info": {"source_type": "custom_tool_format"}, @@ -282,20 +274,13 @@ def example_04_extreme_multimodal_single_message(): "messages": [ { "role": "user", - "content": [ - {"type": "text", "text": "请分析下面这些信息："}, - {"type": "image_url", "image_url": {"url": "https://example.com/x.png"}}, - {"type": "file", "file": {"file_id": "f1", "filename": "xx.pdf"}}, - ], + "content": "请分析下面这些信息：\n[Image: https://example.com/x.png]\n[File: xx.pdf]", "chat_time": "2025-11-24T10:55:00Z", "message_id": "mix-mm-1", }, { "role": "user", - "content": [ - {"type": "text", "text": "请再分析一下下面这些信息："}, - {"type": "file", "file": {"file_id": "f1", "filename": "xx.pdf"}}, - ], + "content": "请再分析一下下面这些信息：\n[File: xx.pdf]", "chat_time": "2025-11-24T10:55:10Z", "message_id": "mix-mm-2", }, @@ -323,19 +308,7 @@ def example_05_multimodal_text_and_image(): "messages": [ { "role": "user", - "content": [ - { - "type": "text", - "text": "帮我看看这张图片大概是什么内容？", - }, - { - "type": "image_url", - "image_url": { - "url": "https://example.com/mountain_lake.jpg", - "detail": "high", - }, - }, - ], + "content": "帮我看看这张图片大概是什么内容？\n[Image: https://example.com/mountain_lake.jpg]", "chat_time": "2025-11-24T10:20:00Z", "message_id": "mm-img-1", } @@ -361,19 +334,7 @@ def example_06_multimodal_text_and_file(): "messages": [ { "role": "user", - "content": [ - { - "type": "text", - "text": "请阅读这个PDF，总结里面的要点。", - }, - { - "type": "file", - "file": { - "file_id": "file_123", - "filename": "report.pdf", # optional, but recommended - }, - }, - ], + "content": "请阅读这个PDF，总结里面的要点。\n[File: report.pdf (ID: file_123)]", "chat_time": "2025-11-24T10:21:00Z", "message_id": "mm-file-1", } @@ -396,15 +357,7 @@ def example_07_audio_only_message(): "messages": [ { "role": "user", - "content": [ - { - "type": "input_audio", - "input_audio": { - "data": "base64_encoded_audio_here", - "format": "mp3", - }, - } - ], + "content": "[Audio Input: base64_encoded_audio_here (mp3)]", "chat_time": "2025-11-24T10:22:00Z", "message_id": "audio-1", } @@ -431,12 +384,12 @@ def example_08_pure_text_input_items(): "writable_cube_ids": [MEM_CUBE_ID], "messages": [ { - "type": "text", - "text": "这是一段独立的文本输入，没有明确的对话上下文。", + "role": "user", + "content": "这是一段独立的文本输入，没有明确的对话上下文。", }, { - "type": "text", - "text": "它依然会被抽取和写入明文记忆。", + "role": "user", + "content": "它依然会被抽取和写入明文记忆。", }, ], "info": {"source_type": "batch_import"}, @@ -453,18 +406,15 @@ def example_09_pure_file_input_by_file_id(): * `file_id`: optional, use when file is already uploaded * `file_data`: optional, use for base64-encoded content * `filename`: optional, but recommended for clarity - - In practice, you need at least `file_id` OR `file_data` to specify the file. + - In practice, you need at least `file_id` OR `file_data` to specify the file. """ payload = { "user_id": USER_ID, "writable_cube_ids": [MEM_CUBE_ID], "messages": [ { - "type": "file", - "file": { - "file_id": "file_uploaded_123", # at least one of file_id/file_data needed - "filename": "document.pdf", # optional - }, + "role": "user", + "content": "[File: document.pdf (ID: file_uploaded_123)]", } ], "info": {"source_type": "file_ingestion"}, @@ -487,11 +437,8 @@ def example_09b_pure_file_input_by_file_data(): "writable_cube_ids": [MEM_CUBE_ID], "messages": [ { - "type": "file", - "file": { - "file_data": "base64_encoded_file_content_here", # at least one of file_id/file_data needed - "filename": "document.pdf", # optional - }, + "role": "user", + "content": "[File: document.pdf (base64_data)]", } ], "info": {"source_type": "file_ingestion_base64"}, @@ -512,11 +459,8 @@ def example_09c_pure_file_input_by_oss_url(): "writable_cube_ids": [MEM_CUBE_ID], "messages": [ { - "type": "file", - "file": { - "file_data": "oss_url", # OSS URL instead of base64 - "filename": "document.pdf", - }, + "role": "user", + "content": "[File: document.pdf (OSS: oss_url)]", } ], "info": {"source_type": "file_ingestion_oss"}, @@ -537,11 +481,8 @@ def example_09d_pure_image_input(): "writable_cube_ids": [MEM_CUBE_ID], "messages": [ { - "type": "image_url", - "image_url": { - "url": "https://example.com/standalone_image.jpg", - "detail": "high", - }, + "role": "user", + "content": "[Image: https://example.com/standalone_image.jpg]", } ], "info": {"source_type": "image_ingestion"}, @@ -561,26 +502,7 @@ def example_10_mixed_text_file_image(): "messages": [ { "role": "user", - "content": [ - { - "type": "text", - "text": "请同时分析这个报告和图表。", - }, - { - "type": "file", - "file": { - "file_id": "file_789", - "filename": "analysis_report.pdf", - }, - }, - { - "type": "image_url", - "image_url": { - "url": "https://example.com/chart.png", - "detail": "auto", - }, - }, - ], + "content": "请同时分析这个报告和图表。\n[File: analysis_report.pdf (ID: file_789)]\n[Image: https://example.com/chart.png]", "chat_time": "2025-11-24T10:23:00Z", "message_id": "mixed-1", } @@ -844,6 +766,82 @@ def example_18_add_with_chat_history(): call_add_api("18_add_with_chat_history", payload) +# =========================================================================== +# 8. Search and Chat examples +# =========================================================================== + + +def example_19_search_memories(): + """ + Search memories using `APISearchRequest`. + + - Searches for memories relevant to a query. + - Demonstrates usage of `readable_cube_ids` for scoping. + """ + payload = { + "user_id": USER_ID, + "query": "What are my hotel preferences?", + "readable_cube_ids": [MEM_CUBE_ID], + "top_k": 5, + "mode": "fast", + "include_preference": True, + } + + print("=" * 80) + print("[*] Example: 19_search_memories") + print("- Payload:") + print(json.dumps(payload, indent=2, ensure_ascii=False)) + + try: + resp = requests.post( + f"{BASE_URL}/search", headers=HEADERS, data=json.dumps(payload), timeout=60 + ) + print("- Response:") + print(resp.status_code, resp.text) + except Exception as e: + print(f"- Request failed with exception: {e!r}") + + print("=" * 80) + print() + + +def example_20_chat_complete(): + """ + Chat completion using `APIChatCompleteRequest`. + + - Sends a chat query to the system. + - System retrieves relevant memories and generates a response. + - please make sure ENABLE_CHAT_API=true in .env or environment variables + - and set up CHAT_MODEL_LIST in .env or environment variables properly with api keys and stuff. + """ + payload = { + "user_id": USER_ID, + "query": "Recommend a hotel for me based on my preferences.", + "readable_cube_ids": [MEM_CUBE_ID], + "writable_cube_ids": [MEM_CUBE_ID], + "mode": "fast", + "top_k": 5, + "add_message_on_answer": True, + } + + print("=" * 80) + print("[*] Example: 20_chat_complete") + print("- Payload:") + print(json.dumps(payload, indent=2, ensure_ascii=False)) + + try: + resp = requests.post( + f"{BASE_URL}/chat/complete", headers=HEADERS, data=json.dumps(payload), timeout=60 + ) + print("- Response:") + print(resp.status_code, resp.text) + except Exception as e: + print(f"- Request failed with exception: {e!r}") + + print("=" * 80) + print() + + # =========================================================================== # Entry point # =========================================================================== @@ -873,3 +871,5 @@ def example_18_add_with_chat_history(): example_16_feedback_add() example_17_family_travel_conversation() example_18_add_with_chat_history() + example_19_search_memories() + example_20_chat_complete() diff --git a/examples/basic_modules/embedder.py b/examples/basic_modules/embedder.py index 7cc7942da..d35dc1797 100644 --- a/examples/basic_modules/embedder.py +++ b/examples/basic_modules/embedder.py @@ -3,7 +3,10 @@ # Scenario 1: Using EmbedderFactory - +# Prerequisites: +# 1. Install Ollama: https://ollama.com/ +# 2. Start Ollama server: `ollama serve` +# 3. Pull the model: `ollama pull nomic-embed-text` config = EmbedderConfigFactory.model_validate( { "backend": "ollama", @@ -33,6 +36,9 @@ # Scenario 3: Using SenTranEmbedder +# Prerequisites: +# 1. Ensure `einops` is installed: `pip install einops` (Required for some HF models like nomic-bert) +# 2. The model `nomic-ai/nomic-embed-text-v1.5` will be downloaded automatically from HuggingFace. config_hf = EmbedderConfigFactory.model_validate( { @@ -49,6 +55,9 @@ print("==" * 20) # === Scenario 4: Using UniversalAPIEmbedder(OpenAI) === +# Prerequisites: +# 1. Set a valid OPENAI_API_KEY +# 2. Ensure the base_url is reachable config_api = EmbedderConfigFactory.model_validate( { @@ -68,6 +77,9 @@ print("Embedding preview:", embedding_api[0][:10]) # === Scenario 5: Using UniversalAPIEmbedder(Azure) === +# Prerequisites: +# 1. Set a valid AZURE_API_KEY +# 2. Ensure the base_url is reachable config_api = EmbedderConfigFactory.model_validate( { diff --git a/examples/basic_modules/llm.py b/examples/basic_modules/llm.py index d33fc9544..fb157c991 100644 --- a/examples/basic_modules/llm.py +++ b/examples/basic_modules/llm.py @@ -5,6 +5,10 @@ # Scenario 1: Using LLMFactory with Ollama Backend # This is the most recommended way! 🌟 +# Prerequisites: +# 1. Install Ollama: https://ollama.com/ +# 2. Start Ollama server: `ollama serve` +# 3. Need python ollama package(>=0.5.0,<0.6.0) config = LLMConfigFactory.model_validate( { @@ -46,6 +50,10 @@ # Scenario 3: Using LLMFactory with OpenAI Backend +# Prerequisites: +# 1. You need a valid OpenAI API key to run this scenario. +# 2. Replace 'sk-xxxx' with your actual API key below. + config = LLMConfigFactory.model_validate( { diff --git a/examples/basic_modules/nebular_example.py b/examples/basic_modules/nebular_example.py deleted file mode 100644 index 13f88e3f3..000000000 --- a/examples/basic_modules/nebular_example.py +++ /dev/null @@ -1,361 +0,0 @@ -import json -import os - -from datetime import datetime, timezone - -import numpy as np - -from dotenv import load_dotenv - -from memos.configs.embedder import EmbedderConfigFactory -from memos.configs.graph_db import GraphDBConfigFactory -from memos.embedders.factory import EmbedderFactory -from memos.graph_dbs.factory import GraphStoreFactory -from memos.memories.textual.item import TextualMemoryItem, TreeNodeTextualMemoryMetadata - - -load_dotenv() - - -def show(nebular_data): - from memos.configs.graph_db import Neo4jGraphDBConfig - from memos.graph_dbs.neo4j import Neo4jGraphDB - - tree_config = Neo4jGraphDBConfig.from_json_file("../../examples/data/config/neo4j_config.json") - tree_config.use_multi_db = True - tree_config.db_name = "nebular-show2" - - neo4j_db = Neo4jGraphDB(tree_config) - neo4j_db.clear() - neo4j_db.import_graph(nebular_data) - - -embedder_config = EmbedderConfigFactory.model_validate( - { - "backend": "universal_api", - "config": { - "provider": "openai", - "api_key": os.getenv("OPENAI_API_KEY", "sk-xxxxx"), - "model_name_or_path": "text-embedding-3-large", - "base_url": os.getenv("OPENAI_API_BASE", "https://api.openai.com/v1"), - }, - } -) -embedder = EmbedderFactory.from_config(embedder_config) -embedder_dimension = 3072 - - -def embed_memory_item(memory: str) -> list[float]: - embedding = embedder.embed([memory])[0] - embedding_np = np.array(embedding, dtype=np.float32) - embedding_list = embedding_np.tolist() - return embedding_list - - -def example_shared_db(db_name: str = "shared-traval-group"): - """ - Example: Single(Shared)-DB multi-tenant (logical isolation) - Multiple users' data in the same Neo4j DB with user_name as a tag. - """ - # users - user_list = ["travel_member_alice", "travel_member_bob"] - - for user_name in user_list: - # Step 1: Build factory config - config = GraphDBConfigFactory( - backend="nebular", - config={ - "uri": json.loads(os.getenv("NEBULAR_HOSTS", "localhost")), - "user": os.getenv("NEBULAR_USER", "root"), - "password": os.getenv("NEBULAR_PASSWORD", "xxxxxx"), - "space": db_name, - "user_name": user_name, - "use_multi_db": False, - "auto_create": True, - "embedding_dimension": embedder_dimension, - }, - ) - - # Step 2: Instantiate graph store - graph = GraphStoreFactory.from_config(config) - print(f"\n[INFO] Working in shared DB: {db_name}, for user: {user_name}") - graph.clear() - - # Step 3: Create topic node - topic = TextualMemoryItem( - memory="This research addresses long-term multi-UAV navigation for energy-efficient communication coverage.", - metadata=TreeNodeTextualMemoryMetadata( - memory_type="LongTermMemory", - key="Multi-UAV Long-Term Coverage", - hierarchy_level="topic", - type="fact", - memory_time="2024-01-01", - source="file", - sources=["paper://multi-uav-coverage/intro"], - status="activated", - confidence=95.0, - tags=["UAV", "coverage", "multi-agent"], - entities=["UAV", "coverage", "navigation"], - visibility="public", - updated_at=datetime.now().isoformat(), - embedding=embed_memory_item( - "This research addresses long-term " - "multi-UAV navigation for " - "energy-efficient communication " - "coverage." - ), - ), - ) - - graph.add_node( - id=topic.id, memory=topic.memory, metadata=topic.metadata.model_dump(exclude_none=True) - ) - - # Step 4: Add a concept for each user - concept = TextualMemoryItem( - memory=f"Itinerary plan for {user_name}", - metadata=TreeNodeTextualMemoryMetadata( - memory_type="LongTermMemory", - key="Multi-UAV Long-Term Coverage", - hierarchy_level="concept", - type="fact", - memory_time="2024-01-01", - source="file", - sources=["paper://multi-uav-coverage/intro"], - status="activated", - confidence=95.0, - tags=["UAV", "coverage", "multi-agent"], - entities=["UAV", "coverage", "navigation"], - visibility="public", - updated_at=datetime.now().isoformat(), - embedding=embed_memory_item(f"Itinerary plan for {user_name}"), - ), - ) - - graph.add_node( - id=concept.id, - memory=concept.memory, - metadata=concept.metadata.model_dump(exclude_none=True), - ) - - # Link concept to topic - graph.add_edge(source_id=concept.id, target_id=topic.id, type="RELATE_TO") - print(f"[INFO] Added nodes for {user_name}") - - # Step 5: Query and print ALL for verification - print("\n=== Export entire DB (for verification, includes ALL users) ===") - graph = GraphStoreFactory.from_config(config) - all_graph_data = graph.export_graph() - print(str(all_graph_data)[:1000]) - - # Step 6: Search for alice's data only - print("\n=== Search for travel_member_alice ===") - config_alice = GraphDBConfigFactory( - backend="nebular", - config={ - "uri": json.loads(os.getenv("NEBULAR_HOSTS", "localhost")), - "user": os.getenv("NEBULAR_USER", "root"), - "password": os.getenv("NEBULAR_PASSWORD", "xxxxxx"), - "space": db_name, - "user_name": user_list[0], - "auto_create": True, - "embedding_dimension": embedder_dimension, - "use_multi_db": False, - }, - ) - graph_alice = GraphStoreFactory.from_config(config_alice) - nodes = graph_alice.search_by_embedding(vector=embed_memory_item("travel itinerary"), top_k=3) - for node in nodes: - print(str(graph_alice.get_node(node["id"]))[:1000]) - - -def run_user_session( - user_name: str, - db_name: str, - topic_text: str, - concept_texts: list[str], - fact_texts: list[str], -): - print(f"\n=== {user_name} starts building their memory graph ===") - - # Manually initialize correct GraphDB class - config = GraphDBConfigFactory( - backend="nebular", - config={ - "uri": json.loads(os.getenv("NEBULAR_HOSTS", "localhost")), - "user": os.getenv("NEBULAR_USER", "root"), - "password": os.getenv("NEBULAR_PASSWORD", "xxxxxx"), - "space": db_name, - "user_name": user_name, - "use_multi_db": False, - "auto_create": True, - "embedding_dimension": embedder_dimension, - }, - ) - graph = GraphStoreFactory.from_config(config) - - # Start with a clean slate for this user - graph.clear() - - now = datetime.now(timezone.utc).isoformat() - - # === Step 1: Create a root topic node (e.g., user's research focus) === - topic = TextualMemoryItem( - memory=topic_text, - metadata=TreeNodeTextualMemoryMetadata( - memory_type="LongTermMemory", - key="Research Topic", - hierarchy_level="topic", - type="fact", - memory_time="2024-01-01", - status="activated", - visibility="public", - tags=["research", "rl"], - updated_at=now, - embedding=embed_memory_item(topic_text), - ), - ) - graph.add_node(topic.id, topic.memory, topic.metadata.model_dump(exclude_none=True)) - - # === Step 2: Create two concept nodes linked to the topic === - concept_items = [] - for i, text in enumerate(concept_texts): - concept = TextualMemoryItem( - memory=text, - metadata=TreeNodeTextualMemoryMetadata( - memory_type="LongTermMemory", - key=f"Concept {i + 1}", - hierarchy_level="concept", - type="fact", - memory_time="2024-01-01", - status="activated", - visibility="public", - updated_at=now, - embedding=embed_memory_item(text), - tags=["concept"], - confidence=90 + i, - ), - ) - graph.add_node(concept.id, concept.memory, concept.metadata.model_dump(exclude_none=True)) - graph.add_edge(topic.id, concept.id, type="PARENT") - concept_items.append(concept) - - # === Step 3: Create supporting facts under each concept === - for i, text in enumerate(fact_texts): - fact = TextualMemoryItem( - memory=text, - metadata=TreeNodeTextualMemoryMetadata( - memory_type="WorkingMemory", - key=f"Fact {i + 1}", - hierarchy_level="fact", - type="fact", - memory_time="2024-01-01", - status="activated", - visibility="public", - updated_at=now, - embedding=embed_memory_item(text), - confidence=85.0, - tags=["fact"], - ), - ) - graph.add_node(fact.id, fact.memory, fact.metadata.model_dump(exclude_none=True)) - graph.add_edge(concept_items[i % len(concept_items)].id, fact.id, type="PARENT") - - # === Step 4: Retrieve memory using semantic search === - vector = embed_memory_item("How is memory retrieved?") - search_result = graph.search_by_embedding(vector, top_k=2) - for r in search_result: - node = graph.get_node(r["id"]) - print("🔍 Search result:", node["memory"]) - - # === Step 5: Tag-based neighborhood discovery === - neighbors = graph.get_neighbors_by_tag(["concept"], exclude_ids=[], top_k=2) - print("📎 Tag-related nodes:", [neighbor["memory"] for neighbor in neighbors]) - - # === Step 6: Retrieve children (facts) of first concept === - children = graph.get_children_with_embeddings(concept_items[0].id) - print("📍 Children of concept:", [child["memory"] for child in children]) - - # === Step 7: Export a local subgraph and grouped statistics === - subgraph = graph.get_subgraph(topic.id, depth=2) - print("📌 Subgraph node count:", len(subgraph["neighbors"])) - - stats = graph.get_grouped_counts(["memory_type", "status"]) - print("📊 Grouped counts:", stats) - - # === Step 8: Demonstrate updates and cleanup === - graph.update_node( - concept_items[0].id, {"confidence": 99.0, "created_at": "2025-07-24T20:11:56.375687"} - ) - graph.remove_oldest_memory("WorkingMemory", keep_latest=1) - graph.delete_edge(topic.id, concept_items[0].id, type="PARENT") - graph.delete_node(concept_items[1].id) - - # === Step 9: Export and re-import the entire graph structure === - exported = graph.export_graph() - graph.import_graph(exported) - print("📦 Graph exported and re-imported, total nodes:", len(exported["nodes"])) - - # ==================================== - # 🔍 Step 10: extra function - # ==================================== - print(f"\n=== 🔍 Extra Tests for user: {user_name} ===") - - print(" - Memory count:", graph.get_memory_count("LongTermMemory")) - print(" - Node count:", graph.count_nodes("LongTermMemory")) - print(" - All LongTermMemory items:", graph.get_all_memory_items("LongTermMemory")) - - if len(exported["edges"]) > 0: - n1, n2 = exported["edges"][0]["source"], exported["edges"][0]["target"] - print(" - Edge exists?", graph.edge_exists(n1, n2, exported["edges"][0]["type"])) - print(" - Edges for node:", graph.get_edges(n1)) - - filters = [{"field": "memory_type", "op": "=", "value": "LongTermMemory"}] - print(" - Metadata query result:", graph.get_by_metadata(filters)) - print( - " - Optimization candidates:", graph.get_structure_optimization_candidates("LongTermMemory") - ) - try: - graph.drop_database() - except ValueError as e: - print(" - drop_database raised ValueError as expected:", e) - - -def example_complex_shared_db(db_name: str = "shared-traval-group-complex"): - # User 1: Alice explores structured memory for LLMs - run_user_session( - user_name="alice", - db_name=db_name, - topic_text="Alice studies structured memory and long-term memory optimization in LLMs.", - concept_texts=[ - "Short-term memory can be simulated using WorkingMemory blocks.", - "A structured memory graph improves retrieval precision for agents.", - ], - fact_texts=[ - "Embedding search is used to find semantically similar memory items.", - "User memories are stored as node-edge structures that support hierarchical reasoning.", - ], - ) - - # User 2: Bob focuses on GNN-based reasoning - run_user_session( - user_name="bob", - db_name=db_name, - topic_text="Bob investigates how graph neural networks can support knowledge reasoning.", - concept_texts=[ - "GNNs can learn high-order relations among entities.", - "Attention mechanisms in graphs improve inference precision.", - ], - fact_texts=[ - "GAT outperforms GCN in graph classification tasks.", - "Multi-hop reasoning helps answer complex queries.", - ], - ) - - -if __name__ == "__main__": - print("\n=== Example: Single-DB ===") - example_shared_db(db_name="shared_traval_group-new") - - print("\n=== Example: Single-DB-Complex ===") - example_complex_shared_db(db_name="shared-traval-group-complex-new2") diff --git a/examples/basic_modules/neo4j_example.py b/examples/basic_modules/neo4j_example.py index ea68975cc..e1c0df317 100644 --- a/examples/basic_modules/neo4j_example.py +++ b/examples/basic_modules/neo4j_example.py @@ -27,6 +27,23 @@ def embed_memory_item(memory: str) -> list[float]: return embedder.embed([memory])[0] +def get_neo4j_graph(db_name: str = "paper"): + config = GraphDBConfigFactory( + backend="neo4j", + config={ + "uri": "bolt://xxxx:7687", + "user": "neo4j", + "password": "xxxx", + "db_name": db_name, + "auto_create": True, + "embedding_dimension": 3072, + "use_multi_db": True, + }, + ) + graph = GraphStoreFactory.from_config(config) + return graph + + def example_multi_db(db_name: str = "paper"): # Step 1: Build factory config config = GraphDBConfigFactory( @@ -537,6 +554,53 @@ def example_complex_shared_db(db_name: str = "shared-traval-group-complex", comm ) +def example_complex_shared_db_search_filter(db): + embedding = embed_memory_item( + "The reward function combines " + "multiple objectives: coverage " + "maximization, energy consumption " + ) + print(f"get_node:{db.get_node(id='5364c28e-1e4b-485a-b1d5-1ba11bc5bc8b')}") + + filter_id = {"id": "a269f2bf-f4a2-43b9-aa8d-1cb2a2eb4691"} + print(f"==filter_id:{db.search_by_embedding(vector=embedding, filter=filter_id)}") + + filter_and_params = { + "and": [{"id": "a269f2bf-f4a2-43b9-aa8d-1cb2a2eb4691"}, {"source": "file123"}] + } + print( + f"==filter_and_params:{db.search_by_embedding(vector=embedding, filter=filter_and_params)}" + ) + + filter_or_params = {"or": [{"id": "a269f2bf-f4a2-43b9-aa8d-1cb2a2eb4691"}, {"id": "xxxxxxxx"}]} + print(f"==filter_or_params:{db.search_by_embedding(vector=embedding, filter=filter_or_params)}") + filter_like_params = { + "and": [ + {"memory_type": {"like": "LongTermMemory"}}, + ] + } + print( + f"==filter_like_params:{db.search_by_embedding(vector=embedding, filter=filter_like_params)}" + ) + + """ + cypher_op_map = {"gt": ">", "lt": "<", "gte": ">=", "lte": "<="} + """ + filter_lt_params = { + "and": [ + {"created_at": {"gt": "2025-11-29"}}, + ] + } + print(f"==filter_lt_params:{db.search_by_embedding(vector=embedding, filter=filter_lt_params)}") + + +def example_complex_shared_db_delete_memory(db): + print("delete node") + db.delete_node(id="582de45f-8f99-4006-8062-76eea5649d94") + print("delete edge") + db.delete_edge(source_id=1, target_id=2, type="PARENT", user_name="") + + if __name__ == "__main__": print("\n=== Example: Multi-DB ===") example_multi_db(db_name="paper") @@ -544,8 +608,17 @@ def example_complex_shared_db(db_name: str = "shared-traval-group-complex", comm print("\n=== Example: Single-DB ===") example_shared_db(db_name="shared-traval-group") + print("\n=== Example: Single-DB ===") + example_shared_db(db_name="shared-traval-group") + print("\n=== Example: Single-DB-Complex ===") example_complex_shared_db(db_name="shared-traval-group-complex-new") print("\n=== Example: Single-Community-DB-Complex ===") example_complex_shared_db(db_name="paper", community=True) + + print("\n=== Example: Single-DB-Complex searchFilter ===") + db = get_neo4j_graph(db_name="paper") + example_complex_shared_db_search_filter(db) + + example_complex_shared_db_delete_memory(db) diff --git a/examples/basic_modules/parser.py b/examples/basic_modules/parser.py deleted file mode 100644 index c063964b7..000000000 --- a/examples/basic_modules/parser.py +++ /dev/null @@ -1,15 +0,0 @@ -from memos.configs.parser import ParserConfigFactory -from memos.parsers.factory import ParserFactory - - -config = ParserConfigFactory.model_validate( - { - "backend": "markitdown", - "config": {}, - } -) -parser = ParserFactory.from_config(config) -file_path = "README.md" -markdown_text = parser.parse(file_path) -print("Markdown text:\n", markdown_text) -print("==" * 20) diff --git a/examples/basic_modules/qdrant_vec_db.py b/examples/basic_modules/qdrant_vec_db.py deleted file mode 100644 index a16cd95ee..000000000 --- a/examples/basic_modules/qdrant_vec_db.py +++ /dev/null @@ -1,143 +0,0 @@ -import uuid - -from memos import settings -from memos.configs.vec_db import VectorDBConfigFactory -from memos.vec_dbs.factory import VecDBFactory - - -config = VectorDBConfigFactory.model_validate( - { - "backend": "qdrant", - "config": { - "collection_name": "test_collection", - "vector_dimension": 4, - "distance_metric": "cosine", - "path": str(settings.MEMOS_DIR / "qdrant"), - }, - } -) -vec_db = VecDBFactory.from_config(config) - -# ============== BATCH OPERATIONS ============== -print("\n" + "=" * 50) -print("BATCH DOCUMENT ADDITION") - -# Add batch data -batch_data = [ - { - "id": str(uuid.uuid4()), - "vector": [0.5, 0.6, 0.7, 0.8], - "payload": {"text": "Document A", "category": "type1"}, - }, - { - "id": str(uuid.uuid4()), - "vector": [0.4, 0.5, 0.6, 0.7], - "payload": {"text": "Document B", "category": "type2"}, - }, - { - "id": str(uuid.uuid4()), - "vector": [0.3, 0.4, 0.5, 0.6], - "payload": {"text": "Document C", "category": "type1"}, - }, -] -vec_db.add(batch_data) -print(f"✓ Added {len(batch_data)} documents") - -# ============== SEARCH OPERATIONS ============== -print("\n" + "=" * 50) -print("VECTOR SEARCH") - -# Search for similar items -query_vector = [5.0, 6.0, 7.0, 8.0] -results = vec_db.search(query_vector, top_k=2) -print(f"Query vector: {query_vector}") -print("\nResults:") -for i, result in enumerate(results, 1): - print(f" {i}. ID: {result.id}") - print(f" Score: {result.score}") - print(f" Payload: {result.payload}") - -# ============== COUNT OPERATIONS ============== -print("\n" + "=" * 50) -print("DOCUMENT COUNT") - -# Count documents in collection -count = vec_db.count() -print(f"Total documents in collection: {count}") - -# Count documents with filter -filtered_count = vec_db.count(filter={"category": "type1"}) -print(f"Documents with category 'type1': {filtered_count}") - -# ============== SINGLE DOCUMENT OPERATIONS ============== -print("\n" + "=" * 50) -print("DOCUMENT OPERATIONS") - -# Add a document -doc_id = str(uuid.uuid4()) -vec_db.add( - [ - { - "id": doc_id, - "vector": [0.1, 0.2, 0.3, 0.4], - "payload": {"text": "Original document", "status": "new"}, - } - ] -) -print(f"✓ Added document with ID: {doc_id}") - -# Update document payload -vec_db.update(doc_id, {"payload": {"text": "Updated document", "status": "updated"}}) -print(f"✓ Updated document payload for ID: {doc_id}") - -# Retrieve updated document -result = vec_db.get_by_id(doc_id) -print("\nRetrieved updated document:") -print(f" ID: {doc_id}") -print(f" Payload: {result.payload if result else 'Not found'}") - -# Delete the document -vec_db.delete([doc_id]) -print(f"\n✓ Deleted document with ID: {doc_id}") - -# Verify deletion -result = vec_db.get_by_id(doc_id) -print("\nDocument after deletion:") -print(f" Result: {'Not found' if result is None else result}") - -# ============== COLLECTION OPERATIONS ============== -print("\n" + "=" * 50) -print("COLLECTION OPERATIONS") - -# List all collections in the database -collections = vec_db.list_collections() -print(f"Available collections: {collections}") - -# ============== FILTER OPERATIONS ============== -print("\n" + "=" * 50) -print("FILTER OPERATIONS") - -# Get documents by filter criteria -filter_results = vec_db.get_by_filter({"category": "type1"}) -print("Documents filtered by category 'type1':") -for i, item in enumerate(filter_results, 1): - print(f" {i}. ID: {item.id}") - print(f" Payload: {item.payload}") - -# Get all documents in the collection -all_docs = vec_db.get_all() -print("\nAll documents in the collection:") -for i, item in enumerate(all_docs, 1): - print(f" {i}. ID: {item.id}") - print(f" Vector: {item.vector}") - print(f" Payload: {item.payload}") - -# ============== CLEANUP ============== -print("\n" + "=" * 50) -print("CLEANUP") - -# Delete the collection -vec_db.delete_collection("test_collection") -print("✓ Collection deleted") -print(f"Available collections after deletion: {vec_db.list_collections()}") -print("\n" + "=" * 50) diff --git a/examples/basic_modules/reranker.py b/examples/basic_modules/reranker.py index 47bf1405c..e5a869789 100644 --- a/examples/basic_modules/reranker.py +++ b/examples/basic_modules/reranker.py @@ -50,6 +50,7 @@ def show_ranked(title: str, ranked: list[tuple[TextualMemoryItem, float]], top_n def main(): # ------------------------------- # 1) Build the embedder (real vectors) + # You may need to set valid OPENAI_API_KEY and OPENAI_API_BASE in your environment variables. # ------------------------------- embedder_cfg = EmbedderConfigFactory.model_validate( { @@ -62,6 +63,22 @@ def main(): }, } ) + """ + # ------------------------------- + # Optional: Build the embedder (using local sentence-transformers) + # ------------------------------- + # Use a local model so no API key is required. + embedder_cfg = EmbedderConfigFactory.model_validate( + { + "backend": "sentence_transformer", + "config": { + "model_name_or_path": "nomic-ai/nomic-embed-text-v1.5", + "trust_remote_code": True, + }, + } + ) + """ + embedder = EmbedderFactory.from_config(embedder_cfg) # ------------------------------- diff --git a/examples/basic_modules/textual_memory_internet_search_example.py b/examples/basic_modules/textual_memory_internet_search_example.py index f26d3afd8..9007d7e67 100644 --- a/examples/basic_modules/textual_memory_internet_search_example.py +++ b/examples/basic_modules/textual_memory_internet_search_example.py @@ -1,7 +1,49 @@ """ -Example: Using InternetRetrieverFactory with BochaAISearchRetriever +Textual Memory Internet Search Example +======================================= + +This example demonstrates how to use MemOS's InternetRetrieverFactory to search +the web and retrieve relevant information as memory items. + +**What you'll learn:** +- How to initialize an embedder for web content embedding +- How to configure and use BochaAI web search retriever +- How to configure and use Google Custom Search API +- How to chunk and process web content into memory items +- How to retrieve structured information from internet searches + +**Use case:** +When you need to answer questions that require real-time web information +(e.g., "What's in Alibaba's 2024 ESG report?"), this retriever can: +1. Search the web using BochaAI API or Google Custom Search API +2. Fetch and parse web page content +3. Chunk the content into manageable pieces +4. Return structured memory items with embeddings + +**Prerequisites:** +- Valid BochaAI API Key (set in environment variable: BOCHA_API_KEY) +- (Optional) Valid Google API Key and Search Engine ID for Google Custom Search + - GOOGLE_API_KEY: Get from https://console.cloud.google.com/ + - GOOGLE_SEARCH_ENGINE_ID: Get from https://programmablesearchengine.google.com/ +- Embedder service running (e.g., Ollama with nomic-embed-text) +- Internet connection for web searches + +Run this example: + # Basic test with BochaAI + export BOCHA_API_KEY='sk-your-bocha-api-key' + python examples/basic_modules/textual_memory_internet_search_example.py + + # Test with both BochaAI and Google + export BOCHA_API_KEY='sk-your-bocha-api-key' + export GOOGLE_API_KEY='your-google-api-key' + export GOOGLE_SEARCH_ENGINE_ID='your-search-engine-id' + python examples/basic_modules/textual_memory_internet_search_example.py """ +import json +import os + +from memos import log from memos.configs.embedder import EmbedderConfigFactory from memos.configs.internet_retriever import InternetRetrieverConfigFactory from memos.embedders.factory import EmbedderFactory @@ -10,43 +52,258 @@ ) -# ========= 1. Create an embedder ========= -embedder_config = EmbedderConfigFactory.model_validate( - { - "backend": "ollama", # Or "sentence_transformer", etc. - "config": { - "model_name_or_path": "nomic-embed-text:latest", - }, - } -) +logger = log.get_logger(__name__) + +# ============================================================================ +# Step 0: Setup - Load configuration files +# ============================================================================ +print("=" * 80) +print("Textual Memory Internet Search Example") +print("=" * 80) + +current_dir = os.path.dirname(os.path.abspath(__file__)) +config_dir = os.path.join(current_dir, "../data/config") + +# Load the shared tree-text memory configuration +config_path = os.path.join(config_dir, "tree_config_shared_database.json") +with open(config_path) as f: + config_data = json.load(f) + +print(f"\n✓ Loaded configuration from: {config_path}") + +# ============================================================================ +# Step 1: Initialize Embedder +# ============================================================================ +print("\n[Step 1] Initializing embedder for web content...") + +# The embedder will convert web content into vector embeddings +embedder_config = EmbedderConfigFactory.model_validate(config_data["embedder"]) embedder = EmbedderFactory.from_config(embedder_config) -# ========= 2. Create retriever config for BochaAI ========= +print(f"✓ Embedder initialized: {embedder_config.backend}") + +# ============================================================================ +# Step 2: Configure Internet Retriever (BochaAI) +# ============================================================================ +print("\n[Step 2] Configuring internet retriever...") + +# Load the simple_struct reader configuration +reader_config_path = os.path.join(config_dir, "simple_struct_reader_config.json") +with open(reader_config_path) as f: + reader_config_data = json.load(f) + +print(f"✓ Loaded reader configuration from: {reader_config_path}") + +# NOTE: You need to set your BochaAI API key here or in environment variable +# For this example, we'll read from environment variable +bocha_api_key = os.environ.get("BOCHA_API_KEY", "sk-your-bocha-api-key-here") + +if bocha_api_key == "sk-your-bocha-api-key-here": + print("⚠️ Warning: Using placeholder API key. Set BOCHA_API_KEY environment variable.") + retriever_config = InternetRetrieverConfigFactory.model_validate( { "backend": "bocha", "config": { - "api_key": "sk-xxx", # Your BochaAI API Key - "max_results": 5, - "reader": { # Reader config for chunking web content + "api_key": bocha_api_key, + "max_results": 5, # Maximum number of search results to retrieve + "reader": { + # The reader chunks web content into memory items "backend": "simple_struct", - "config": { # your simple struct reader config - }, + "config": reader_config_data, # Use loaded configuration }, }, } ) -# ========= 3. Build retriever instance via factory ========= +print(f"✓ Retriever configured: {retriever_config.backend}") +print(f" Max results per search: {retriever_config.config.max_results}") + +# ============================================================================ +# Step 3: Create Retriever Instance +# ============================================================================ +print("\n[Step 3] Creating internet retriever instance...") + retriever = InternetRetrieverFactory.from_config(retriever_config, embedder) -# ========= 4. Run BochaAI Web Search ========= -print("=== Scenario 1: Web Search (BochaAI) ===") -query_web = "Alibaba 2024 ESG report" -results_web = retriever.retrieve_from_internet(query_web) +print("✓ Retriever initialized and ready") + +# ============================================================================ +# Step 4: Perform Web Search +# ============================================================================ +print("\n[Step 4] Performing web search...") + +# Define the search query +query = "Alibaba 2024 ESG report" +print(f" 🔍 Query: '{query}'") +print(" ⏳ Searching the web and processing results...\n") + +# Execute the search +# This will: +# 1. Search using BochaAI API +# 2. Fetch web page content +# 3. Parse and chunk the content +# 4. Generate embeddings for each chunk +# 5. Return as TextualMemoryItem objects +results = retriever.retrieve_from_internet(query) + +print("✓ Search completed!") +print(f"✓ Retrieved {len(results)} memory items from web search\n") + +# ============================================================================ +# Step 5: Display Results +# ============================================================================ +print("=" * 80) +print("WEB SEARCH RESULTS") +print("=" * 80) + +if not results: + print("\n❌ No results found.") + print(" This might indicate:") + print(" - Invalid or missing BochaAI API key") + print(" - Network connectivity issues") + print(" - The query returned no relevant web pages") + print(" - The web content couldn't be parsed") +else: + for idx, item in enumerate(results, 1): + print(f"\n[Result #{idx}]") + print("-" * 80) + + # Display the memory content (truncated for readability) + content = item.memory + if len(content) > 300: + print(f"Content: {content[:300]}...") + print(f" (... {len(content) - 300} more characters)") + else: + print(f"Content: {content}") + + # Display metadata if available + if hasattr(item, "metadata") and item.metadata: + metadata = item.metadata + if hasattr(metadata, "sources") and metadata.sources: + print(f"Source: {metadata.sources[0] if metadata.sources else 'N/A'}") + + print() + +print("=" * 80) +print("Example completed successfully!") +print("=" * 80) +print("\n💡 Next steps:") +print(" - Set your BochaAI API key in environment variable: export BOCHA_API_KEY='sk-...'") +print(" - Try different search queries to test various topics") +print(" - Adjust max_results in config to control number of results") +print(" - Use the retrieved memory items in your retrieval pipeline") +print(" - Combine internet search with local memory retrieval for hybrid systems\n") + +print("\n⚠️ Note:") +print(" If you see 'No results found', make sure:") +print(" 1. Your BochaAI API key is valid and set correctly") +print(" 2. You have internet connectivity") +print(" 3. The embedder service is running\n") + +# ============================================================================ +# Step 6: Test Google Custom Search API (Optional) +# ============================================================================ +print("\n" + "=" * 80) +print("GOOGLE CUSTOM SEARCH API TEST") +print("=" * 80) + +# NOTE: You need to set your Google API key and Search Engine ID +# Get these from: https://developers.google.com/custom-search/v1/overview +google_api_key = os.environ.get("GOOGLE_API_KEY", "") +google_search_engine_id = os.environ.get("GOOGLE_SEARCH_ENGINE_ID", "") + +if google_api_key and google_search_engine_id: + print("\n[Step 6.1] Configuring Google Custom Search retriever...") + + google_retriever_config = InternetRetrieverConfigFactory.model_validate( + { + "backend": "google", + "config": { + "api_key": google_api_key, + "search_engine_id": google_search_engine_id, + "max_results": 5, # Maximum number of search results to retrieve + "num_per_request": 5, # Number of results per API request (max 10 for Google) + }, + } + ) + + print("✓ Google retriever configured") + print(f" Max results: {google_retriever_config.config.max_results}") + + print("\n[Step 6.2] Creating Google retriever instance...") + google_retriever = InternetRetrieverFactory.from_config(google_retriever_config, embedder) + print("✓ Google retriever initialized") + + print("\n[Step 6.3] Performing Google web search...") + google_query = "Python best practices 2024" + print(f" 🔍 Query: '{google_query}'") + print(" ⏳ Searching via Google Custom Search API...\n") + + google_results = google_retriever.retrieve_from_internet(google_query) + + print("✓ Google search completed!") + print(f"✓ Retrieved {len(google_results)} memory items from Google search\n") + + # Display Google search results + print("=" * 80) + print("GOOGLE SEARCH RESULTS") + print("=" * 80) + + if not google_results: + print("\n❌ No results found from Google.") + print(" This might indicate:") + print(" - Invalid Google API key or Search Engine ID") + print(" - API quota exceeded") + print(" - Network connectivity issues") + else: + for idx, item in enumerate(google_results, 1): + print(f"\n[Google Result #{idx}]") + print("-" * 80) + + # Display the memory content (truncated for readability) + content = item.memory + if len(content) > 300: + print(f"Content: {content[:300]}...") + print(f" (... {len(content) - 300} more characters)") + else: + print(f"Content: {content}") + + # Display metadata if available + if hasattr(item, "metadata") and item.metadata: + metadata = item.metadata + if hasattr(metadata, "sources") and metadata.sources: + print(f"Source: {metadata.sources[0] if metadata.sources else 'N/A'}") + + print() -print(f"Retrieved {len(results_web)} memory items.") -for idx, item in enumerate(results_web, 1): - print(f"[{idx}] {item.memory[:500]}...") # preview first 100 chars + print("=" * 80) + print("Google Search Test completed!") + print("=" * 80) +else: + print("\n⏭️ Skipping Google Custom Search API test") + print(" To enable this test, set the following environment variables:") + print(" - GOOGLE_API_KEY: Your Google API key") + print(" - GOOGLE_SEARCH_ENGINE_ID: Your Google Custom Search Engine ID (cx parameter)") + print("\n Get your credentials from:") + print(" https://developers.google.com/custom-search/v1/overview") -print("==" * 20) +print("\n" + "=" * 80) +print("ALL TESTS COMPLETED") +print("=" * 80) +print("\n💡 Summary:") +print(" ✓ Tested BochaAI web search retriever") +if google_api_key and google_search_engine_id: + print(" ✓ Tested Google Custom Search API") +else: + print(" ⏭️ Skipped Google Custom Search API (credentials not set)") +print("\n💡 Quick Start:") +print(" # Set BochaAI API key") +print(" export BOCHA_API_KEY='sk-your-bocha-api-key'") +print(" ") +print(" # Set Google Custom Search credentials (optional)") +print(" export GOOGLE_API_KEY='your-google-api-key'") +print(" export GOOGLE_SEARCH_ENGINE_ID='your-search-engine-id'") +print(" ") +print(" # Run the example") +print(" python examples/basic_modules/textual_memory_internet_search_example.py\n") diff --git a/examples/basic_modules/tree_textual_memory_reasoner.py b/examples/basic_modules/tree_textual_memory_reasoner.py deleted file mode 100644 index 369787458..000000000 --- a/examples/basic_modules/tree_textual_memory_reasoner.py +++ /dev/null @@ -1,167 +0,0 @@ -from memos import log -from memos.configs.embedder import EmbedderConfigFactory -from memos.configs.graph_db import GraphDBConfigFactory -from memos.configs.llm import LLMConfigFactory -from memos.embedders.factory import EmbedderFactory -from memos.graph_dbs.factory import GraphStoreFactory -from memos.llms.factory import LLMFactory -from memos.memories.textual.item import TextualMemoryItem, TreeNodeTextualMemoryMetadata -from memos.memories.textual.tree_text_memory.retrieve.reasoner import MemoryReasoner -from memos.memories.textual.tree_text_memory.retrieve.retrieval_mid_structs import ParsedTaskGoal - - -logger = log.get_logger(__name__) - -embedder_config = EmbedderConfigFactory.model_validate( - { - "backend": "ollama", - "config": { - "model_name_or_path": "nomic-embed-text:latest", - }, - } -) -embedder = EmbedderFactory.from_config(embedder_config) - -# Step 1: Load LLM config and instantiate -config = LLMConfigFactory.model_validate( - { - "backend": "ollama", - "config": { - "model_name_or_path": "qwen3:0.6b", - "temperature": 0.7, - "max_tokens": 1024, - }, - } -) -llm = LLMFactory.from_config(config) - -# Step 1: Prepare a mock ParsedTaskGoal -parsed_goal = ParsedTaskGoal( - memories=[ - "Multi-UAV Long-Term Coverage", - "Coverage Metrics", - "Reward Function Design", - "Energy Model", - "CT and FT Definition", - "Reward Components", - "Energy Cost Components", - ], - keys=["UAV", "coverage", "energy", "reward"], - tags=[], - goal_type="explanation", -) - -query = "How can multiple UAVs coordinate to maximize coverage while saving energy?" -query_embedding = embedder.embed([query])[0] - - -# Step 2: Initialize graph store -graph_config = GraphDBConfigFactory( - backend="neo4j", - config={ - "uri": "bolt://localhost:7687", - "user": "neo4j", - "password": "12345678", - "db_name": "user06alice", - "auto_create": True, - }, -) -graph_store = GraphStoreFactory.from_config(graph_config) - -ranked_memories = [ - TextualMemoryItem( - id="a88db9ce-3c77-4e83-8d61-aa9ef95c957e", - memory="Coverage performance is measured using CT (Coverage Time) and FT (Fairness Time) metrics.", - metadata=TreeNodeTextualMemoryMetadata( - user_id=None, - session_id=None, - status="activated", - type="fact", - memory_time="2024-01-01", - source="file", - confidence=91.0, - entities=["CT", "FT"], - tags=["coverage", "fairness", "metrics"], - visibility="public", - updated_at="2025-06-11T11:51:24.438001", - memory_type="LongTermMemory", - key="Coverage Metrics", - value="CT and FT used for long-term area and fairness evaluation", - hierarchy_level="concept", - sources=["paper://multi-uav-coverage/metrics"], - embedding=[0.01] * 768, - ), - ), - TextualMemoryItem( - id="c34f5e6b-2d34-4e6f-8c9b-abcdef123456", - memory="The capital of France is Paris, which is known for the Eiffel Tower.", - metadata=TreeNodeTextualMemoryMetadata( - user_id=None, - session_id=None, - status="activated", - type="fact", - memory_time="2024-01-01", - source="file", - confidence=90.0, - entities=["France", "Paris", "Eiffel Tower"], - tags=["geography", "city", "landmark"], - visibility="public", - updated_at="2025-06-11T11:51:24.438001", - memory_type="LongTermMemory", - key="Geography Fact", - value="Paris is the capital of France", - hierarchy_level="concept", - sources=["wikipedia://paris"], - embedding=[0.03] * 768, - ), - ), - TextualMemoryItem( - id="d56a7b8c-3e45-4f7a-9dab-fedcba654321", - memory="Total energy cost is calculated from both mechanical movement and communication transmission.", - metadata=TreeNodeTextualMemoryMetadata( - user_id=None, - session_id=None, - status="activated", - type="fact", - memory_time="2024-01-01", - source="file", - confidence=89.0, - entities=["movement power", "transmission power"], - tags=["energy", "movement", "transmission"], - visibility="public", - updated_at="2025-06-11T11:51:24.438001", - memory_type="LongTermMemory", - key="Energy Cost Components", - value="Includes movement and communication energy", - hierarchy_level="fact", - sources=["paper://multi-uav-coverage/energy-detail"], - embedding=[0.04] * 768, - ), - ), -] - -# Step 7: Init memory retriever -reasoner = MemoryReasoner(llm=llm) - - -# Step 8: Print retrieved memory items before ranking -print("\n=== Retrieved Memory Items (Before Rerank) ===") -for idx, item in enumerate(ranked_memories): - print(f"[Original #{idx + 1}] ID: {item.id}") - print(f"Memory: {item.memory[:200]}...\n") - -# Step 9: Rerank -reasoned_memories = reasoner.reason( - query=query, - ranked_memories=ranked_memories, - parsed_goal=parsed_goal, -) - -# Step 10: Print ranked reasoned memory items with original positions -print("\n=== Memory Items After Reason (Sorted) ===") -id_to_original_rank = {item.id: i + 1 for i, item in enumerate(ranked_memories)} - -for idx, item in enumerate(reasoned_memories): - original_rank = id_to_original_rank.get(item.id, "-") - print(f"[Reasoned #{idx + 1}] ID: {item.id} (Original #{original_rank})") - print(f"Memory: {item.memory[:200]}...\n") diff --git a/examples/basic_modules/tree_textual_memory_recall.py b/examples/basic_modules/tree_textual_memory_recall.py index 935513784..10831cbeb 100644 --- a/examples/basic_modules/tree_textual_memory_recall.py +++ b/examples/basic_modules/tree_textual_memory_recall.py @@ -1,26 +1,152 @@ +""" +Tree Textual Memory Recall Example +=================================== + +This example demonstrates how to use MemOS's GraphMemoryRetriever to recall memories +from a shared graph database. + +**What you'll learn:** +- How to load embedder and graph database configurations +- How to insert memories into the graph store with embeddings +- How to build a ParsedTaskGoal to guide retrieval +- How to retrieve relevant memories using hybrid search + +**Use case:** +You have stored various long-term memories about a user (e.g., "Caroline") +in a graph database, and now you want to answer a natural language question +by retrieving the most relevant memories. + +Run this example: + python examples/basic_modules/tree_textual_memory_recall.py +""" + +import json +import os + from memos import log from memos.configs.embedder import EmbedderConfigFactory from memos.configs.graph_db import GraphDBConfigFactory from memos.embedders.factory import EmbedderFactory from memos.graph_dbs.factory import GraphStoreFactory -from memos.memories.textual.item import TextualMemoryItem +from memos.memories.textual.item import TextualMemoryItem, TreeNodeTextualMemoryMetadata from memos.memories.textual.tree_text_memory.retrieve.recall import GraphMemoryRetriever from memos.memories.textual.tree_text_memory.retrieve.retrieval_mid_structs import ParsedTaskGoal logger = log.get_logger(__name__) -embedder_config = EmbedderConfigFactory.model_validate( - { - "backend": "ollama", - "config": { - "model_name_or_path": "nomic-embed-text:latest", - }, - } -) +# ============================================================================ +# Step 0: Setup - Load configuration files +# ============================================================================ +print("=" * 70) +print("Tree Textual Memory Recall Example") +print("=" * 70) + +current_dir = os.path.dirname(os.path.abspath(__file__)) +config_dir = os.path.join(current_dir, "../data/config") + +# Load the shared tree-text memory configuration +# This config includes both embedder settings and graph database settings +config_path = os.path.join(config_dir, "tree_config_shared_database.json") +with open(config_path) as f: + config_data = json.load(f) + +print(f"\n✓ Loaded configuration from: {config_path}") + +# ============================================================================ +# Step 1: Initialize Embedder +# ============================================================================ +# The embedder converts text into vector embeddings for semantic search +embedder_config = EmbedderConfigFactory.model_validate(config_data["embedder"]) embedder = EmbedderFactory.from_config(embedder_config) -# Step 1: Prepare a mock ParsedTaskGoal +print(f"✓ Initialized embedder: {embedder_config.backend}") + +# ============================================================================ +# Step 2: Initialize Graph Store +# ============================================================================ +# The graph store persists memories and supports both graph queries and vector search +graph_config = GraphDBConfigFactory(**config_data["graph_db"]) +graph_store = GraphStoreFactory.from_config(graph_config) + +print(f"✓ Initialized graph store: {graph_config.backend}") + +# ============================================================================ +# Step 3: Clean up old mock data (optional) +# ============================================================================ +# If you're running this example multiple times, clean up previous test data +# to avoid duplicates. This is optional in production. +print("\nCleaning up old mock data...") +try: + if hasattr(graph_store, "delete_node_by_prams"): + graph_store.delete_node_by_prams(filter={"key": "LGBTQ support group"}) + graph_store.delete_node_by_prams(filter={"key": "LGBTQ community"}) + print("✓ Old mock data cleaned") + else: + print("⚠ Graph store doesn't support delete_node_by_prams, skipping cleanup") +except Exception as exc: + print(f"⚠ Cleanup warning: {exc}") + +# ============================================================================ +# Step 4: Insert mock memories into the graph store +# ============================================================================ +# In a real application, these would be memories extracted from user conversations +# or documents. Here we use a few hardcoded examples about "Caroline". +print("\nInserting mock memories...") + +mock_memories = [ + { + "memory": "Caroline joined the LGBTQ support group in 2023.", + "tags": ["LGBTQ", "support group"], + "key": "LGBTQ support group", + }, + { + "memory": "Caroline has been an active member of the LGBTQ community since college.", + "tags": ["LGBTQ", "community"], + "key": "LGBTQ community", + }, + { + "memory": "She attended the weekly LGBTQ support group meetings every Friday.", + "tags": ["LGBTQ", "support group", "meetings"], + "key": "LGBTQ support group", + }, +] + +for idx, mem_data in enumerate(mock_memories, 1): + # Generate embedding for this memory + mem_embedding = embedder.embed([mem_data["memory"]])[0] + + # Create a TextualMemoryItem with metadata + item = TextualMemoryItem( + memory=mem_data["memory"], + metadata=TreeNodeTextualMemoryMetadata( + memory_type="LongTermMemory", # Can be ShortTermMemory, LongTermMemory, etc. + key=mem_data["key"], + tags=mem_data["tags"], + embedding=mem_embedding, + sources=[], + ), + ) + + # Add the memory node to the graph store + graph_store.add_node(item.id, item.memory, item.metadata.model_dump()) + print(f" [{idx}/{len(mock_memories)}] Added: {mem_data['memory'][:60]}...") + +print("✓ Mock memories inserted successfully") + +# ============================================================================ +# Step 5: Define a query and retrieval goal +# ============================================================================ +# This is the natural language question we want to answer +query = "When did Caroline go to the LGBTQ support group?" +print(f"\n{'=' * 70}") +print(f"Query: {query}") +print(f"{'=' * 70}") + +# ParsedTaskGoal provides hints to guide the retrieval process: +# - memories: semantic descriptions of what we're looking for +# - keys: specific keywords to match +# - tags: categorical tags to filter by parsed_goal = ParsedTaskGoal( memories=[ "Caroline's participation in the LGBTQ community", @@ -34,37 +160,46 @@ tags=["LGBTQ", "support group"], ) -# Step 2: Initialize graph store -graph_config = GraphDBConfigFactory( - backend="neo4j", - config={ - "uri": "bolt://localhost:7687", - "user": "neo4j", - "password": "12345678", - "db_name": "caroline", - "auto_create": True, - }, -) -graph_store = GraphStoreFactory.from_config(graph_config) +# ============================================================================ +# Step 6: Perform hybrid retrieval +# ============================================================================ +# The retriever uses both semantic similarity (embeddings) and graph structure +# to find the most relevant memories +print("\nPerforming hybrid retrieval...") -# Step 6: Create embedding for query -query = "When did Caroline go to the LGBTQ support group?" query_embedding = embedder.embed([query])[0] - -# Step 7: Init memory retriever retriever = GraphMemoryRetriever(graph_store=graph_store, embedder=embedder) -# Step 8: Run hybrid retrieval retrieved_items: list[TextualMemoryItem] = retriever.retrieve( query=query, parsed_goal=parsed_goal, - top_k=10, - memory_scope="LongTermMemory", + top_k=10, # Maximum number of memories to retrieve + memory_scope="LongTermMemory", # Filter by memory type query_embedding=[query_embedding], ) -# Step 9: Print retrieved memory items -print("\n=== Retrieved Memory Items ===") -for item in retrieved_items: - print(f"ID: {item.id}") - print(f"Memory: {item.memory}") +print(f"✓ Retrieved {len(retrieved_items)} memories") + +# ============================================================================ +# Step 7: Display results +# ============================================================================ +print(f"\n{'=' * 70}") +print("Retrieved Memory Items:") +print(f"{'=' * 70}\n") + +if not retrieved_items: + print("❌ No memories retrieved.") + print(" This might indicate:") + print(" - The mock data wasn't inserted correctly") + print(" - The query doesn't match any stored memories") + print(" - The retrieval parameters are too restrictive") +else: + for idx, item in enumerate(retrieved_items, 1): + print(f"[{idx}] ID: {item.id}") + print(f" Memory: {item.memory}") + print(f" Tags: {item.metadata.tags if hasattr(item.metadata, 'tags') else 'N/A'}") + print() + +print(f"{'=' * 70}") +print("Example completed successfully!") +print(f"{'=' * 70}\n") diff --git a/examples/basic_modules/tree_textual_memory_relation_reason_detector.py b/examples/basic_modules/tree_textual_memory_relation_reason_detector.py index 72e4deb60..dd6bca67f 100644 --- a/examples/basic_modules/tree_textual_memory_relation_reason_detector.py +++ b/examples/basic_modules/tree_textual_memory_relation_reason_detector.py @@ -1,3 +1,31 @@ +""" +Tree Textual Memory Relation & Reasoning Detector Example +========================================================== + +This example demonstrates how to use MemOS's RelationAndReasoningDetector to +automatically discover relationships between memories and infer new knowledge. + +**What you'll learn:** +- How to initialize embedder, graph store, and LLM for relation detection +- How to create mock memory nodes with rich metadata +- How to detect pairwise relations between memory nodes (e.g., causal, temporal) +- How to infer new facts through multi-hop reasoning chains +- How to generate aggregate concepts from related memories +- How to identify sequential patterns (FOLLOWS relationships) + +**Use case:** +You have stored multiple facts about a user (e.g., "Caroline's work stress", +"joining support group", "improved mental health"). This detector can: +1. Find causal links: "Work stress" → "Joining support group" → "Better mental health" +2. Infer new facts: "Support groups help reduce work-related stress" +3. Build aggregate concepts: "Caroline's stress management journey" + +Run this example: + python examples/basic_modules/tree_textual_memory_relation_reason_detector.py +""" + +import json +import os import uuid from memos import log @@ -16,52 +44,79 @@ logger = log.get_logger(__name__) -# === Step 1: Initialize embedder === -embedder_config = EmbedderConfigFactory.model_validate( - { - "backend": "ollama", - "config": { - "model_name_or_path": "nomic-embed-text:latest", - }, - } -) +# ============================================================================ +# Step 0: Setup - Load configuration files +# ============================================================================ +print("=" * 80) +print("Tree Textual Memory Relation & Reasoning Detector Example") +print("=" * 80) +print("\nThis example will:") +print(" 1. Create a set of related memories about Caroline") +print(" 2. Detect causal and temporal relationships between them") +print(" 3. Infer new knowledge through reasoning chains") +print(" 4. Generate aggregate concepts") +print("=" * 80) + +current_dir = os.path.dirname(os.path.abspath(__file__)) +config_dir = os.path.join(current_dir, "../data/config") + +# Load the shared tree-text memory configuration +# This includes embedder, graph DB, and LLM configurations +config_path = os.path.join(config_dir, "tree_config_shared_database.json") +with open(config_path) as f: + config_data = json.load(f) + +print(f"\n✓ Loaded configuration from: {config_path}") + +# ============================================================================ +# Step 1: Initialize Embedder +# ============================================================================ +print("\n[Step 1] Initializing embedder...") + +embedder_config = EmbedderConfigFactory.model_validate(config_data["embedder"]) embedder = EmbedderFactory.from_config(embedder_config) -# === Step 2: Initialize Neo4j GraphStore === -graph_config = GraphDBConfigFactory( - backend="neo4j", - config={ - "uri": "bolt://localhost:7687", - "user": "neo4j", - "password": "12345678", - "db_name": "lucy4", - "auto_create": True, - }, -) +print(f"✓ Embedder initialized: {embedder_config.backend}") + +# ============================================================================ +# Step 2: Initialize Graph Store +# ============================================================================ +print("\n[Step 2] Initializing graph database...") + +# Load graph database configuration from the config file +graph_config = GraphDBConfigFactory(**config_data["graph_db"]) graph_store = GraphStoreFactory.from_config(graph_config) -# === Step 3: Initialize LLM for pairwise relation detection === -# Step 1: Load LLM config and instantiate -config = LLMConfigFactory.model_validate( - { - "backend": "ollama", - "config": { - "model_name_or_path": "qwen3:0.6b", - "temperature": 0.7, - "max_tokens": 1024, - }, - } -) -llm = LLMFactory.from_config(config) +print(f"✓ Graph store initialized: {graph_config.backend}") +print(f" Connected to: {graph_config.config.get('uri', 'N/A')}") +print(f" Database: {graph_config.config.get('db_name', 'N/A')}") + +# ============================================================================ +# Step 3: Initialize LLM +# ============================================================================ +print("\n[Step 3] Initializing LLM for relation detection...") -# === Step 4: Create a mock GraphDBNode to test relation detection === +# The LLM analyzes pairs of memories to detect semantic relationships +# (e.g., "causes", "leads to", "happens before", etc.) +# We use the extractor_llm from the config file +llm_config = LLMConfigFactory.model_validate(config_data["extractor_llm"]) +llm = LLMFactory.from_config(llm_config) +print(f"✓ LLM initialized: {llm_config.backend}") + +# ============================================================================ +# Step 4: Create Mock Memory Nodes +# ============================================================================ +print("\n[Step 4] Creating mock memory nodes...") +print(" Building a scenario about Caroline's stress and support journey...\n") + +# Node A: Caroline's work stress node_a = GraphDBNode( id=str(uuid.uuid4()), memory="Caroline faced increased workload stress during the project deadline.", metadata=TreeNodeTextualMemoryMetadata( memory_type="LongTermMemory", - embedding=[0.1] * 10, + embedding=[0.1] * 10, # Placeholder embedding (real one will be generated) key="Workload stress", tags=["stress", "workload"], type="fact", @@ -70,7 +125,7 @@ updated_at="2024-06-28T09:00:00Z", ), ) - +# Node B: Improved mental health after joining support group node_b = GraphDBNode( id=str(uuid.uuid4()), memory="After joining the support group, Caroline reported improved mental health.", @@ -85,7 +140,9 @@ updated_at="2024-07-10T12:00:00Z", ), ) +print(" ✓ Node B: Improved mental health") +# Node C: General research about support groups node_c = GraphDBNode( id=str(uuid.uuid4()), memory="Peer support groups are effective in reducing stress for LGBTQ individuals.", @@ -100,8 +157,9 @@ updated_at="2024-06-29T14:00:00Z", ), ) +print(" ✓ Node C: Support group benefits") -# === D: Work pressure ➜ stress === +# Node D: Work pressure → stress (causal chain element) node_d = GraphDBNode( id=str(uuid.uuid4()), memory="Excessive work pressure increases stress levels among employees.", @@ -116,8 +174,9 @@ updated_at="2024-06-15T08:00:00Z", ), ) +print(" ✓ Node D: Work pressure → stress") -# === E: Stress ➜ poor sleep === +# Node E: Stress → poor sleep (causal chain element) node_e = GraphDBNode( id=str(uuid.uuid4()), memory="High stress levels often result in poor sleep quality.", @@ -132,8 +191,9 @@ updated_at="2024-06-18T10:00:00Z", ), ) +print(" ✓ Node E: Stress → poor sleep") -# === F: Poor sleep ➜ low performance === +# Node F: Poor sleep → low performance (causal chain element) node_f = GraphDBNode( id=str(uuid.uuid4()), memory="Employees with poor sleep show reduced work performance.", @@ -148,7 +208,10 @@ updated_at="2024-06-20T12:00:00Z", ), ) +print(" ✓ Node F: Poor sleep → low performance") +# Main Node: The central fact we want to analyze +# This node will be used as the "anchor" to find related memories node = GraphDBNode( id="a88db9ce-3c77-4e83-8d61-aa9ef95c957e", memory="Caroline joined an LGBTQ support group to cope with work-related stress.", @@ -156,7 +219,7 @@ memory_type="LongTermMemory", embedding=embedder.embed( ["Caroline joined an LGBTQ support group to cope with work-related stress."] - )[0], + )[0], # Generate real embedding for the main node key="Caroline LGBTQ stress", tags=["LGBTQ", "support group", "stress"], type="fact", @@ -165,49 +228,122 @@ updated_at="2024-07-01T10:00:00Z", ), ) +print(" ✓ Main Node: Caroline's support group action\n") +# ============================================================================ +# Step 5: Insert Nodes into Graph Store +# ============================================================================ +print("[Step 5] Inserting all nodes into graph database...") -for n in [node, node_a, node_b, node_c, node_d, node_e, node_f]: +all_nodes = [node, node_a, node_b, node_c, node_d, node_e, node_f] +for n in all_nodes: graph_store.add_node(n.id, n.memory, n.metadata.dict()) +print(f"✓ Successfully inserted {len(all_nodes)} memory nodes into the graph\n") + +# ============================================================================ +# Step 6: Initialize Relation & Reasoning Detector +# ============================================================================ +print("[Step 6] Initializing RelationAndReasoningDetector...") -# === Step 5: Initialize RelationDetector and run detection === relation_detector = RelationAndReasoningDetector( - graph_store=graph_store, llm=llm, embedder=embedder + graph_store=graph_store, + llm=llm, + embedder=embedder, ) +print("✓ Detector initialized and ready\n") + +# ============================================================================ +# Step 7: Run Relation Detection & Reasoning +# ============================================================================ +print("[Step 7] Running relation detection and reasoning...") +print(f" Analyzing relationships for: '{node.memory[:60]}...'\n") + +# This will: +# 1. Find semantically similar nodes using embeddings +# 2. Detect pairwise relations (causal, temporal, etc.) using LLM +# 3. Infer new facts through multi-hop reasoning +# 4. Generate aggregate concepts +# 5. Identify sequential patterns results = relation_detector.process_node( node=node, - exclude_ids=[node.id], # Exclude self when searching for neighbors - top_k=5, + exclude_ids=[node.id], # Don't compare the node with itself + top_k=5, # Consider top 5 most similar nodes ) -# === Step 6: Print detected relations === -print("\n=== Detected Global Relations ===") +print("✓ Analysis complete!\n") + +# ============================================================================ +# Step 8: Display Results +# ============================================================================ +print("=" * 80) +print("ANALYSIS RESULTS") +print("=" * 80) + +# Display detected pairwise relations +print("\n📊 [1] Detected Pairwise Relations") +print("-" * 80) +if results["relations"]: + for idx, rel in enumerate(results["relations"], 1): + print(f"\n Relation #{idx}:") + print(f" Source: {rel['source_id'][:8]}...") + print(f" Target: {rel['target_id'][:8]}...") + print(f" Type: {rel['relation_type']}") +else: + print(" ❌ No pairwise relations detected") + print(" Try adjusting similarity threshold or adding more related nodes") +# Display inferred new facts +print("\n\n💡 [2] Inferred New Facts (through reasoning)") +print("-" * 80) +if results["inferred_nodes"]: + for idx, inferred_node in enumerate(results["inferred_nodes"], 1): + print(f"\n Inferred Fact #{idx}:") + print(f" 💬 {inferred_node.memory}") + print(f" 📌 Sources: {inferred_node.metadata.sources}") + print(f" 🏷️ Key: {inferred_node.metadata.key}") +else: + print(" ℹ️ No new facts inferred") + print(" This is normal if relations are simple or insufficient for reasoning") -# === Step 6: Pretty-print detected results === -print("\n=== Detected Pairwise Relations ===") -for rel in results["relations"]: - print(f" Source ID: {rel['source_id']}") - print(f" Target ID: {rel['target_id']}") - print(f" Relation Type: {rel['relation_type']}") - print("------") +# Display sequence links (temporal ordering) +print("\n\n⏱️ [3] Sequence Links (FOLLOWS relationships)") +print("-" * 80) +if results["sequence_links"]: + for idx, link in enumerate(results["sequence_links"], 1): + print(f" {idx}. {link['from_id'][:8]}... → {link['to_id'][:8]}...") +else: + print(" ℹ️ No sequential patterns detected") -print("\n=== Inferred Nodes ===") -for node in results["inferred_nodes"]: - print(f" New Fact: {node.memory}") - print(f" Sources: {node.metadata.sources}") - print("------") +# Display aggregate concepts +print("\n\n🎯 [4] Aggregate Concepts") +print("-" * 80) +if results["aggregate_nodes"]: + for idx, agg in enumerate(results["aggregate_nodes"], 1): + print(f"\n Concept #{idx}:") + print(f" 📖 {agg.memory}") + print(f" 🔑 Key: {agg.metadata.key}") + print(f" 📎 Aggregates from: {agg.metadata.sources}") +else: + print(" ℹ️ No aggregate concepts generated") + print(" Aggregates are created when multiple related memories share themes") -print("\n=== Sequence Links (FOLLOWS) ===") -for link in results["sequence_links"]: - print(f" From: {link['from_id']} -> To: {link['to_id']}") - print("------") +print("\n" + "=" * 80) +print("Example completed successfully!") +print("=" * 80) +print("\n💡 Next steps:") +print(" - Modify the mock memories to test different scenarios") +print(" - Adjust top_k parameter to control how many neighbors are considered") +print(" - Experiment with different LLM models for relation detection") +print(" - Check the Neo4j database to visualize the created graph\n") print("\n=== Aggregate Concepts ===") -for agg in results["aggregate_nodes"]: - print(f" Concept Key: {agg.metadata.key}") - print(f" Concept Memory: {agg.memory}") - print(f" Sources: {agg.metadata.sources}") - print("------") +if not results["aggregate_nodes"]: + print("No aggregate concepts generated.") +else: + for agg in results["aggregate_nodes"]: + print(f" Concept Key: {agg.metadata.key}") + print(f" Concept Memory: {agg.memory}") + print(f" Sources: {agg.metadata.sources}") + print("------") diff --git a/examples/basic_modules/tree_textual_memory_task_goal_parser.py b/examples/basic_modules/tree_textual_memory_task_goal_parser.py index c36d47cf1..11c2c8087 100644 --- a/examples/basic_modules/tree_textual_memory_task_goal_parser.py +++ b/examples/basic_modules/tree_textual_memory_task_goal_parser.py @@ -1,52 +1,185 @@ +""" +Tree Textual Memory Task Goal Parser Example +============================================= + +This example demonstrates how to use MemOS's TaskGoalParser to parse natural +language queries into structured retrieval goals. + +**What you'll learn:** +- How to initialize an LLM for task parsing +- How to parse a natural language query into structured components +- The difference between "fast" and "fine" parsing modes +- How the parser extracts memories, keys, tags, and goal types + +**Use case:** +When a user asks "When did Caroline go to the LGBTQ support group?", you need to: +1. Extract semantic descriptions (memories to look for) +2. Identify key phrases and keywords +3. Determine relevant tags for filtering +4. Classify the goal type (retrieval, update, etc.) + +The TaskGoalParser does this automatically using an LLM. + +Run this example: + python examples/basic_modules/tree_textual_memory_task_goal_parser.py +""" + +import json +import os import time +from memos import log from memos.configs.llm import LLMConfigFactory from memos.llms.factory import LLMFactory from memos.memories.textual.tree_text_memory.retrieve.task_goal_parser import TaskGoalParser -# Step 1: Load LLM config and instantiate -config = LLMConfigFactory.model_validate( - { - "backend": "ollama", - "config": { - "model_name_or_path": "qwen3:0.6b", - "temperature": 0.7, - "max_tokens": 1024, - "remove_think_prefix": True, - }, - } -) -llm = LLMFactory.from_config(config) +logger = log.get_logger(__name__) + +# ============================================================================ +# Step 0: Setup - Load configuration files +# ============================================================================ +print("=" * 80) +print("Tree Textual Memory Task Goal Parser Example") +print("=" * 80) + +current_dir = os.path.dirname(os.path.abspath(__file__)) +config_dir = os.path.join(current_dir, "../data/config") + +# Load the shared tree-text memory configuration +config_path = os.path.join(config_dir, "tree_config_shared_database.json") +with open(config_path) as f: + config_data = json.load(f) + +print(f"\n✓ Loaded configuration from: {config_path}") + +# ============================================================================ +# Step 1: Initialize LLM for Task Parsing +# ============================================================================ +print("\n[Step 1] Initializing LLM for task goal parsing...") + +# The LLM will analyze the natural language query and extract structured information +# We use the extractor_llm from the config file +llm_config = LLMConfigFactory.model_validate(config_data["extractor_llm"]) +llm = LLMFactory.from_config(llm_config) + +print(f"✓ LLM initialized: {llm_config.backend}") -# Task input +# ============================================================================ +# Step 2: Define a natural language task/query +# ============================================================================ +# This is the user's question that needs to be parsed task = "When did Caroline go to the LGBTQ support group?" -parser = TaskGoalParser(llm, mode="fast") +print("\n[Step 2] Task to parse:") +print(f" 📝 '{task}'") +print() -time_init = time.time() -# Parse task goal -result = parser.parse(task) +# ============================================================================ +# Step 3: Parse using FAST mode +# ============================================================================ +print("[Step 3] Parsing with FAST mode...") +print(" (Fast mode uses a simpler prompt for quick parsing)") -# Print output -print("=== Parsed Result ===") -print("memories:", result.memories) -print("keys:", result.keys) -print("tags:", result.tags) -print("goal_type:", result.goal_type) -print("time:", time.time() - time_init) +parser = TaskGoalParser(llm) -parser = TaskGoalParser(llm, mode="fine") +time_start = time.time() +result_fast = parser.parse(task, mode="fast") +time_fast = time.time() - time_start -time_init = time.time() +print(f"✓ Fast mode parsing completed in {time_fast:.3f}s\n") -# Parse task goal -result = parser.parse(task) +# Display fast mode results +print("=" * 80) +print("FAST MODE RESULTS") +print("=" * 80) +print("\n📋 Memories (semantic descriptions):") +if result_fast.memories: + for idx, mem in enumerate(result_fast.memories, 1): + print(f" {idx}. {mem}") +else: + print(" (None extracted)") + +print("\n🔑 Keys (important keywords):") +if result_fast.keys: + for idx, key in enumerate(result_fast.keys, 1): + print(f" {idx}. {key}") +else: + print(" (None extracted)") + +print("\n🏷️ Tags (categorical labels):") +if result_fast.tags: + print(f" {', '.join(result_fast.tags)}") +else: + print(" (None extracted)") + +print(f"\n🎯 Goal Type: {result_fast.goal_type}") +print(f"⏱️ Processing Time: {time_fast:.3f}s") + +# ============================================================================ +# Step 4: Parse using FINE mode +# ============================================================================ +print(f"\n{'=' * 80}") +print("[Step 4] Parsing with FINE mode...") +print(" (Fine mode uses more detailed prompts for better accuracy)") + +time_start = time.time() +result_fine = parser.parse(task, mode="fine") +time_fine = time.time() - time_start + +print(f"✓ Fine mode parsing completed in {time_fine:.3f}s\n") + +# Display fine mode results +print("=" * 80) +print("FINE MODE RESULTS") +print("=" * 80) +print("\n📋 Memories (semantic descriptions):") +if result_fine.memories: + for idx, mem in enumerate(result_fine.memories, 1): + print(f" {idx}. {mem}") +else: + print(" (None extracted)") + +print("\n🔑 Keys (important keywords):") +if result_fine.keys: + for idx, key in enumerate(result_fine.keys, 1): + print(f" {idx}. {key}") +else: + print(" (None extracted)") + +print("\n🏷️ Tags (categorical labels):") +if result_fine.tags: + print(f" {', '.join(result_fine.tags)}") +else: + print(" (None extracted)") + +print(f"\n🎯 Goal Type: {result_fine.goal_type}") +print(f"⏱️ Processing Time: {time_fine:.3f}s") + +# ============================================================================ +# Step 5: Compare Results +# ============================================================================ +print(f"\n{'=' * 80}") +print("COMPARISON") +print("=" * 80) +print("\nSpeed:") +print(f" Fast mode: {time_fast:.3f}s") +print(f" Fine mode: {time_fine:.3f}s") +print(f" Difference: {abs(time_fast - time_fine):.3f}s") + +print("\nExtracted Components:") +print( + f" Fast mode: {len(result_fast.memories)} memories, {len(result_fast.keys)} keys, {len(result_fast.tags)} tags" +) +print( + f" Fine mode: {len(result_fine.memories)} memories, {len(result_fine.keys)} keys, {len(result_fine.tags)} tags" +) -# Print output -print("=== Parsed Result ===") -print("memories:", result.memories) -print("keys:", result.keys) -print("tags:", result.tags) -print("goal_type:", result.goal_type) -print("time:", time.time() - time_init) +print(f"\n{'=' * 80}") +print("Example completed successfully!") +print("=" * 80) +print("\n💡 Next steps:") +print(" - Try different queries to see how the parser handles various inputs") +print(" - Use the parsed result as input for GraphMemoryRetriever") +print(" - Experiment with 'fast' vs 'fine' mode based on your accuracy/speed needs") +print(" - The parsed ParsedTaskGoal can be passed directly to retrieval functions\n") diff --git a/examples/core_memories/general_textual_memory.py b/examples/core_memories/general_textual_memory.py index f71e2ef2e..d5c765b01 100644 --- a/examples/core_memories/general_textual_memory.py +++ b/examples/core_memories/general_textual_memory.py @@ -1,14 +1,25 @@ +import os +import pprint + from memos.configs.memory import MemoryConfigFactory from memos.memories.factory import MemoryFactory +# Initialize the memory configuration +# This configuration specifies the extractor, vector database, and embedder backend. +# Here we use OpenAI for extraction, Qdrant for vector storage, and Ollama for embedding. config = MemoryConfigFactory( backend="general_text", config={ "extractor_llm": { - "backend": "ollama", + "backend": "openai", "config": { - "model_name_or_path": "qwen3:0.6b", + "model_name_or_path": "gpt-4o-mini", + "api_key": os.environ.get("OPENAI_API_KEY"), + "api_base": os.environ.get( + "OPENAI_BASE_URL", + os.environ.get("OPENAI_API_BASE", "https://api.openai.com/v1"), + ), "temperature": 0.0, "remove_think_prefix": True, "max_tokens": 8192, @@ -30,6 +41,8 @@ }, }, ) + +# Create the memory instance from the configuration m = MemoryFactory.from_config(config) example_memories = [ @@ -52,20 +65,25 @@ }, }, ] + example_id = "a19b6caa-5d59-42ad-8c8a-e4f7118435b4" print("===== Extract memories =====") +# Extract memories from a conversation +# The extractor LLM processes the conversation to identify relevant information. memories = m.extract( [ {"role": "user", "content": "I love tomatoes."}, {"role": "assistant", "content": "Great! Tomatoes are delicious."}, ] ) -print(memories) +pprint.pprint(memories) print() print("==== Add memories ====") +# Add the extracted memories to the memory store m.add(memories) +# Add a manually created memory item m.add( [ { @@ -80,19 +98,27 @@ } ] ) -print(m.get_all()) +print("All memories after addition:") +pprint.pprint(m.get_all()) print() + print("==== Search memories ====") +# Search for memories related to a query search_results = m.search("Tell me more about the user", top_k=2) -print(search_results) +pprint.pprint(search_results) print() print("==== Get memories ====") -print(m.get(example_id)) -print(m.get_by_ids([example_id])) +# Retrieve a specific memory by its ID +print(f"Memory with ID {example_id}:") +pprint.pprint(m.get(example_id)) +# Retrieve multiple memories by IDs +print(f"Memories by IDs [{example_id}]:") +pprint.pprint(m.get_by_ids([example_id])) print() print("==== Update memories ====") +# Update an existing memory m.update( example_id, { @@ -106,19 +132,26 @@ }, }, ) -print(m.get(example_id)) +print(f"Memory after update (ID {example_id}):") +pprint.pprint(m.get(example_id)) +print() + +print("==== Dump memory ====") +# Dump the current state of memory to a file +m.dump("tmp/general_mem") +print("Memory dumped to 'tmp/general_mem'.") print() print("==== Delete memories ====") +# Delete a memory by its ID m.delete([example_id]) -print(m.get_all()) +print("All memories after deletion:") +pprint.pprint(m.get_all()) print() print("==== Delete all memories ====") +# Clear all memories from the store m.delete_all() -print(m.get_all()) +print("All memories after delete_all:") +pprint.pprint(m.get_all()) print() - -print("==== Dump memory ====") -m.dump("tmp/mem") -print("Memory dumped to 'tmp/mem'.") diff --git a/examples/core_memories/kv_cache_memory.py b/examples/core_memories/kv_cache_memory.py index dbe3ac4d7..470ad56a7 100644 --- a/examples/core_memories/kv_cache_memory.py +++ b/examples/core_memories/kv_cache_memory.py @@ -1,68 +1,132 @@ +import json + +from transformers import DynamicCache + from memos.configs.memory import MemoryConfigFactory +from memos.memories.activation.item import KVCacheItem from memos.memories.factory import MemoryFactory -# ===== Example: Use factory and HFLLM to build and manage KVCacheMemory ===== +def get_cache_info(cache): + if not cache: + return None + + num_layers = 0 + total_size_bytes = 0 + + if hasattr(cache, "layers"): + num_layers = len(cache.layers) + for layer in cache.layers: + if hasattr(layer, "key_cache") and layer.key_cache is not None: + total_size_bytes += layer.key_cache.nelement() * layer.key_cache.element_size() + if hasattr(layer, "value_cache") and layer.value_cache is not None: + total_size_bytes += layer.value_cache.nelement() * layer.value_cache.element_size() + + if hasattr(layer, "keys") and layer.keys is not None: + total_size_bytes += layer.keys.nelement() * layer.keys.element_size() + if hasattr(layer, "values") and layer.values is not None: + total_size_bytes += layer.values.nelement() * layer.values.element_size() + + elif hasattr(cache, "key_cache") and hasattr(cache, "value_cache"): + num_layers = len(cache.key_cache) + for k, v in zip(cache.key_cache, cache.value_cache, strict=False): + if k is not None: + total_size_bytes += k.nelement() * k.element_size() + if v is not None: + total_size_bytes += v.nelement() * v.element_size() + + return { + "num_layers": num_layers, + "size_bytes": total_size_bytes, + "size_mb": f"{total_size_bytes / (1024 * 1024):.2f} MB", + } + + +def serialize_item(obj): + if isinstance(obj, list): + return [serialize_item(x) for x in obj] + + if isinstance(obj, KVCacheItem): + return { + "id": obj.id, + "metadata": obj.metadata, + "records": obj.records.model_dump() + if hasattr(obj.records, "model_dump") + else obj.records, + "memory": get_cache_info(obj.memory), + } + + if isinstance(obj, DynamicCache): + return get_cache_info(obj) -# 1. Create config for KVCacheMemory (using HuggingFace backend) -config = MemoryConfigFactory( - backend="kv_cache", - config={ - "extractor_llm": { - "backend": "huggingface", - "config": { - "model_name_or_path": "Qwen/Qwen3-0.6B", # Use a valid HuggingFace model name - "max_tokens": 32, - "add_generation_prompt": True, - "remove_think_prefix": True, + return str(obj) + + +if __name__ == "__main__": + # ===== Example: Use factory and HFLLM to build and manage KVCacheMemory ===== + + # 1. Create config for KVCacheMemory (using HuggingFace backend) + config = MemoryConfigFactory( + backend="kv_cache", + config={ + "extractor_llm": { + "backend": "huggingface", + "config": { + "model_name_or_path": "Qwen/Qwen3-0.6B", # Use a valid HuggingFace model name + "max_tokens": 32, + "add_generation_prompt": True, + "remove_think_prefix": True, + }, }, }, - }, -) - -# 2. Instantiate KVCacheMemory using the factory -kv_mem = MemoryFactory.from_config(config) - -# 3. Extract a KVCacheItem (DynamicCache) from a prompt (uses HFLLM.build_kv_cache internally) -prompt = [ - {"role": "user", "content": "What is MemOS?"}, - {"role": "assistant", "content": "MemOS is a memory operating system for LLMs."}, -] -print("===== Extract KVCacheItem =====") -cache_item = kv_mem.extract(prompt) -print(cache_item) -print() - -# 4. Add the extracted KVCacheItem -print("===== Add KVCacheItem =====") -kv_mem.add([cache_item]) -print(kv_mem.get_all()) -print() - -# 5. Get by id -print("===== Get KVCacheItem by id =====") -retrieved = kv_mem.get(cache_item.id) -print(retrieved) -print() - -# 6. Merge caches (simulate with two items) -print("===== Merge DynamicCache =====") -item2 = kv_mem.extract([{"role": "user", "content": "Tell me a joke."}]) -kv_mem.add([item2]) -merged_cache = kv_mem.get_cache([cache_item.id, item2.id]) -print(merged_cache) -print() - -# 7. Delete one -print("===== Delete one KVCacheItem =====") -kv_mem.delete([cache_item.id]) -print(kv_mem.get_all()) -print() - -# 8. Dump and load -print("===== Dump and Load KVCacheMemory =====") -kv_mem.dump("tmp/kv_mem") -print("Memory dumped to 'tmp/kv_mem'.") -kv_mem.delete_all() -kv_mem.load("tmp/kv_mem") -print("Memory loaded from 'tmp/kv_mem':", kv_mem.get_all()) + ) + + # 2. Instantiate KVCacheMemory using the factory + kv_mem = MemoryFactory.from_config(config) + + # 3. Extract a KVCacheItem (DynamicCache) from a prompt (uses HFLLM.build_kv_cache internally) + prompt = [ + {"role": "user", "content": "What is MemOS?"}, + {"role": "assistant", "content": "MemOS is a memory operating system for LLMs."}, + ] + print("===== Extract KVCacheItem =====") + cache_item = kv_mem.extract(prompt) + print(json.dumps(serialize_item(cache_item), indent=2, default=str)) + print() + + # 4. Add the extracted KVCacheItem + print("===== Add KVCacheItem =====") + kv_mem.add([cache_item]) + print(json.dumps(serialize_item(kv_mem.get_all()), indent=2, default=str)) + print() + + # 5. Get by id + print("===== Get KVCacheItem by id =====") + retrieved = kv_mem.get(cache_item.id) + print(json.dumps(serialize_item(retrieved), indent=2, default=str)) + print() + + # 6. Merge caches (simulate with two items) + print("===== Merge DynamicCache =====") + item2 = kv_mem.extract([{"role": "user", "content": "Tell me a joke."}]) + kv_mem.add([item2]) + merged_cache = kv_mem.get_cache([cache_item.id, item2.id]) + print(json.dumps(serialize_item(merged_cache), indent=2, default=str)) + print() + + # 7. Delete one + print("===== Delete one KVCacheItem =====") + kv_mem.delete([cache_item.id]) + print(json.dumps(serialize_item(kv_mem.get_all()), indent=2, default=str)) + print() + + # 8. Dump and load + print("===== Dump and Load KVCacheMemory =====") + kv_mem.dump("tmp/kv_mem") + print("Memory dumped to 'tmp/kv_mem'.") + kv_mem.delete_all() + kv_mem.load("tmp/kv_mem") + print( + "Memory loaded from 'tmp/kv_mem':", + json.dumps(serialize_item(kv_mem.get_all()), indent=2, default=str), + ) diff --git a/examples/core_memories/naive_textual_memory.py b/examples/core_memories/naive_textual_memory.py index ff61334dc..ab73060c7 100644 --- a/examples/core_memories/naive_textual_memory.py +++ b/examples/core_memories/naive_textual_memory.py @@ -1,53 +1,77 @@ +import json +import os import uuid from memos.configs.memory import MemoryConfigFactory from memos.memories.factory import MemoryFactory +def print_result(title, result): + """Helper function: Pretty print the result.""" + print(f"\n{'=' * 10} {title} {'=' * 10}") + if isinstance(result, list | dict): + print(json.dumps(result, indent=2, ensure_ascii=False, default=str)) + else: + print(result) + + +# Configure memory backend with OpenAI extractor config = MemoryConfigFactory( backend="naive_text", config={ "extractor_llm": { - "backend": "ollama", + "backend": "openai", "config": { - "model_name_or_path": "qwen3:0.6b", + "model_name_or_path": "gpt-4o-mini", + "api_key": os.environ.get("OPENAI_API_KEY"), + "api_base": os.environ.get( + "OPENAI_BASE_URL", + os.environ.get("OPENAI_API_BASE", "https://api.openai.com/v1"), + ), "temperature": 0.0, "remove_think_prefix": True, }, } }, ) + +# Create memory instance m = MemoryFactory.from_config(config) -print("===== Extract memories =====") +# Extract memories from a simulated conversation memories = m.extract( [ {"role": "user", "content": "I love tomatoes."}, {"role": "assistant", "content": "Great! Tomatoes are delicious."}, ] ) -print(memories) -print() +print_result("Extract memories", memories) + -print("==== Add memories ====") +# Add the extracted memories to storage m.add(memories) + +# Manually create a memory item and add it example_id = str(uuid.uuid4()) -m.add([{"id": example_id, "memory": "User is Chinese.", "metadata": {"type": "opinion"}}]) -print(m.get_all()) -print() +manual_memory = [{"id": example_id, "memory": "User is Chinese.", "metadata": {"type": "opinion"}}] +m.add(manual_memory) + +# Print all current memories +print_result("Add memories (Check all after adding)", m.get_all()) -print("==== Search memories ====") + +# Search for relevant memories based on the query search_results = m.search("Tell me more about the user", top_k=2) -print(search_results) -print() +print_result("Search memories", search_results) + + +# Get specific memory item by ID +memory_item = m.get(example_id) +print_result("Get memory", memory_item) -print("==== Get memories ====") -memories = m.get(example_id) -print(memories) -print() -print("==== Update memories ====") +# Update the memory content for the specified ID m.update( example_id, { @@ -56,15 +80,22 @@ "metadata": {"type": "opinion", "confidence": 85}, }, ) -print(m.get(example_id)) +updated_memory = m.get(example_id) +print_result("Update memory", updated_memory) + + +print("==== Dump memory ====") +# Dump the current state of memory to a file +m.dump("tmp/naive_mem") +print("Memory dumped to 'tmp/naive_mem'.") print() -print("==== Delete memories ====") + +# Delete memory with the specified ID m.delete([example_id]) -print(m.get_all()) -print() +print_result("Delete memory (Check all after deleting)", m.get_all()) -print("==== Delete all memories ====") + +# Delete all memories in storage m.delete_all() -print(m.get_all()) -print() +print_result("Delete all", m.get_all()) diff --git a/examples/core_memories/textual_internet_memoy.py b/examples/core_memories/pref_textual_memory.py similarity index 76% rename from examples/core_memories/textual_internet_memoy.py rename to examples/core_memories/pref_textual_memory.py index 21979e890..ca9d471b9 100644 --- a/examples/core_memories/textual_internet_memoy.py +++ b/examples/core_memories/pref_textual_memory.py @@ -1,52 +1,18 @@ +import time + from memos import log -from memos.configs.embedder import EmbedderConfigFactory -from memos.configs.internet_retriever import InternetRetrieverConfigFactory -from memos.configs.mem_reader import SimpleStructMemReaderConfig -from memos.configs.memory import TreeTextMemoryConfig -from memos.embedders.factory import EmbedderFactory -from memos.mem_reader.simple_struct import SimpleStructMemReader -from memos.memories.textual.tree import TreeTextMemory +from memos.configs.memory import PreferenceTextMemoryConfig +from memos.memories.textual.preference import PreferenceTextMemory logger = log.get_logger(__name__) - -embedder_config = EmbedderConfigFactory.model_validate( - { - "backend": "ollama", - "config": { - "model_name_or_path": "nomic-embed-text:latest", - }, - } -) -embedder = EmbedderFactory.from_config(embedder_config) - - -def embed_memory_item(memory: str) -> list[float]: - return embedder.embed([memory])[0] - - -# google or xinyu pelase change tree_config_with_internet api key or token or url -tree_config = TreeTextMemoryConfig.from_json_file("examples/data/config/tree_config.json") -interconfig = InternetRetrieverConfigFactory( - backend="xinyu", - config={ - "api_key": "google or xinyu token", - "search_engine_id": "google engine id or xinyu url", - "max_results": 15, - "num_per_request": 10, - }, +preference_config = PreferenceTextMemoryConfig.from_json_file( + "examples/data/config/preference_config.json" ) -tree_config.internet_retriever = interconfig +my_preference_textual_memory = PreferenceTextMemory(preference_config) +my_preference_textual_memory.delete_all() -my_tree_textual_memory = TreeTextMemory(tree_config) -my_tree_textual_memory.delete_all() - -# Create a memory reader instance -reader_config = SimpleStructMemReaderConfig.from_json_file( - "examples/data/config/simple_struct_reader_config.json" -) -reader = SimpleStructMemReader(reader_config) scene_data = [ [ @@ -191,51 +157,36 @@ def embed_memory_item(memory: str) -> list[float]: ], ] -# Acquiring memories -memory = reader.get_memory(scene_data, type="chat", info={"user_id": "1234", "session_id": "2222"}) - -for m_list in memory: - my_tree_textual_memory.add(m_list) - -results = my_tree_textual_memory.search( - "Talk about the user's childhood story?", - top_k=10, - info={"query": "Talk about the user's childhood story?", "user_id": "111", "session": "2234"}, +memories = my_preference_textual_memory.get_memory( + scene_data, type="chat", info={"user_id": "1234", "session_id": "2222"} ) -for i, r in enumerate(results): - r = r.to_dict() - print(f"{i}'th similar result is: " + str(r["memory"])) -print(f"Successfully search {len(results)} memories") -# find related nodes -related_nodes = my_tree_textual_memory.get_relevant_subgraph("Painting") +added_ids = my_preference_textual_memory.add(memories) -# get current memory_size -print(f"Current Memory Size is {my_tree_textual_memory.get_current_memory_size()}") +time.sleep(10) -logger.info("Start doc search example...") -# Processing Documents -doc_paths = [ - "./text1.txt", - "./text2.txt", -] -# Acquiring memories from documents -doc_memory = reader.get_memory(doc_paths, "doc", info={"user_id": "1111", "session_id": "2222"}) - -for m_list in doc_memory: - my_tree_textual_memory.add(m_list) +init_time = time.time() +# search preference memories +results = my_preference_textual_memory.search("Talk about childhood story of the user", top_k=10) -results = my_tree_textual_memory.search( - "Tell me about what memos consist of?", - top_k=30, - info={"query": "Tell me about what memos consist of?", "user_id": "111", "session": "2234"}, -) for i, r in enumerate(results): r = r.to_dict() print(f"{i}'th similar result is: " + str(r["memory"])) -print(f"Successfully search {len(results)} memories") +print(f"Successfully search {len(results)} memories in {round(time.time() - init_time)}s") +# get all preference memories +all_preference_memories = my_preference_textual_memory.get_all() +for key, value in all_preference_memories.items(): + for i, m in enumerate(value): + print(f"{i}'th {key} memory is: " + str(m.memory)) + +# use filter to get all implicit preference memories +all_implicit_memories = my_preference_textual_memory.get_memory_by_filter( + {"preference_type": "implicit_preference"} +) +for i, m in enumerate(all_implicit_memories[0]): + print(f"{i}'th filtered memory is: " + str(m.memory)) -# my_tree_textual_memory.dump -my_tree_textual_memory.dump("tmp/my_tree_textual_memory") -my_tree_textual_memory.drop() +# dump preference memories +dumped_memories_dir = "tmp/my_preference_textual_memory" +my_preference_textual_memory.dump(dumped_memories_dir) diff --git a/examples/core_memories/tree_textual_memory.py b/examples/core_memories/tree_textual_memory.py index d2e197e5b..641335495 100644 --- a/examples/core_memories/tree_textual_memory.py +++ b/examples/core_memories/tree_textual_memory.py @@ -3,6 +3,7 @@ from memos import log from memos.configs.mem_reader import SimpleStructMemReaderConfig from memos.configs.memory import TreeTextMemoryConfig +from memos.mem_reader.multi_modal_struct import MultiModalStructMemReader from memos.mem_reader.simple_struct import SimpleStructMemReader from memos.memories.textual.tree import TreeTextMemory @@ -240,6 +241,33 @@ top_k=30, info={"query": "Tell me about what memos consist of?", "user_id": "111", "session": "2234"}, ) + +for i, r in enumerate(results): + r = r.to_dict() + print(f"{i}'th similar result is: " + str(r["memory"])) +print(f"Successfully search {len(results)} memories") + +logger.info("start multi-modal memory search example...") + +multi_modal_reader = MultiModalStructMemReader(reader_config) +doc_paths = ["examples/data/one_page_example.pdf"] +multi_modal_memory = multi_modal_reader.get_memory( + doc_paths, "doc", info={"user_id": "1111", "session_id": "2222"} +) + +for m_list in multi_modal_memory: + added_ids = my_tree_textual_memory.add(m_list) + my_tree_textual_memory.memory_manager.wait_reorganizer() + +results = my_tree_textual_memory.search( + "Give me one poem from Tagore's 'Stray birds'", + top_k=30, + info={ + "query": "Give me one poem from Tagore's 'Stray birds'", + "user_id": "111", + "session": "2234", + }, +) for i, r in enumerate(results): r = r.to_dict() print(f"{i}'th similar result is: " + str(r["memory"])) diff --git a/examples/core_memories/vllm_kv_cache_memory.py b/examples/core_memories/vllm_kv_cache_memory.py index 65ae64911..7122200f6 100644 --- a/examples/core_memories/vllm_kv_cache_memory.py +++ b/examples/core_memories/vllm_kv_cache_memory.py @@ -20,7 +20,7 @@ def main(): "extractor_llm": { "backend": "vllm", "config": { - "model_name_or_path": "/mnt/afs/models/hf_models/Qwen2.5-7B", + "model_name_or_path": "Qwen/Qwen3-0.6B", "api_base": "http://localhost:8088/v1", "temperature": 0.7, "max_tokens": 1024, diff --git a/examples/data/config/preference_config.json b/examples/data/config/preference_config.json new file mode 100644 index 000000000..4db7475a0 --- /dev/null +++ b/examples/data/config/preference_config.json @@ -0,0 +1,53 @@ +{ + "extractor_llm": { + "backend": "ollama", + "config": { + "model_name_or_path": "qwen3:0.6b", + "temperature": 0.0, + "remove_think_prefix": true, + "max_tokens": 8192 + } + }, + "vector_db": { + "backend": "milvus", + "config": { + "collection_name": [ + "explicit_preference", + "implicit_preference" + ], + "vector_dimension": 768, + "distance_metric": "cosine", + "uri": "./milvus_demo.db" + } + }, + "embedder": { + "backend": "ollama", + "config": { + "model_name_or_path": "nomic-embed-text:latest" + } + }, + "reranker": { + "backend": "cosine_local", + "config": { + "level_weights": { + "topic": 1.0, + "concept": 1.0, + "fact": 1.0 + }, + "level_field": "background" + } + }, + "extractor": { + "backend": "naive", + "config": {} + }, + "adder": { + "backend": "naive", + "config": {} + }, + "retriever": { + "backend": "naive", + "config": {} + }, + "memory_filename": "preference_textual_memories.json" +} diff --git a/examples/data/config/simple_struct_reader_config.json b/examples/data/config/simple_struct_reader_config.json index 8a27de457..abfcdfa92 100644 --- a/examples/data/config/simple_struct_reader_config.json +++ b/examples/data/config/simple_struct_reader_config.json @@ -1,24 +1,30 @@ { "llm": { - "backend": "ollama", + "backend": "openai", "config": { - "model_name_or_path": "qwen3:0.6b", - "temperature": 0.0, - "remove_think_prefix": true, - "max_tokens": 8192 + "model_name_or_path": "gpt-4o-mini", + "temperature": 0.8, + "max_tokens": 1024, + "top_p": 0.9, + "top_k": 50, + "api_key": "sk-xxx", + "api_base": "http://" } }, "embedder": { - "backend": "ollama", + "backend": "universal_api", "config": { - "model_name_or_path": "nomic-embed-text:latest" + "provider": "openai", + "api_key": "EMPTY", + "model_name_or_path": "bge-m3", + "base_url": "http://" } }, "chunker": { "backend": "sentence", "config": { "tokenizer_or_token_counter": "gpt2", - "chunk_size": 512, + "chunk_size": 2048, "chunk_overlap": 128, "min_sentences_per_chunk": 1 } diff --git a/examples/data/config/tree_config_shared_database.json b/examples/data/config/tree_config_shared_database.json index 914f3c725..6dd1a9220 100644 --- a/examples/data/config/tree_config_shared_database.json +++ b/examples/data/config/tree_config_shared_database.json @@ -1,40 +1,63 @@ { "extractor_llm": { - "backend": "ollama", + "backend": "openai", "config": { - "model_name_or_path": "qwen3:0.6b", - "temperature": 0.0, - "remove_think_prefix": true, - "max_tokens": 8192 + "model_name_or_path": "gpt-4o", + "temperature": 0.8, + "max_tokens": 1024, + "top_p": 0.9, + "top_k": 50, + "api_key": "sk-xxx", + "api_base": "http://" } }, "dispatcher_llm": { - "backend": "ollama", + "backend": "openai", "config": { - "model_name_or_path": "qwen3:0.6b", - "temperature": 0.0, - "remove_think_prefix": true, - "max_tokens": 8192 + "model_name_or_path": "gpt-4o", + "temperature": 0.8, + "max_tokens": 1024, + "top_p": 0.9, + "top_k": 50, + "api_key": "sk-xxx", + "api_base": "http://" } }, "embedder": { - "backend": "ollama", + "backend": "universal_api", "config": { - "model_name_or_path": "nomic-embed-text:latest" + "provider": "openai", + "api_key": "EMPTY", + "model_name_or_path": "bge-m3", + "base_url": "http://" + } + }, + "reranker": { + "backend": "http_bge", + "config": { + "url": "http://", + "model": "bge-reranker-v2-m3", + "timeout": 10 } }, "graph_db": { "backend": "neo4j", "config": { - "uri": "bolt://localhost:7687", + "uri": "neo4j://127.0.0.1:7687", "user": "neo4j", "password": "12345678", - "db_name": "shared-tree-textual-memory", - "user_name": "alice", + "db_name": "neo4j", + "user_name": "xxx", "auto_create": true, - "use_multi_db": false, - "embedding_dimension": 768 + "use_multi_db": true, + "embedding_dimension": 1024 } }, - "reorganize": true + "reorganize": false, + "memory_size": { + "WorkingMemory": 200, + "LongTermMemory": 20000, + "UserMemory": 30000 + }, + "memory_filename": "tree_textual_memories.json" } diff --git a/examples/data/mem_cube_tree/textual_memory.json b/examples/data/mem_cube_tree/textual_memory.json new file mode 100644 index 000000000..91f426ca2 --- /dev/null +++ b/examples/data/mem_cube_tree/textual_memory.json @@ -0,0 +1,4219 @@ +{ + "nodes": [ + { + "id": "c1e2bfb3-4dd7-40f9-a347-a25a72223b2c", + "memory": "On January 13, 2026, at 8:51 PM, the user added another memory to demonstrate the concept of persistence.", + "metadata": { + "sources": [ + "\"{\\\"type\\\": \\\"chat\\\", \\\"role\\\": \\\"user\\\", \\\"chat_time\\\": \\\"08:51 PM on 13 January, 2026\\\", \\\"content\\\": \\\"This is a test memory for dump example\\\", \\\"lang\\\": \\\"en\\\"}\"", + "\"{\\\"type\\\": \\\"chat\\\", \\\"role\\\": \\\"user\\\", \\\"chat_time\\\": \\\"08:51 PM on 13 January, 2026\\\", \\\"content\\\": \\\"Another memory to demonstrate persistence\\\", \\\"lang\\\": \\\"en\\\"}\"" + ], + "confidence": 0.99, + "usage": [], + "created_at": "2026-01-13T20:51:42.394870000+00:00", + "session_id": "default_session", + "memory_type": "UserMemory", + "type": "fact", + "tags": [ + "persistence", + "memory", + "demonstration" + ], + "updated_at": "2026-01-13T20:51:42.396309000+00:00", + "user_id": "example_user", + "background": "On January 13, 2026, at 8:51 PM, the user engaged in a memory demonstration by creating a test memory example and another memory to illustrate the concept of persistence. These actions reflect the user's interest in exploring and demonstrating memory-related concepts.", + "file_ids": [], + "embedding": [ + -0.022425159811973572, + 0.01004026923328638, + 0.013415387831628323, + -0.013093948364257812, + -0.04768655076622963, + -0.047837819904088974, + 0.02147974818944931, + 0.019097311422228813, + 0.005332120228558779, + -0.005180854815989733, + 0.0216499213129282, + -0.006977136246860027, + -0.014067722484469414, + 0.014010997489094734, + -0.01164746843278408, + 0.037948813289403915, + 0.049539558589458466, + -0.028778323903679848, + -0.0363227054476738, + -0.006282258778810501, + 0.03201163187623024, + -0.043678008019924164, + 0.06640569865703583, + 0.04734620451927185, + -0.03552856296300888, + 0.008948319591581821, + -0.018199170008301735, + 0.015788370743393898, + 0.08047342300415039, + 0.04632515832781792, + -0.01887986622750759, + 0.00966683216392994, + -0.02516685239970684, + -0.00974719226360321, + -0.01631780155003071, + -0.01085332315415144, + 0.0284947007894516, + 0.012914320454001427, + -0.02427816577255726, + 0.006589517462998629, + -0.013954272493720055, + -0.0017856458434835076, + 0.004043997265398502, + -0.004202353768050671, + 0.032805778086185455, + -0.00395891023799777, + -0.002543156733736396, + -0.018889321014285088, + -0.02066669426858425, + -0.006225534249097109, + 0.008262895978987217, + 0.0266795102506876, + 0.055325478315353394, + -0.019683465361595154, + -0.013576108030974865, + 0.015372389927506447, + -0.05816170945763588, + -0.00842361617833376, + -0.009548655711114407, + -0.0019262757850810885, + -0.047535285353660583, + 0.026603877544403076, + -0.0035098399966955185, + -0.030290981754660606, + -0.015334573574364185, + 0.04798908531665802, + 0.028400158509612083, + -0.00544084282591939, + 0.008106903173029423, + -0.03639834001660347, + 0.011515110731124878, + 0.042278800159692764, + -0.07343956083059311, + -0.02628243714570999, + -0.08886867761611938, + 0.02773837186396122, + -0.0017962816637009382, + -0.007761827670037746, + -0.011363845318555832, + 0.031860366463661194, + 0.017783189192414284, + -0.026849685236811638, + 0.04144683852791786, + -0.00212244875729084, + 0.03042333945631981, + 0.026755142956972122, + -0.04500158503651619, + 0.03989636152982712, + 0.04318639263510704, + -0.03339193016290665, + 0.025412660092115402, + -0.01126930397003889, + 0.027587104588747025, + -0.03471550717949867, + -0.006041178945451975, + -0.007709830068051815, + -0.02686859294772148, + 0.0020196351688355207, + 0.05143038183450699, + 0.029402295127511024, + 0.059598736464977264, + 0.01989145576953888, + 0.013944818638265133, + -0.010550791397690773, + 0.011004588566720486, + -0.02439161390066147, + 0.05952310189604759, + -0.0021389934699982405, + -0.011723102070391178, + 0.007941455580294132, + 0.010248259641230106, + 0.029893908649683, + 0.033032674342393875, + 0.008130538277328014, + 0.025242485105991364, + -0.009198852814733982, + 0.04307294636964798, + 0.005294303875416517, + -0.030272074043750763, + -0.004263805691152811, + 0.04027452692389488, + 0.017197033390402794, + 0.06016598269343376, + -0.025620650500059128, + -0.02253860794007778, + -0.015088766813278198, + -0.007000771816819906, + -0.018870411440730095, + -0.015325118787586689, + -0.0027582377661019564, + 0.025734098628163338, + -0.017972271889448166, + 0.014824051409959793, + -0.025072310119867325, + -0.05165728181600571, + -0.00673132948577404, + 0.004667968954890966, + 0.03293813392519951, + 0.02083686739206314, + -0.03785427287220955, + 0.013188489712774754, + -0.02533702552318573, + 0.002592790871858597, + -0.03611471503973007, + 0.022084811702370644, + -0.007724011316895485, + 0.0028811413794755936, + 0.007497112732380629, + 0.026755142956972122, + 0.014861867763102055, + -0.012413252145051956, + 0.014880775474011898, + -0.03142547607421875, + -0.03377009555697441, + -0.024202533066272736, + 0.0020539062097668648, + 0.02323821373283863, + 0.09204525500535965, + -0.03365664556622505, + -0.031633466482162476, + -0.005308485124260187, + -0.06735111027956009, + -0.06126265972852707, + 0.04545538127422333, + 0.03220071271061897, + 0.028343435376882553, + 0.010843869298696518, + 0.02891068160533905, + -0.034488607197999954, + 0.006887322291731834, + -0.0065280660055577755, + -0.003923457581549883, + 0.02036416158080101, + 0.008220352232456207, + 0.0012183990329504013, + 0.0293266624212265, + 0.04390490800142288, + 0.01639343425631523, + -0.0038076445925980806, + -0.04159810394048691, + -0.057556647807359695, + -0.01890822872519493, + 0.04326202720403671, + -0.025355935096740723, + -0.045379746705293655, + -0.014578244648873806, + -0.0410686731338501, + 0.03469659760594368, + 0.01117476262152195, + 0.036190349608659744, + 0.020439794287085533, + -0.013093948364257812, + -0.03879968449473381, + 0.005251760594546795, + -0.028040902689099312, + -0.028229985386133194, + 0.005209216848015785, + -0.00709531269967556, + -0.029875000938773155, + 0.0010895866435021162, + -0.012006725184619427, + 0.07343956083059311, + 0.010541337542235851, + 0.02246297523379326, + -0.02633916214108467, + -0.020401978865265846, + -0.01178928092122078, + 0.025715190917253494, + 0.01966455765068531, + -0.07551946491003036, + -0.02388109266757965, + 0.005265941843390465, + -0.0036445611622184515, + -0.004528520628809929, + 0.018482793122529984, + -0.04511503502726555, + 0.012120174244046211, + 0.010635878890752792, + -0.008995589800179005, + -0.01553311012685299, + -0.030158624053001404, + 0.060014717280864716, + -0.056913767009973526, + 0.004339438397437334, + 0.032975949347019196, + -0.008612697944045067, + 0.0023138944525271654, + 0.03199272230267525, + 0.004360710270702839, + -0.008277077227830887, + 0.018066812306642532, + 0.02879723161458969, + 0.02745474874973297, + -0.010200989432632923, + 0.022746598348021507, + -0.03348647430539131, + -0.04084177315235138, + -0.006783327087759972, + 0.005696103908121586, + 0.0018281893571838737, + -0.01589236594736576, + -0.03482895717024803, + 0.0062491693533957005, + 0.0023871639277786016, + -0.034091535955667496, + -0.03195490688085556, + -0.014294620603322983, + 0.01899331621825695, + 0.017537381500005722, + 0.006069541443139315, + 0.03144438564777374, + 0.02019398845732212, + -0.06375854462385178, + -0.032087262719869614, + 0.005927729420363903, + 0.02123394049704075, + 0.034091535955667496, + -0.010465704835951328, + 0.041219938546419144, + 0.03170910105109215, + 0.03439406678080559, + -0.024996677413582802, + -0.0038998222444206476, + 0.03887531906366348, + 0.041219938546419144, + -0.01626107655465603, + -0.01804790459573269, + -0.0051997629925608635, + 0.023502927273511887, + -0.026547152549028397, + 0.029572470113635063, + -0.03140656650066376, + -0.023105856031179428, + 0.017924999818205833, + -0.005507021676748991, + -0.010134810581803322, + -0.024315981194376945, + -0.007908366620540619, + -0.024259258061647415, + 0.0360390841960907, + 0.02802199497818947, + -0.03195490688085556, + 0.023200396448373795, + -0.018123537302017212, + 0.022576425224542618, + -0.0413333885371685, + -0.03955601528286934, + 5.5653712479397655e-05, + 0.0209503173828125, + -0.04159810394048691, + -0.024070175364613533, + 0.02679296024143696, + 0.06493085622787476, + -0.033789005130529404, + -0.01567492075264454, + -0.02656606025993824, + -0.005653560161590576, + -0.16155190765857697, + 0.010730419307947159, + -0.021025950089097023, + 0.028532516211271286, + -0.010900593362748623, + -0.008319620974361897, + -0.03637943044304848, + -0.01583564095199108, + -0.04239225015044212, + 0.07003607600927353, + -0.03594454377889633, + -0.04863196238875389, + 0.027360206469893456, + 0.021252848207950592, + -0.02497776970267296, + -0.004530884325504303, + -0.02703876607120037, + -0.019815823063254356, + -0.03985854610800743, + -0.030877137556672096, + 0.012186353094875813, + -0.01614762656390667, + 0.03165237605571747, + -0.035131487995386124, + -0.008636333048343658, + 0.008645787835121155, + 0.002212262712419033, + -0.009577018208801746, + -0.055438924580812454, + 0.006929865572601557, + -0.010834414511919022, + 0.03229525312781334, + -0.011987816542387009, + -0.015410206280648708, + 0.02745474874973297, + 0.05207326263189316, + 0.009000317193567753, + -0.01176091842353344, + 0.023068038746714592, + -0.030971677973866463, + 0.026849685236811638, + 0.05237579345703125, + -0.024561788886785507, + 0.0284947007894516, + 0.020874684676527977, + -0.02405126579105854, + -0.020685601979494095, + 0.0006671059527434409, + -0.049653008580207825, + -0.010815506801009178, + -0.02312476374208927, + -0.028475793078541756, + -0.01626107655465603, + -0.008579608984291553, + -0.02042088657617569, + 0.011486749164760113, + -0.0293266624212265, + 0.013538291677832603, + -0.035377293825149536, + 0.009870095178484917, + -0.012413252145051956, + 0.00476250983774662, + -0.0022571696899831295, + 0.006447705905884504, + -0.04447215422987938, + 0.016828322783112526, + 0.033505380153656006, + -0.04341329261660576, + 7.6445379818324e-05, + -0.034791141748428345, + 0.05759446322917938, + -0.006835324689745903, + 0.002079905243590474, + -0.01290486566722393, + 0.010768236592411995, + 0.07635142654180527, + -0.011609652079641819, + -0.04632515832781792, + 0.01966455765068531, + -0.10089430958032608, + 0.019929273054003716, + 0.050711870193481445, + -0.0033467565663158894, + 0.00785164162516594, + -0.03653069585561752, + 0.022368434816598892, + 0.006537519861012697, + 0.004381982143968344, + 0.03955601528286934, + 0.2114696353673935, + 0.031690191477537155, + -0.05438006669282913, + 0.013112856075167656, + 0.041219938546419144, + -0.005327393300831318, + -0.01223362423479557, + 0.025242485105991364, + -0.056270889937877655, + -0.046552058309316635, + 0.015154944732785225, + -0.016459612175822258, + 0.00955338217318058, + 0.013727374374866486, + 0.03424280136823654, + 0.004329984541982412, + -0.024675238877534866, + 0.015438568778336048, + 0.08584335446357727, + 0.012999407015740871, + 0.016686512157320976, + -0.009118493646383286, + -0.004190536215901375, + 0.005601562559604645, + -0.034318435937166214, + -0.01955110765993595, + -0.004864141810685396, + 0.045606646686792374, + 0.02949683740735054, + 0.053396835923194885, + -0.01936202682554722, + 0.030688054859638214, + -0.002644788473844528, + 0.0024911591317504644, + -0.0174617487937212, + -0.014861867763102055, + -0.025488292798399925, + -0.06754019111394882, + -0.011590744368731976, + 0.018123537302017212, + 0.02720894105732441, + -0.009737737476825714, + 0.0340726263821125, + 0.0259799063205719, + -0.01553311012685299, + -0.011316575109958649, + 0.017471203580498695, + -0.03048006445169449, + -0.04371582344174385, + -0.022954588755965233, + -0.022028086706995964, + 0.024372706189751625, + 0.02620680443942547, + -0.023086946457624435, + 0.019967088475823402, + -0.03685213625431061, + -0.005171400494873524, + -0.010541337542235851, + 0.0012302165850996971, + 0.06027943268418312, + 0.029875000938773155, + 0.03199272230267525, + -0.0051146759651601315, + 0.03204944729804993, + -0.0076814680360257626, + 0.0007687376928515732, + 0.029515745118260384, + -0.02446724846959114, + 0.03348647430539131, + 0.02340838685631752, + 0.03902658447623253, + 0.013377571478486061, + 0.01896495372056961, + 0.020515428856015205, + 0.010938410647213459, + -0.0005574973183684051, + 0.02427816577255726, + 0.021876821294426918, + 0.016128718852996826, + 0.05600617453455925, + -0.011921637691557407, + -0.02773837186396122, + -0.07733465731143951, + 0.007166218478232622, + 0.020969225093722343, + -0.04950174316763878, + -0.04844288155436516, + 0.07752373814582825, + -0.0036587421782314777, + -0.03898876905441284, + -0.01647852174937725, + 0.049010127782821655, + -0.0048286886885762215, + -0.019626740366220474, + 0.05732974782586098, + -0.020628876984119415, + 0.07241851836442947, + 0.0036209258250892162, + -0.036549605429172516, + -0.016610879451036453, + -0.0497286394238472, + -0.026225712150335312, + 0.01734829880297184, + -0.0354718379676342, + -0.026433704420924187, + -0.005355755798518658, + -0.023673102259635925, + -0.0011675831628963351, + -0.0009330029133707285, + 0.005828461609780788, + 0.01135439146310091, + 0.011278758756816387, + -0.059371836483478546, + 0.010787144303321838, + 0.01670541986823082, + 0.008470886386930943, + -0.005596835631877184, + 0.005970273166894913, + -0.016270531341433525, + 0.026055539026856422, + 0.014701147563755512, + 0.004741238430142403, + 0.02008053846657276, + 0.05604398995637894, + 0.004296895116567612, + -0.002104722196236253, + 0.016043631359934807, + -0.019182397052645683, + -0.009676286019384861, + -0.01740502379834652, + 0.02580973133444786, + -0.019570017233490944, + 0.04764873534440994, + -0.022746598348021507, + 0.04935047775506973, + -0.009510839357972145, + 0.014578244648873806, + 0.017660286277532578, + -0.0019392751855775714, + 0.02645261213183403, + 0.008243987336754799, + -0.06999826431274414, + -0.00033325751428492367, + 0.010739874094724655, + 0.013169581070542336, + 0.03482895717024803, + -0.0050059533677995205, + -0.01709303818643093, + -0.021971361711621284, + 0.016941772773861885, + 0.0410686731338501, + 0.026698417961597443, + 0.04243006557226181, + -0.049426108598709106, + -0.03530166298151016, + -0.021631013602018356, + -0.038648419082164764, + 0.022860048338770866, + 0.02896740660071373, + -0.0329570434987545, + 0.0008916411316022277, + 0.025677373632788658, + 0.05101440101861954, + 0.0709436759352684, + 0.057443197816610336, + -0.0075018396601080894, + -0.019702374935150146, + 0.008073813281953335, + -0.02323821373283863, + 0.0043016220442950726, + 0.005232852417975664, + -0.018671875819563866, + -0.01896495372056961, + 0.012696875259280205, + -0.003824189305305481, + -0.021498655900359154, + -0.040463607758283615, + 0.06292658299207687, + -0.008726147934794426, + 0.00958174467086792, + 0.008849050849676132, + -0.021044857800006866, + -0.025355935096740723, + -0.02193354442715645, + 0.041219938546419144, + 0.03013971634209156, + 0.03628489002585411, + -0.03866732865571976, + 0.03148220106959343, + -0.006480795331299305, + -0.06251060217618942, + 0.0854651927947998, + 0.028248893097043037, + -0.01910676434636116, + 0.02620680443942547, + -0.05116566643118858, + 0.019380934536457062, + 0.021555380895733833, + 0.0031009495723992586, + -0.0020385433454066515, + -0.021801186725497246, + 0.013935364782810211, + -0.015400752425193787, + -0.0027086036279797554, + -0.001123857800848782, + 0.022028086706995964, + 0.01639343425631523, + 0.01609090156853199, + -0.016998497769236565, + -0.02265205793082714, + -0.0029922272078692913, + -0.04299731180071831, + -0.021460838615894318, + 0.017282120883464813, + -0.04364019259810448, + -0.019475474953651428, + 0.06614098697900772, + 0.03450751677155495, + -1.5242912922985852e-05, + -0.016941772773861885, + 0.0027251483406871557, + 0.03061242215335369, + 0.03271123394370079, + -0.051695097237825394, + -0.011023497208952904, + 0.017527928575873375, + 0.027133308351039886, + -0.027757279574871063, + 0.023313846439123154, + 0.017887184396386147, + -0.01006863173097372, + -0.030536789447069168, + -0.022084811702370644, + -0.0037651010788977146, + 0.04095522314310074, + -0.06035506725311279, + 0.015485838986933231, + 0.014285166747868061, + 0.0011563563020899892, + -0.02253860794007778, + -0.005549564957618713, + -0.015079312026500702, + 0.0031553106382489204, + 0.012895411811769009, + 0.03948038071393967, + 0.04492595046758652, + -0.006863687187433243, + -0.02352183684706688, + 0.018331527709960938, + 0.0300262663513422, + 0.004287440795451403, + 0.004887776914983988, + -0.0340726263821125, + -0.024958861991763115, + 0.0438670888543129, + 0.03278686851263046, + 0.020742326974868774, + -0.013519383035600185, + -0.06307785212993622, + -0.04855633154511452, + -0.004703422077000141, + 0.036133624613285065, + -0.03594454377889633, + 0.0012633060105144978, + -0.06288877129554749, + -0.0413333885371685, + 0.01810462959110737, + -0.06330475211143494, + 0.000763419724535197, + -0.027171123772859573, + 0.0096384696662426, + -0.009543928317725658, + -0.012971044518053532, + -0.03136875107884407, + 0.058426424860954285, + 0.03386463597416878, + -0.007010225672274828, + -0.05173291265964508, + 0.05850205942988396, + 0.03488568216562271, + -0.011751463636755943, + 0.046438608318567276, + 0.039934176951646805, + 0.04927484318614006, + -0.043980538845062256, + -0.00198654574342072, + 0.01614762656390667, + 0.0009625470265746117, + 0.03339193016290665, + -0.0865996852517128, + 0.021252848207950592, + 0.007208762224763632, + -0.03583109378814697, + 0.04904794320464134, + 0.006003362592309713, + 0.008792325854301453, + -0.0311418529599905, + 0.01097622700035572, + -0.02458069659769535, + 0.018085720017552376, + -0.017669739201664925, + -0.011430024169385433, + -0.039820730686187744, + -0.04084177315235138, + -0.005908821243792772, + -0.003512203460559249, + -0.00965265091508627, + -0.0047388747334480286, + 0.0592205710709095, + -0.0059135486371815205, + -0.021517563611268997, + -0.04401835426688194, + 0.04284604638814926, + 0.04632515832781792, + -0.011458386667072773, + -0.018482793122529984, + 0.027927452698349953, + -0.06493085622787476, + 0.0351882129907608, + 0.0034814774990081787, + 0.03804335743188858, + 0.025885364040732384, + -0.04159810394048691, + 0.008589062839746475, + -0.0035500198137015104, + -0.008442523889243603, + 0.005483386106789112, + -0.026320254430174828, + 0.008669422939419746, + -0.027492564171552658, + -0.012148536741733551, + -0.015495292842388153, + -0.029648102819919586, + -0.006665150634944439, + 0.05721629783511162, + 0.026603877544403076, + 0.015240032225847244, + -0.024429431185126305, + 0.04295949637889862, + -0.04927484318614006, + -0.05165728181600571, + -0.036719780415296555, + 0.036833230406045914, + -0.04791345074772835, + -0.02650933712720871, + 0.020855776965618134, + -0.0036776503548026085, + 0.006182990502566099, + -0.029893908649683, + 0.007118948269635439, + 0.0009324120474047959, + 0.04556883126497269, + -0.024202533066272736, + -0.07865823060274124, + -0.004027452785521746, + -0.004242533817887306, + -0.03898876905441284, + -0.006078995298594236, + 0.033618830144405365, + -0.013245213776826859, + -0.01378409843891859, + 0.01681886985898018, + -0.014842959120869637, + 0.0032687599305063486, + -0.04700585454702377, + 0.009071222506463528, + 0.026471519842743874, + 0.030045175924897194, + 0.006863687187433243, + -0.03429952636361122, + 0.018842048943042755, + -0.01849224790930748, + 0.004533248022198677, + -0.021574288606643677, + -0.023673102259635925, + -0.010730419307947159, + 0.012243078090250492, + -0.03978291153907776, + 0.03781645745038986, + 0.047724369913339615, + -0.03592563420534134, + -0.021215032786130905, + 0.014757872559130192, + -0.040652692317962646, + 0.004173991270363331, + -0.04352674260735512, + -0.028116535395383835, + -0.009619561024010181, + 0.011836551129817963, + 0.022897865623235703, + -0.05528765916824341, + -0.025355935096740723, + -0.017442841082811356, + 0.017111947759985924, + -0.16896392405033112, + 0.029856093227863312, + 0.016194898635149002, + 0.016866140067577362, + -0.04326202720403671, + -0.013509929180145264, + -0.05611962080001831, + 0.006154628470540047, + 0.014616061002016068, + -0.021290665492415428, + 0.0066131530329585075, + -0.04564446210861206, + 0.06096012890338898, + -0.03384573012590408, + -0.027416931465268135, + 0.010229351930320263, + 0.006901503540575504, + 0.057027217000722885, + -0.05388845130801201, + 0.06186772510409355, + 0.009699921123683453, + -0.01977800764143467, + -0.0142473503947258, + -0.0002583632303867489, + -0.017244303598999977, + -0.01770755648612976, + 0.019796915352344513, + 0.03055569715797901, + -0.030158624053001404, + 0.05131693184375763, + 0.0026967860758304596, + 0.02441052347421646, + 0.019182397052645683, + 0.041749369353055954, + 0.03246542811393738, + 0.0048452336341142654, + -0.0012266712728887796, + -0.03808117285370827, + -0.0015800188994035125, + -0.01989145576953888, + -0.021044857800006866, + 0.02153647318482399, + -0.01650688424706459, + -0.01559928897768259, + 0.004192899446934462, + 0.09136456251144409, + -0.028929589316248894, + -0.003526384709402919, + -0.02592318132519722, + 0.03384573012590408, + 0.015240032225847244, + 0.03781645745038986, + 0.009520293213427067, + 0.011524565517902374, + -0.017631923779845238, + -0.005559019278734922, + 0.01158128958195448, + -0.019031131640076637, + -0.024032358080148697, + 0.014776780270040035, + 0.014086630195379257, + -0.019314754754304886, + -0.03161455690860748, + -0.0729101300239563, + -0.017074130475521088, + 0.006750237662345171, + -0.07007389515638351, + -0.007246578577905893, + -0.025091219693422318, + -0.013349208980798721, + -0.025885364040732384, + 0.029950633645057678, + 0.026887500658631325, + -0.0005920638795942068, + -0.012838686816394329, + 0.048064716160297394, + 0.03757065162062645, + 0.01751847378909588, + -0.013821914792060852, + -0.026244621723890305, + -0.027019858360290527, + -0.011808188632130623, + -0.0006871959194540977, + -0.03278686851263046, + 0.035377293825149536, + 0.013056132011115551, + -0.025960998609662056, + -0.012469977140426636, + -0.0031482200138270855, + -0.0007728738128207624, + -0.08077595382928848, + 0.015088766813278198, + 0.056497786194086075, + 0.01623271405696869, + -0.009293394163250923, + 0.06659477949142456, + -0.01989145576953888, + -0.008598516695201397, + -0.015192762017250061, + -0.0026802413631230593, + 0.00839525368064642, + -0.001349574769847095, + 0.02866487391293049, + 0.005842642392963171, + -0.025942089036107063, + 0.0029638647101819515, + -0.024240348488092422, + -0.03524493798613548, + 0.021876821294426918, + -0.010720965452492237, + -0.005984454415738583, + -0.025771915912628174, + -0.04863196238875389, + 0.008239260874688625, + -0.06856123358011246, + -0.03305158391594887, + -0.032162897288799286, + 0.03318393975496292, + -0.0441318042576313, + -0.02000490576028824, + 0.03225743770599365, + 0.03131202608346939, + 0.048896677792072296, + -0.03499913215637207, + 0.024297073483467102, + 0.000977909890934825, + 0.030101899057626724, + 0.015882911160588264, + 0.018000634387135506, + -0.006197171751409769, + -0.006892049219459295, + -0.00835270993411541, + 0.02195245400071144, + 0.039593830704689026, + -0.010238805785775185, + 0.00586627796292305, + 0.0156465582549572, + 0.005800099112093449, + -0.015230578370392323, + -0.031690191477537155, + 0.006849505938589573, + -0.028986314311623573, + 0.02253860794007778, + -0.01421898789703846, + -0.014795688912272453, + -0.019872548058629036, + 0.018331527709960938, + -0.01570328325033188, + -0.019116219133138657, + -0.011864913627505302, + 0.002848051954060793, + 0.00756329158321023, + 0.032919224351644516, + 0.06387199461460114, + -0.01949438452720642, + -0.013112856075167656, + 0.004374891519546509, + 0.022311709821224213, + 0.01765083149075508, + 0.02304913103580475, + -0.034564241766929626, + -0.00717094587162137, + -0.05332120507955551, + -0.05154383182525635, + -0.003493295283988118, + 0.022444067522883415, + 0.037589557468891144, + -0.0087970532476902, + -0.06220807135105133, + 0.004157446790486574, + 0.009151582606136799, + 0.06288877129554749, + -0.030820412561297417, + 0.011902729980647564, + 0.003717830404639244, + -0.056157439947128296, + 0.024883229285478592, + 0.0030016812961548567, + 0.02505340240895748, + -0.026717327535152435, + 0.025185760110616684, + -0.01331139262765646, + 0.005445569753646851, + 0.02475087158381939, + 0.024202533066272736, + 0.036833230406045914, + -0.002345802029594779, + 0.023427294567227364, + -0.011184217408299446, + -0.01375573594123125, + 0.02312476374208927, + -0.029515745118260384, + -0.05910712108016014, + -0.017631923779845238, + 0.01743338629603386, + -0.024486156180500984, + -0.0026613331865519285, + 0.01509822066873312, + -0.011042405851185322, + 0.03392136096954346, + -0.036247074604034424, + -0.04787563532590866, + 0.030064083635807037, + -0.022670965641736984, + 0.005308485124260187, + 0.014786235056817532, + -0.027284573763608932, + 0.009671558625996113, + -0.0360390841960907, + 0.02153647318482399, + -0.05188417807221413, + -0.008995589800179005, + 0.011051859706640244, + -0.0006470159278251231, + -0.020269621163606644, + 0.02229280211031437, + 0.011108584702014923, + -0.013736828230321407, + -0.009841732680797577, + 0.034148260951042175, + -0.0014807506231591105, + -0.0018388251774013042, + 0.013500475324690342, + -0.026849685236811638, + -0.060014717280864716, + -0.008404707536101341, + 0.011212579905986786, + 0.02047761157155037, + -0.0016426523216068745, + 0.0483294315636158, + 0.022444067522883415, + 0.036719780415296555, + -0.03861060366034508, + 0.037362661212682724, + -0.006286986172199249, + -0.02773837186396122, + 0.033448655158281326, + 0.019985998049378395, + -0.03745720162987709, + -0.005204489920288324, + 0.016487974673509598, + -0.019314754754304886, + -0.0555901937186718, + -0.008636333048343658 + ], + "key": "Persistence Demonstration", + "status": "activated" + } + }, + { + "id": "76374204-35a7-49af-a437-29410ae353e3", + "memory": "On January 13, 2026, at 8:51 PM, the user added another memory to demonstrate the concept of persistence.", + "metadata": { + "sources": [ + "\"{\\\"type\\\": \\\"chat\\\", \\\"role\\\": \\\"user\\\", \\\"chat_time\\\": \\\"08:51 PM on 13 January, 2026\\\", \\\"content\\\": \\\"This is a test memory for dump example\\\", \\\"lang\\\": \\\"en\\\"}\"", + "\"{\\\"type\\\": \\\"chat\\\", \\\"role\\\": \\\"user\\\", \\\"chat_time\\\": \\\"08:51 PM on 13 January, 2026\\\", \\\"content\\\": \\\"Another memory to demonstrate persistence\\\", \\\"lang\\\": \\\"en\\\"}\"" + ], + "confidence": 0.99, + "usage": [], + "created_at": "2026-01-13T20:51:42.394870000+00:00", + "session_id": "default_session", + "memory_type": "WorkingMemory", + "type": "fact", + "tags": [ + "persistence", + "memory", + "demonstration" + ], + "updated_at": "2026-01-13T20:51:42.396281000+00:00", + "user_id": "example_user", + "background": "On January 13, 2026, at 8:51 PM, the user engaged in a memory demonstration by creating a test memory example and another memory to illustrate the concept of persistence. These actions reflect the user's interest in exploring and demonstrating memory-related concepts.", + "file_ids": [], + "embedding": [ + -0.022425159811973572, + 0.01004026923328638, + 0.013415387831628323, + -0.013093948364257812, + -0.04768655076622963, + -0.047837819904088974, + 0.02147974818944931, + 0.019097311422228813, + 0.005332120228558779, + -0.005180854815989733, + 0.0216499213129282, + -0.006977136246860027, + -0.014067722484469414, + 0.014010997489094734, + -0.01164746843278408, + 0.037948813289403915, + 0.049539558589458466, + -0.028778323903679848, + -0.0363227054476738, + -0.006282258778810501, + 0.03201163187623024, + -0.043678008019924164, + 0.06640569865703583, + 0.04734620451927185, + -0.03552856296300888, + 0.008948319591581821, + -0.018199170008301735, + 0.015788370743393898, + 0.08047342300415039, + 0.04632515832781792, + -0.01887986622750759, + 0.00966683216392994, + -0.02516685239970684, + -0.00974719226360321, + -0.01631780155003071, + -0.01085332315415144, + 0.0284947007894516, + 0.012914320454001427, + -0.02427816577255726, + 0.006589517462998629, + -0.013954272493720055, + -0.0017856458434835076, + 0.004043997265398502, + -0.004202353768050671, + 0.032805778086185455, + -0.00395891023799777, + -0.002543156733736396, + -0.018889321014285088, + -0.02066669426858425, + -0.006225534249097109, + 0.008262895978987217, + 0.0266795102506876, + 0.055325478315353394, + -0.019683465361595154, + -0.013576108030974865, + 0.015372389927506447, + -0.05816170945763588, + -0.00842361617833376, + -0.009548655711114407, + -0.0019262757850810885, + -0.047535285353660583, + 0.026603877544403076, + -0.0035098399966955185, + -0.030290981754660606, + -0.015334573574364185, + 0.04798908531665802, + 0.028400158509612083, + -0.00544084282591939, + 0.008106903173029423, + -0.03639834001660347, + 0.011515110731124878, + 0.042278800159692764, + -0.07343956083059311, + -0.02628243714570999, + -0.08886867761611938, + 0.02773837186396122, + -0.0017962816637009382, + -0.007761827670037746, + -0.011363845318555832, + 0.031860366463661194, + 0.017783189192414284, + -0.026849685236811638, + 0.04144683852791786, + -0.00212244875729084, + 0.03042333945631981, + 0.026755142956972122, + -0.04500158503651619, + 0.03989636152982712, + 0.04318639263510704, + -0.03339193016290665, + 0.025412660092115402, + -0.01126930397003889, + 0.027587104588747025, + -0.03471550717949867, + -0.006041178945451975, + -0.007709830068051815, + -0.02686859294772148, + 0.0020196351688355207, + 0.05143038183450699, + 0.029402295127511024, + 0.059598736464977264, + 0.01989145576953888, + 0.013944818638265133, + -0.010550791397690773, + 0.011004588566720486, + -0.02439161390066147, + 0.05952310189604759, + -0.0021389934699982405, + -0.011723102070391178, + 0.007941455580294132, + 0.010248259641230106, + 0.029893908649683, + 0.033032674342393875, + 0.008130538277328014, + 0.025242485105991364, + -0.009198852814733982, + 0.04307294636964798, + 0.005294303875416517, + -0.030272074043750763, + -0.004263805691152811, + 0.04027452692389488, + 0.017197033390402794, + 0.06016598269343376, + -0.025620650500059128, + -0.02253860794007778, + -0.015088766813278198, + -0.007000771816819906, + -0.018870411440730095, + -0.015325118787586689, + -0.0027582377661019564, + 0.025734098628163338, + -0.017972271889448166, + 0.014824051409959793, + -0.025072310119867325, + -0.05165728181600571, + -0.00673132948577404, + 0.004667968954890966, + 0.03293813392519951, + 0.02083686739206314, + -0.03785427287220955, + 0.013188489712774754, + -0.02533702552318573, + 0.002592790871858597, + -0.03611471503973007, + 0.022084811702370644, + -0.007724011316895485, + 0.0028811413794755936, + 0.007497112732380629, + 0.026755142956972122, + 0.014861867763102055, + -0.012413252145051956, + 0.014880775474011898, + -0.03142547607421875, + -0.03377009555697441, + -0.024202533066272736, + 0.0020539062097668648, + 0.02323821373283863, + 0.09204525500535965, + -0.03365664556622505, + -0.031633466482162476, + -0.005308485124260187, + -0.06735111027956009, + -0.06126265972852707, + 0.04545538127422333, + 0.03220071271061897, + 0.028343435376882553, + 0.010843869298696518, + 0.02891068160533905, + -0.034488607197999954, + 0.006887322291731834, + -0.0065280660055577755, + -0.003923457581549883, + 0.02036416158080101, + 0.008220352232456207, + 0.0012183990329504013, + 0.0293266624212265, + 0.04390490800142288, + 0.01639343425631523, + -0.0038076445925980806, + -0.04159810394048691, + -0.057556647807359695, + -0.01890822872519493, + 0.04326202720403671, + -0.025355935096740723, + -0.045379746705293655, + -0.014578244648873806, + -0.0410686731338501, + 0.03469659760594368, + 0.01117476262152195, + 0.036190349608659744, + 0.020439794287085533, + -0.013093948364257812, + -0.03879968449473381, + 0.005251760594546795, + -0.028040902689099312, + -0.028229985386133194, + 0.005209216848015785, + -0.00709531269967556, + -0.029875000938773155, + 0.0010895866435021162, + -0.012006725184619427, + 0.07343956083059311, + 0.010541337542235851, + 0.02246297523379326, + -0.02633916214108467, + -0.020401978865265846, + -0.01178928092122078, + 0.025715190917253494, + 0.01966455765068531, + -0.07551946491003036, + -0.02388109266757965, + 0.005265941843390465, + -0.0036445611622184515, + -0.004528520628809929, + 0.018482793122529984, + -0.04511503502726555, + 0.012120174244046211, + 0.010635878890752792, + -0.008995589800179005, + -0.01553311012685299, + -0.030158624053001404, + 0.060014717280864716, + -0.056913767009973526, + 0.004339438397437334, + 0.032975949347019196, + -0.008612697944045067, + 0.0023138944525271654, + 0.03199272230267525, + 0.004360710270702839, + -0.008277077227830887, + 0.018066812306642532, + 0.02879723161458969, + 0.02745474874973297, + -0.010200989432632923, + 0.022746598348021507, + -0.03348647430539131, + -0.04084177315235138, + -0.006783327087759972, + 0.005696103908121586, + 0.0018281893571838737, + -0.01589236594736576, + -0.03482895717024803, + 0.0062491693533957005, + 0.0023871639277786016, + -0.034091535955667496, + -0.03195490688085556, + -0.014294620603322983, + 0.01899331621825695, + 0.017537381500005722, + 0.006069541443139315, + 0.03144438564777374, + 0.02019398845732212, + -0.06375854462385178, + -0.032087262719869614, + 0.005927729420363903, + 0.02123394049704075, + 0.034091535955667496, + -0.010465704835951328, + 0.041219938546419144, + 0.03170910105109215, + 0.03439406678080559, + -0.024996677413582802, + -0.0038998222444206476, + 0.03887531906366348, + 0.041219938546419144, + -0.01626107655465603, + -0.01804790459573269, + -0.0051997629925608635, + 0.023502927273511887, + -0.026547152549028397, + 0.029572470113635063, + -0.03140656650066376, + -0.023105856031179428, + 0.017924999818205833, + -0.005507021676748991, + -0.010134810581803322, + -0.024315981194376945, + -0.007908366620540619, + -0.024259258061647415, + 0.0360390841960907, + 0.02802199497818947, + -0.03195490688085556, + 0.023200396448373795, + -0.018123537302017212, + 0.022576425224542618, + -0.0413333885371685, + -0.03955601528286934, + 5.5653712479397655e-05, + 0.0209503173828125, + -0.04159810394048691, + -0.024070175364613533, + 0.02679296024143696, + 0.06493085622787476, + -0.033789005130529404, + -0.01567492075264454, + -0.02656606025993824, + -0.005653560161590576, + -0.16155190765857697, + 0.010730419307947159, + -0.021025950089097023, + 0.028532516211271286, + -0.010900593362748623, + -0.008319620974361897, + -0.03637943044304848, + -0.01583564095199108, + -0.04239225015044212, + 0.07003607600927353, + -0.03594454377889633, + -0.04863196238875389, + 0.027360206469893456, + 0.021252848207950592, + -0.02497776970267296, + -0.004530884325504303, + -0.02703876607120037, + -0.019815823063254356, + -0.03985854610800743, + -0.030877137556672096, + 0.012186353094875813, + -0.01614762656390667, + 0.03165237605571747, + -0.035131487995386124, + -0.008636333048343658, + 0.008645787835121155, + 0.002212262712419033, + -0.009577018208801746, + -0.055438924580812454, + 0.006929865572601557, + -0.010834414511919022, + 0.03229525312781334, + -0.011987816542387009, + -0.015410206280648708, + 0.02745474874973297, + 0.05207326263189316, + 0.009000317193567753, + -0.01176091842353344, + 0.023068038746714592, + -0.030971677973866463, + 0.026849685236811638, + 0.05237579345703125, + -0.024561788886785507, + 0.0284947007894516, + 0.020874684676527977, + -0.02405126579105854, + -0.020685601979494095, + 0.0006671059527434409, + -0.049653008580207825, + -0.010815506801009178, + -0.02312476374208927, + -0.028475793078541756, + -0.01626107655465603, + -0.008579608984291553, + -0.02042088657617569, + 0.011486749164760113, + -0.0293266624212265, + 0.013538291677832603, + -0.035377293825149536, + 0.009870095178484917, + -0.012413252145051956, + 0.00476250983774662, + -0.0022571696899831295, + 0.006447705905884504, + -0.04447215422987938, + 0.016828322783112526, + 0.033505380153656006, + -0.04341329261660576, + 7.6445379818324e-05, + -0.034791141748428345, + 0.05759446322917938, + -0.006835324689745903, + 0.002079905243590474, + -0.01290486566722393, + 0.010768236592411995, + 0.07635142654180527, + -0.011609652079641819, + -0.04632515832781792, + 0.01966455765068531, + -0.10089430958032608, + 0.019929273054003716, + 0.050711870193481445, + -0.0033467565663158894, + 0.00785164162516594, + -0.03653069585561752, + 0.022368434816598892, + 0.006537519861012697, + 0.004381982143968344, + 0.03955601528286934, + 0.2114696353673935, + 0.031690191477537155, + -0.05438006669282913, + 0.013112856075167656, + 0.041219938546419144, + -0.005327393300831318, + -0.01223362423479557, + 0.025242485105991364, + -0.056270889937877655, + -0.046552058309316635, + 0.015154944732785225, + -0.016459612175822258, + 0.00955338217318058, + 0.013727374374866486, + 0.03424280136823654, + 0.004329984541982412, + -0.024675238877534866, + 0.015438568778336048, + 0.08584335446357727, + 0.012999407015740871, + 0.016686512157320976, + -0.009118493646383286, + -0.004190536215901375, + 0.005601562559604645, + -0.034318435937166214, + -0.01955110765993595, + -0.004864141810685396, + 0.045606646686792374, + 0.02949683740735054, + 0.053396835923194885, + -0.01936202682554722, + 0.030688054859638214, + -0.002644788473844528, + 0.0024911591317504644, + -0.0174617487937212, + -0.014861867763102055, + -0.025488292798399925, + -0.06754019111394882, + -0.011590744368731976, + 0.018123537302017212, + 0.02720894105732441, + -0.009737737476825714, + 0.0340726263821125, + 0.0259799063205719, + -0.01553311012685299, + -0.011316575109958649, + 0.017471203580498695, + -0.03048006445169449, + -0.04371582344174385, + -0.022954588755965233, + -0.022028086706995964, + 0.024372706189751625, + 0.02620680443942547, + -0.023086946457624435, + 0.019967088475823402, + -0.03685213625431061, + -0.005171400494873524, + -0.010541337542235851, + 0.0012302165850996971, + 0.06027943268418312, + 0.029875000938773155, + 0.03199272230267525, + -0.0051146759651601315, + 0.03204944729804993, + -0.0076814680360257626, + 0.0007687376928515732, + 0.029515745118260384, + -0.02446724846959114, + 0.03348647430539131, + 0.02340838685631752, + 0.03902658447623253, + 0.013377571478486061, + 0.01896495372056961, + 0.020515428856015205, + 0.010938410647213459, + -0.0005574973183684051, + 0.02427816577255726, + 0.021876821294426918, + 0.016128718852996826, + 0.05600617453455925, + -0.011921637691557407, + -0.02773837186396122, + -0.07733465731143951, + 0.007166218478232622, + 0.020969225093722343, + -0.04950174316763878, + -0.04844288155436516, + 0.07752373814582825, + -0.0036587421782314777, + -0.03898876905441284, + -0.01647852174937725, + 0.049010127782821655, + -0.0048286886885762215, + -0.019626740366220474, + 0.05732974782586098, + -0.020628876984119415, + 0.07241851836442947, + 0.0036209258250892162, + -0.036549605429172516, + -0.016610879451036453, + -0.0497286394238472, + -0.026225712150335312, + 0.01734829880297184, + -0.0354718379676342, + -0.026433704420924187, + -0.005355755798518658, + -0.023673102259635925, + -0.0011675831628963351, + -0.0009330029133707285, + 0.005828461609780788, + 0.01135439146310091, + 0.011278758756816387, + -0.059371836483478546, + 0.010787144303321838, + 0.01670541986823082, + 0.008470886386930943, + -0.005596835631877184, + 0.005970273166894913, + -0.016270531341433525, + 0.026055539026856422, + 0.014701147563755512, + 0.004741238430142403, + 0.02008053846657276, + 0.05604398995637894, + 0.004296895116567612, + -0.002104722196236253, + 0.016043631359934807, + -0.019182397052645683, + -0.009676286019384861, + -0.01740502379834652, + 0.02580973133444786, + -0.019570017233490944, + 0.04764873534440994, + -0.022746598348021507, + 0.04935047775506973, + -0.009510839357972145, + 0.014578244648873806, + 0.017660286277532578, + -0.0019392751855775714, + 0.02645261213183403, + 0.008243987336754799, + -0.06999826431274414, + -0.00033325751428492367, + 0.010739874094724655, + 0.013169581070542336, + 0.03482895717024803, + -0.0050059533677995205, + -0.01709303818643093, + -0.021971361711621284, + 0.016941772773861885, + 0.0410686731338501, + 0.026698417961597443, + 0.04243006557226181, + -0.049426108598709106, + -0.03530166298151016, + -0.021631013602018356, + -0.038648419082164764, + 0.022860048338770866, + 0.02896740660071373, + -0.0329570434987545, + 0.0008916411316022277, + 0.025677373632788658, + 0.05101440101861954, + 0.0709436759352684, + 0.057443197816610336, + -0.0075018396601080894, + -0.019702374935150146, + 0.008073813281953335, + -0.02323821373283863, + 0.0043016220442950726, + 0.005232852417975664, + -0.018671875819563866, + -0.01896495372056961, + 0.012696875259280205, + -0.003824189305305481, + -0.021498655900359154, + -0.040463607758283615, + 0.06292658299207687, + -0.008726147934794426, + 0.00958174467086792, + 0.008849050849676132, + -0.021044857800006866, + -0.025355935096740723, + -0.02193354442715645, + 0.041219938546419144, + 0.03013971634209156, + 0.03628489002585411, + -0.03866732865571976, + 0.03148220106959343, + -0.006480795331299305, + -0.06251060217618942, + 0.0854651927947998, + 0.028248893097043037, + -0.01910676434636116, + 0.02620680443942547, + -0.05116566643118858, + 0.019380934536457062, + 0.021555380895733833, + 0.0031009495723992586, + -0.0020385433454066515, + -0.021801186725497246, + 0.013935364782810211, + -0.015400752425193787, + -0.0027086036279797554, + -0.001123857800848782, + 0.022028086706995964, + 0.01639343425631523, + 0.01609090156853199, + -0.016998497769236565, + -0.02265205793082714, + -0.0029922272078692913, + -0.04299731180071831, + -0.021460838615894318, + 0.017282120883464813, + -0.04364019259810448, + -0.019475474953651428, + 0.06614098697900772, + 0.03450751677155495, + -1.5242912922985852e-05, + -0.016941772773861885, + 0.0027251483406871557, + 0.03061242215335369, + 0.03271123394370079, + -0.051695097237825394, + -0.011023497208952904, + 0.017527928575873375, + 0.027133308351039886, + -0.027757279574871063, + 0.023313846439123154, + 0.017887184396386147, + -0.01006863173097372, + -0.030536789447069168, + -0.022084811702370644, + -0.0037651010788977146, + 0.04095522314310074, + -0.06035506725311279, + 0.015485838986933231, + 0.014285166747868061, + 0.0011563563020899892, + -0.02253860794007778, + -0.005549564957618713, + -0.015079312026500702, + 0.0031553106382489204, + 0.012895411811769009, + 0.03948038071393967, + 0.04492595046758652, + -0.006863687187433243, + -0.02352183684706688, + 0.018331527709960938, + 0.0300262663513422, + 0.004287440795451403, + 0.004887776914983988, + -0.0340726263821125, + -0.024958861991763115, + 0.0438670888543129, + 0.03278686851263046, + 0.020742326974868774, + -0.013519383035600185, + -0.06307785212993622, + -0.04855633154511452, + -0.004703422077000141, + 0.036133624613285065, + -0.03594454377889633, + 0.0012633060105144978, + -0.06288877129554749, + -0.0413333885371685, + 0.01810462959110737, + -0.06330475211143494, + 0.000763419724535197, + -0.027171123772859573, + 0.0096384696662426, + -0.009543928317725658, + -0.012971044518053532, + -0.03136875107884407, + 0.058426424860954285, + 0.03386463597416878, + -0.007010225672274828, + -0.05173291265964508, + 0.05850205942988396, + 0.03488568216562271, + -0.011751463636755943, + 0.046438608318567276, + 0.039934176951646805, + 0.04927484318614006, + -0.043980538845062256, + -0.00198654574342072, + 0.01614762656390667, + 0.0009625470265746117, + 0.03339193016290665, + -0.0865996852517128, + 0.021252848207950592, + 0.007208762224763632, + -0.03583109378814697, + 0.04904794320464134, + 0.006003362592309713, + 0.008792325854301453, + -0.0311418529599905, + 0.01097622700035572, + -0.02458069659769535, + 0.018085720017552376, + -0.017669739201664925, + -0.011430024169385433, + -0.039820730686187744, + -0.04084177315235138, + -0.005908821243792772, + -0.003512203460559249, + -0.00965265091508627, + -0.0047388747334480286, + 0.0592205710709095, + -0.0059135486371815205, + -0.021517563611268997, + -0.04401835426688194, + 0.04284604638814926, + 0.04632515832781792, + -0.011458386667072773, + -0.018482793122529984, + 0.027927452698349953, + -0.06493085622787476, + 0.0351882129907608, + 0.0034814774990081787, + 0.03804335743188858, + 0.025885364040732384, + -0.04159810394048691, + 0.008589062839746475, + -0.0035500198137015104, + -0.008442523889243603, + 0.005483386106789112, + -0.026320254430174828, + 0.008669422939419746, + -0.027492564171552658, + -0.012148536741733551, + -0.015495292842388153, + -0.029648102819919586, + -0.006665150634944439, + 0.05721629783511162, + 0.026603877544403076, + 0.015240032225847244, + -0.024429431185126305, + 0.04295949637889862, + -0.04927484318614006, + -0.05165728181600571, + -0.036719780415296555, + 0.036833230406045914, + -0.04791345074772835, + -0.02650933712720871, + 0.020855776965618134, + -0.0036776503548026085, + 0.006182990502566099, + -0.029893908649683, + 0.007118948269635439, + 0.0009324120474047959, + 0.04556883126497269, + -0.024202533066272736, + -0.07865823060274124, + -0.004027452785521746, + -0.004242533817887306, + -0.03898876905441284, + -0.006078995298594236, + 0.033618830144405365, + -0.013245213776826859, + -0.01378409843891859, + 0.01681886985898018, + -0.014842959120869637, + 0.0032687599305063486, + -0.04700585454702377, + 0.009071222506463528, + 0.026471519842743874, + 0.030045175924897194, + 0.006863687187433243, + -0.03429952636361122, + 0.018842048943042755, + -0.01849224790930748, + 0.004533248022198677, + -0.021574288606643677, + -0.023673102259635925, + -0.010730419307947159, + 0.012243078090250492, + -0.03978291153907776, + 0.03781645745038986, + 0.047724369913339615, + -0.03592563420534134, + -0.021215032786130905, + 0.014757872559130192, + -0.040652692317962646, + 0.004173991270363331, + -0.04352674260735512, + -0.028116535395383835, + -0.009619561024010181, + 0.011836551129817963, + 0.022897865623235703, + -0.05528765916824341, + -0.025355935096740723, + -0.017442841082811356, + 0.017111947759985924, + -0.16896392405033112, + 0.029856093227863312, + 0.016194898635149002, + 0.016866140067577362, + -0.04326202720403671, + -0.013509929180145264, + -0.05611962080001831, + 0.006154628470540047, + 0.014616061002016068, + -0.021290665492415428, + 0.0066131530329585075, + -0.04564446210861206, + 0.06096012890338898, + -0.03384573012590408, + -0.027416931465268135, + 0.010229351930320263, + 0.006901503540575504, + 0.057027217000722885, + -0.05388845130801201, + 0.06186772510409355, + 0.009699921123683453, + -0.01977800764143467, + -0.0142473503947258, + -0.0002583632303867489, + -0.017244303598999977, + -0.01770755648612976, + 0.019796915352344513, + 0.03055569715797901, + -0.030158624053001404, + 0.05131693184375763, + 0.0026967860758304596, + 0.02441052347421646, + 0.019182397052645683, + 0.041749369353055954, + 0.03246542811393738, + 0.0048452336341142654, + -0.0012266712728887796, + -0.03808117285370827, + -0.0015800188994035125, + -0.01989145576953888, + -0.021044857800006866, + 0.02153647318482399, + -0.01650688424706459, + -0.01559928897768259, + 0.004192899446934462, + 0.09136456251144409, + -0.028929589316248894, + -0.003526384709402919, + -0.02592318132519722, + 0.03384573012590408, + 0.015240032225847244, + 0.03781645745038986, + 0.009520293213427067, + 0.011524565517902374, + -0.017631923779845238, + -0.005559019278734922, + 0.01158128958195448, + -0.019031131640076637, + -0.024032358080148697, + 0.014776780270040035, + 0.014086630195379257, + -0.019314754754304886, + -0.03161455690860748, + -0.0729101300239563, + -0.017074130475521088, + 0.006750237662345171, + -0.07007389515638351, + -0.007246578577905893, + -0.025091219693422318, + -0.013349208980798721, + -0.025885364040732384, + 0.029950633645057678, + 0.026887500658631325, + -0.0005920638795942068, + -0.012838686816394329, + 0.048064716160297394, + 0.03757065162062645, + 0.01751847378909588, + -0.013821914792060852, + -0.026244621723890305, + -0.027019858360290527, + -0.011808188632130623, + -0.0006871959194540977, + -0.03278686851263046, + 0.035377293825149536, + 0.013056132011115551, + -0.025960998609662056, + -0.012469977140426636, + -0.0031482200138270855, + -0.0007728738128207624, + -0.08077595382928848, + 0.015088766813278198, + 0.056497786194086075, + 0.01623271405696869, + -0.009293394163250923, + 0.06659477949142456, + -0.01989145576953888, + -0.008598516695201397, + -0.015192762017250061, + -0.0026802413631230593, + 0.00839525368064642, + -0.001349574769847095, + 0.02866487391293049, + 0.005842642392963171, + -0.025942089036107063, + 0.0029638647101819515, + -0.024240348488092422, + -0.03524493798613548, + 0.021876821294426918, + -0.010720965452492237, + -0.005984454415738583, + -0.025771915912628174, + -0.04863196238875389, + 0.008239260874688625, + -0.06856123358011246, + -0.03305158391594887, + -0.032162897288799286, + 0.03318393975496292, + -0.0441318042576313, + -0.02000490576028824, + 0.03225743770599365, + 0.03131202608346939, + 0.048896677792072296, + -0.03499913215637207, + 0.024297073483467102, + 0.000977909890934825, + 0.030101899057626724, + 0.015882911160588264, + 0.018000634387135506, + -0.006197171751409769, + -0.006892049219459295, + -0.00835270993411541, + 0.02195245400071144, + 0.039593830704689026, + -0.010238805785775185, + 0.00586627796292305, + 0.0156465582549572, + 0.005800099112093449, + -0.015230578370392323, + -0.031690191477537155, + 0.006849505938589573, + -0.028986314311623573, + 0.02253860794007778, + -0.01421898789703846, + -0.014795688912272453, + -0.019872548058629036, + 0.018331527709960938, + -0.01570328325033188, + -0.019116219133138657, + -0.011864913627505302, + 0.002848051954060793, + 0.00756329158321023, + 0.032919224351644516, + 0.06387199461460114, + -0.01949438452720642, + -0.013112856075167656, + 0.004374891519546509, + 0.022311709821224213, + 0.01765083149075508, + 0.02304913103580475, + -0.034564241766929626, + -0.00717094587162137, + -0.05332120507955551, + -0.05154383182525635, + -0.003493295283988118, + 0.022444067522883415, + 0.037589557468891144, + -0.0087970532476902, + -0.06220807135105133, + 0.004157446790486574, + 0.009151582606136799, + 0.06288877129554749, + -0.030820412561297417, + 0.011902729980647564, + 0.003717830404639244, + -0.056157439947128296, + 0.024883229285478592, + 0.0030016812961548567, + 0.02505340240895748, + -0.026717327535152435, + 0.025185760110616684, + -0.01331139262765646, + 0.005445569753646851, + 0.02475087158381939, + 0.024202533066272736, + 0.036833230406045914, + -0.002345802029594779, + 0.023427294567227364, + -0.011184217408299446, + -0.01375573594123125, + 0.02312476374208927, + -0.029515745118260384, + -0.05910712108016014, + -0.017631923779845238, + 0.01743338629603386, + -0.024486156180500984, + -0.0026613331865519285, + 0.01509822066873312, + -0.011042405851185322, + 0.03392136096954346, + -0.036247074604034424, + -0.04787563532590866, + 0.030064083635807037, + -0.022670965641736984, + 0.005308485124260187, + 0.014786235056817532, + -0.027284573763608932, + 0.009671558625996113, + -0.0360390841960907, + 0.02153647318482399, + -0.05188417807221413, + -0.008995589800179005, + 0.011051859706640244, + -0.0006470159278251231, + -0.020269621163606644, + 0.02229280211031437, + 0.011108584702014923, + -0.013736828230321407, + -0.009841732680797577, + 0.034148260951042175, + -0.0014807506231591105, + -0.0018388251774013042, + 0.013500475324690342, + -0.026849685236811638, + -0.060014717280864716, + -0.008404707536101341, + 0.011212579905986786, + 0.02047761157155037, + -0.0016426523216068745, + 0.0483294315636158, + 0.022444067522883415, + 0.036719780415296555, + -0.03861060366034508, + 0.037362661212682724, + -0.006286986172199249, + -0.02773837186396122, + 0.033448655158281326, + 0.019985998049378395, + -0.03745720162987709, + -0.005204489920288324, + 0.016487974673509598, + -0.019314754754304886, + -0.0555901937186718, + -0.008636333048343658 + ], + "key": "Persistence Demonstration", + "status": "activated" + } + }, + { + "id": "6c0c977f-2e75-4e3a-a661-88674da5b14e", + "memory": "On January 13, 2026, at 8:51 PM, the user created a test memory for a demonstration purpose.", + "metadata": { + "sources": [ + "\"{\\\"type\\\": \\\"chat\\\", \\\"role\\\": \\\"user\\\", \\\"chat_time\\\": \\\"08:51 PM on 13 January, 2026\\\", \\\"content\\\": \\\"This is a test memory for dump example\\\", \\\"lang\\\": \\\"en\\\"}\"", + "\"{\\\"type\\\": \\\"chat\\\", \\\"role\\\": \\\"user\\\", \\\"chat_time\\\": \\\"08:51 PM on 13 January, 2026\\\", \\\"content\\\": \\\"Another memory to demonstrate persistence\\\", \\\"lang\\\": \\\"en\\\"}\"" + ], + "confidence": 0.99, + "usage": [], + "created_at": "2026-01-13T20:51:42.359643000+00:00", + "session_id": "default_session", + "memory_type": "UserMemory", + "type": "fact", + "tags": [ + "test", + "memory", + "demonstration" + ], + "updated_at": "2026-01-13T20:51:42.396236000+00:00", + "user_id": "example_user", + "background": "On January 13, 2026, at 8:51 PM, the user engaged in a memory demonstration by creating a test memory example and another memory to illustrate the concept of persistence. These actions reflect the user's interest in exploring and demonstrating memory-related concepts.", + "file_ids": [], + "embedding": [ + -0.023225771263241768, + 0.009711744263768196, + 0.01401975005865097, + -0.023113388568162918, + -0.044953107833862305, + -0.025454696267843246, + 0.004780950024724007, + 0.012184164486825466, + 0.003397237276658416, + 0.006766378879547119, + 0.033059265464544296, + -0.0018625102238729596, + -0.00135678774677217, + -0.0159396231174469, + -0.0017864176770672202, + 0.03302180394530296, + 0.05941302329301834, + -0.013467201963067055, + -0.011341294273734093, + -0.03324656933546066, + 0.01599581353366375, + -0.05263259634375572, + 0.07222465425729752, + 0.022851161658763885, + -0.007220592815428972, + 0.0017969536129385233, + -0.012418295256793499, + 0.013214340433478355, + 0.08645980805158615, + 0.0018215373856946826, + 0.02071589045226574, + -0.028620144352316856, + -0.015490091405808926, + -0.025454696267843246, + -0.04469088092446327, + -0.02105303853750229, + 0.027945848181843758, + 3.424162423470989e-05, + -0.028526492416858673, + -0.019479678943753242, + -0.008882921189069748, + 0.012596234679222107, + 0.015480726025998592, + 0.006949000991880894, + 0.015499456785619259, + 0.014825159683823586, + -0.012549408711493015, + -0.02382514625787735, + -0.025342313572764397, + -0.013364183716475964, + 0.002884491113945842, + 0.03133606165647507, + 0.04525279253721237, + -0.037179965525865555, + 0.027009325101971626, + 0.021689873188734055, + -0.05034748092293739, + -0.043342288583517075, + -0.019816827028989792, + 0.007492184638977051, + -0.03575645014643669, + 0.007862110622227192, + 0.0021774161141365767, + 0.01226845197379589, + -0.013504662550985813, + 0.03937143087387085, + 0.0068834442645311356, + 0.0019292375072836876, + -0.00272059952840209, + -0.023169580847024918, + 0.0013368866639211774, + 0.04581470787525177, + -0.05514247715473175, + -0.01210924331098795, + -0.0809905156493187, + 0.01955460198223591, + 0.0017419328214600682, + -0.015677396208047867, + -0.009426104836165905, + 0.021408917382359505, + 0.03892189636826515, + -0.022476553916931152, + 0.01390736736357212, + -0.0002572511730249971, + 0.047350604087114334, + 0.01260560005903244, + -0.06349626183509827, + 0.057502515614032745, + 0.040345411747694016, + -0.04274291172623634, + 0.028788719326257706, + 0.0003658293280750513, + 0.010807476006448269, + -0.02504262700676918, + 0.004296299535781145, + 0.010404771193861961, + -0.05158368870615959, + 0.003305926453322172, + 0.029069675132632256, + 0.011032241396605968, + 0.05813935026526451, + 0.022532744333148003, + 0.02037874236702919, + -0.017878225073218346, + -0.0010945613030344248, + -0.028788719326257706, + 0.02927570976316929, + 0.014375628903508186, + -0.009627456776797771, + 0.0005191849777474999, + 0.001597942435182631, + 0.04281783476471901, + 0.04918619245290756, + 0.01210924331098795, + 0.04203115403652191, + 0.013766888529062271, + 0.013485931791365147, + -0.01013317983597517, + -0.05083446949720383, + -0.014778333716094494, + 0.017484884709119797, + 0.014703411608934402, + 0.038341253995895386, + -0.03693646937608719, + -0.06076161563396454, + -0.014787699095904827, + -0.01294274814426899, + -0.023019736632704735, + -0.0010770015651360154, + -0.004511700011789799, + 0.016042640432715416, + -0.000730487983673811, + 0.00883141253143549, + -0.008058780804276466, + -0.026971863582730293, + -0.021614952012896538, + 0.011247641406953335, + 0.05281990021467209, + -0.010498423129320145, + -0.03399578854441643, + 0.020341280847787857, + -0.003252076217904687, + -0.008681568317115307, + -0.026859480887651443, + 0.015930257737636566, + -0.02498643472790718, + 0.005469294730573893, + -0.003645415883511305, + 0.02721535973250866, + 0.001160117913968861, + -0.01229654811322689, + 0.019891750067472458, + -0.02288862317800522, + -0.04469088092446327, + -0.022851161658763885, + -0.003551763715222478, + 0.01916126161813736, + 0.06192290410399437, + -0.01468468178063631, + 0.012624330818653107, + -0.0015429217601194978, + -0.048511892557144165, + -0.06076161563396454, + 0.0465264655649662, + 0.02032255008816719, + 0.02860141359269619, + 0.013457836583256721, + 0.025248661637306213, + -0.0460769347846508, + 0.03639328479766846, + 0.002751036547124386, + -0.008306959643959999, + -0.008747125044465065, + -0.008157115429639816, + -0.008611329831182957, + 0.033845942467451096, + 0.038734592497348785, + -0.000227984826778993, + -0.01032984908670187, + -0.0321040116250515, + -0.05158368870615959, + 0.0031490586698055267, + 0.030287155881524086, + 0.013794984668493271, + -0.04135685786604881, + -0.003430015640333295, + -0.012390200048685074, + 0.004427412524819374, + 0.015049925073981285, + 0.01107906736433506, + 0.008161798119544983, + -0.03180432319641113, + -0.000778484798502177, + 0.00315842404961586, + -0.005722155794501305, + -0.027908386662602425, + -0.0014211736852303147, + 0.0008768197149038315, + -0.02448071353137493, + 0.005408420693129301, + -0.011060337536036968, + 0.035737719386816025, + 0.015668030828237534, + 0.02127780392765999, + -0.01746615581214428, + -0.013823079876601696, + -0.01416022889316082, + 0.0026854798197746277, + 0.0031677891965955496, + -0.0812152773141861, + -0.024836590513586998, + 0.0003201738290954381, + -0.016014544293284416, + -0.011032241396605968, + 0.02238290011882782, + -0.013261166401207447, + -0.0022523379884660244, + -0.005698742810636759, + 0.012764809653162956, + -0.015499456785619259, + -0.05371896177530289, + 0.04941095784306526, + -0.017447425052523613, + 0.030343346297740936, + 0.030811607837677002, + -0.01826220005750656, + 0.008536407724022865, + 0.05016017332673073, + 0.02848903089761734, + -0.0326659232378006, + 0.026110263541340828, + 0.018721096217632294, + 0.012015590444207191, + -0.007585836574435234, + 0.03513834625482559, + -0.020847003906965256, + -0.04154416173696518, + 0.0050150807946920395, + 0.02605407126247883, + 0.0016857414739206433, + -0.01627677120268345, + 0.004181575495749712, + 0.01961079239845276, + 0.018365217372775078, + -0.04154416173696518, + -0.026241375133395195, + 0.0030554065015167, + 0.01035794522613287, + 0.02227051742374897, + 0.01488135103136301, + 0.01074191927909851, + 0.007899572141468525, + -0.05918825790286064, + -0.001568676088936627, + 0.01715710200369358, + 0.009051495231688023, + 0.029032215476036072, + -0.006073351949453354, + 0.03877205401659012, + -0.0013439105823636055, + 0.04064510017633438, + -0.03774188086390495, + 0.025810575112700462, + 0.03341514244675636, + 0.022944815456867218, + -0.019854288548231125, + -0.019535871222615242, + -0.013411009684205055, + 0.014591028913855553, + -0.008742442354559898, + 0.02993127703666687, + -0.05525486171245575, + -0.02365657314658165, + 0.03487611934542656, + -0.015705492347478867, + 0.008349102921783924, + -0.02970651164650917, + 5.427992618933786e-06, + -0.020584776997566223, + 0.038846977055072784, + 0.016632650047540665, + -0.026409950107336044, + 0.031298600137233734, + -0.0122216260060668, + 0.0071503533981740475, + -0.017335042357444763, + -0.027796003967523575, + -0.027140438556671143, + 0.03952127322554588, + -0.019704444333910942, + -0.019460948184132576, + 0.01024556253105402, + 0.06357118487358093, + -0.04360451176762581, + -0.018206007778644562, + -0.021952100098133087, + 0.00883141253143549, + -0.15463869273662567, + -0.008232037536799908, + -0.013879271224141121, + -0.0018063188763335347, + -0.014413089491426945, + 0.016632650047540665, + -0.04682615399360657, + -0.0486617386341095, + -0.010479693301022053, + 0.08968144655227661, + -0.005267942324280739, + -0.04244322329759598, + 0.017306946218013763, + 0.026353757828474045, + -0.025866767391562462, + -0.005651916842907667, + -0.03865967318415642, + -0.027908386662602425, + -0.024312138557434082, + -0.011931303888559341, + 0.010030161589384079, + -0.02148383855819702, + 0.05424341559410095, + -0.01921745389699936, + -0.002258191118016839, + -0.00631684809923172, + 0.04941095784306526, + 0.009739839471876621, + -0.020734621211886406, + 0.002176245441660285, + -0.014169593341648579, + 0.04128193482756615, + -0.006579074542969465, + -0.005895412527024746, + -0.008999986574053764, + 0.05971270799636841, + -0.0005150876822881401, + 0.007473453879356384, + 0.011453676968812943, + -0.020416202023625374, + 0.015396438539028168, + 0.066230908036232, + -0.03858475014567375, + 0.045964550226926804, + 0.03455770015716553, + 0.0163891538977623, + -0.004703687038272619, + -0.012184164486825466, + -0.0476502925157547, + -0.011369389481842518, + -0.007332975510507822, + -0.020022863522171974, + -0.015096751973032951, + -0.036505669355392456, + -0.020528584718704224, + -0.013401645235717297, + -0.054955173283815384, + 0.017035353928804398, + -0.08526106178760529, + 0.021614952012896538, + -0.0310363732278347, + -0.05192083865404129, + -0.010760650038719177, + 0.040345411747694016, + -0.02970651164650917, + 0.00763734569773078, + 0.04030795395374298, + -0.06394579261541367, + -0.0025051990523934364, + -0.03637455403804779, + 0.04847443476319313, + -0.018449503928422928, + 0.012736713513731956, + 0.001112121157348156, + 0.021352725103497505, + 0.06166067719459534, + 0.002046302892267704, + -0.03247861936688423, + 0.011013510636985302, + -0.08885730803012848, + -0.027796003967523575, + 0.053681500256061554, + -0.015452630817890167, + -0.03388340398669243, + -0.01715710200369358, + 0.00902808178216219, + -0.01121954619884491, + 0.008096241392195225, + 0.04113209247589111, + 0.20933163166046143, + -0.0004925525863654912, + -0.0620727464556694, + -0.013055131770670414, + 0.03787299245595932, + -0.010076988488435745, + -0.053606580942869186, + 0.01327053178101778, + -0.04233084246516228, + -0.04536517709493637, + 0.020285088568925858, + -0.0007509744027629495, + 0.01352339331060648, + 0.010208101011812687, + 0.0115192336961627, + -0.0243870597332716, + -0.01263369619846344, + -0.007501549553126097, + 0.08908206969499588, + 0.016585823148489, + 0.035793911665678024, + -0.014338168315589428, + -0.001372006256133318, + -0.001101585221476853, + -0.04525279253721237, + -0.02648487128317356, + -0.00762329762801528, + 0.07971683889627457, + -0.012708617374300957, + 0.056565992534160614, + -0.018917765468358994, + 0.011697173118591309, + 0.007997906766831875, + 0.004799680784344673, + 0.023469267413020134, + 0.009664918296039104, + 0.007894889451563358, + -0.04281783476471901, + -0.017569173127412796, + 0.02921951934695244, + 0.040794942528009415, + -0.023787684738636017, + 0.013841810636222363, + 0.04124447703361511, + -0.002535636071115732, + 0.01271798275411129, + 0.011978129856288433, + -0.02232670970261097, + -0.04135685786604881, + -0.002619923325255513, + -0.03480119630694389, + 0.016838684678077698, + 0.0310363732278347, + -0.016005178913474083, + 0.013682601973414421, + -0.006705504842102528, + -0.011097798123955727, + -0.02687821164727211, + 0.002102494239807129, + 0.03644947707653046, + 0.03888443857431412, + 0.0315420962870121, + -0.028751257807016373, + 0.051995761692523956, + -0.014937542378902435, + 0.0018285613041371107, + 0.0020779105834662914, + -0.05240783095359802, + 0.03425801172852516, + 0.026953132823109627, + 0.06267212331295013, + 0.010020796209573746, + 0.03259100392460823, + 0.0234318058937788, + 0.008007272146642208, + 0.012689887546002865, + 0.03985841944813728, + 0.05488025024533272, + 0.021858448162674904, + 0.07214973866939545, + 0.0159115269780159, + -0.020453663542866707, + -0.04027049243450165, + 0.008372516371309757, + 0.005876682233065367, + -0.05701552331447601, + -0.024143563583493233, + 0.052445292472839355, + -0.024518173187971115, + -0.03463262319564819, + -0.0329468809068203, + 0.045739784836769104, + -0.025510888546705246, + -0.027833465486764908, + 0.048174746334552765, + 0.0156586654484272, + 0.04124447703361511, + -0.0003596833848860115, + -0.029743971303105354, + 0.0014972662320360541, + -0.05083446949720383, + -0.02843284048140049, + 0.013663871213793755, + -0.0015663348603993654, + -0.010320483706891537, + -0.03315291553735733, + -0.028414109721779823, + -0.003565811552107334, + 0.008011954836547375, + 0.012764809653162956, + -0.0057455687783658504, + 0.03433293476700783, + -0.07361071556806564, + -0.04446611553430557, + -0.0036758529022336006, + 0.003003897611051798, + -0.014619125053286552, + 0.009084274061024189, + -0.009491661563515663, + 0.047425527125597, + 0.009936509653925896, + 0.00795576348900795, + 0.010620171204209328, + 0.05540470406413078, + 0.009861587546765804, + -0.022364171221852303, + -0.0022628738079220057, + -0.012006225995719433, + 0.004530430305749178, + -0.01372006256133318, + 0.026615984737873077, + -0.036187250167131424, + 0.04094478860497475, + -0.03877205401659012, + 0.02732774242758751, + 0.014862621203064919, + 0.021858448162674904, + 0.01667010970413685, + -0.01502182986587286, + 0.033621177077293396, + 0.0031748132314532995, + -0.09402691572904587, + -0.000989787746220827, + 0.027346473187208176, + 0.025585809722542763, + 0.01667947508394718, + 0.024181025102734566, + -0.02427467703819275, + -0.01249321736395359, + 0.02448071353137493, + 0.03933396935462952, + 0.01710091158747673, + 0.02449944242835045, + -0.05960032716393471, + -0.03335895016789436, + -0.039071742445230484, + -0.04547755792737007, + 0.017475519329309464, + 0.03644947707653046, + -0.04236830398440361, + -0.006752331275492907, + 0.031841784715652466, + 0.03397705778479576, + 0.08226418495178223, + 0.0621851310133934, + -0.04214353859424591, + -0.011397485621273518, + 0.030006198212504387, + 0.0031162805389612913, + -0.023356884717941284, + 0.006532248109579086, + -0.005928190890699625, + -0.016407884657382965, + 0.021184151992201805, + 0.0018777287332341075, + -0.047425527125597, + -0.04903634637594223, + 0.026840750128030777, + -0.021502569317817688, + -0.01077001541852951, + -0.008461485616862774, + -0.02521120011806488, + -0.037685688585042953, + -0.011650347150862217, + 0.02054731547832489, + 0.021090500056743622, + 0.01893649622797966, + -0.028357917442917824, + 0.012689887546002865, + -0.001998306019231677, + -0.08511121571063995, + 0.06967731565237045, + 0.030399538576602936, + -0.015471360646188259, + 0.04203115403652191, + -0.043791819363832474, + 0.02854522317647934, + 0.010301753878593445, + 0.011388120241463184, + 0.022233057767152786, + -0.012661791406571865, + 0.0016037956811487675, + -0.019685715436935425, + -0.01944221928715706, + -0.019292375072836876, + -0.012905287556350231, + 0.009908413514494896, + 0.0161643885076046, + -0.0068834442645311356, + 0.02015397697687149, + -0.018589982762932777, + -0.019966671243309975, + 0.0020123538561165333, + 0.014328802935779095, + -0.050722088664770126, + -0.014890716411173344, + 0.07177512347698212, + 0.044503577053546906, + -0.011921938508749008, + 0.010320483706891537, + -0.014478646218776703, + 0.01338291447609663, + 0.010526519268751144, + -0.01812172122299671, + -0.02142764814198017, + 0.01088239811360836, + 0.00478329136967659, + -0.05225798487663269, + 0.02693440392613411, + 0.022682588547468185, + -0.005764299537986517, + -0.02238290011882782, + -0.024087373167276382, + -0.019311105832457542, + 0.05296974256634712, + -0.0651070848107338, + 0.020809542387723923, + 0.015780413523316383, + -0.024780400097370148, + -0.0027042103465646505, + -0.02914459817111492, + 0.014207054860889912, + 0.020079053938388824, + 0.04768775403499603, + 0.0315982885658741, + 0.03526945784687996, + 0.013439105823636055, + 0.012886556796729565, + 0.007670123595744371, + 0.023974990472197533, + 0.017597267404198647, + 0.007866793312132359, + -0.048174746334552765, + -0.0489988848567009, + 0.040457796305418015, + 0.08016636967658997, + 0.013682601973414421, + -0.018046800047159195, + -0.06683028489351273, + -0.034351665526628494, + -0.006799157243221998, + 0.0028048865497112274, + -0.020341280847787857, + 0.01605200581252575, + -0.04637662321329117, + -0.06993953883647919, + 0.017569173127412796, + -0.06008731946349144, + -0.006916222628206015, + -0.03309672325849533, + -0.027758542448282242, + 0.004675591364502907, + -0.00883141253143549, + -0.03899681940674782, + 0.07012684643268585, + -0.0009002327569760382, + -0.02620391547679901, + -0.025023896247148514, + 0.06802903115749359, + 0.030193503946065903, + -0.00706606637686491, + 0.052332907915115356, + 0.02976270206272602, + 0.03502596169710159, + -0.012886556796729565, + -3.263197504566051e-05, + 0.019179992377758026, + -0.02637248858809471, + 0.017278850078582764, + -0.06031208485364914, + -0.012249721214175224, + 0.00553953368216753, + -0.029182057827711105, + 0.040907327085733414, + -0.010788745246827602, + 0.01755044236779213, + -0.042293380945920944, + 0.02043493278324604, + -0.0234318058937788, + 0.02427467703819275, + -0.030455728992819786, + 0.011584790423512459, + -0.045515019446611404, + -0.015181038528680801, + 0.0005320621421560645, + 0.00795576348900795, + -0.01710091158747673, + 0.017082180827856064, + 0.029687780886888504, + 0.003418309148401022, + 0.020734621211886406, + -0.07994160801172256, + 0.007215910125523806, + 0.022457823157310486, + -0.0006052280077710748, + -0.015817875042557716, + 0.02682201936841011, + -0.058176811784505844, + 0.05413103103637695, + -0.004389951936900616, + 0.03639328479766846, + 0.011172720231115818, + -0.0489988848567009, + -0.005675329826772213, + -0.031111296266317368, + -0.0019268961623311043, + -0.0032825132366269827, + -0.017981242388486862, + 0.036955200135707855, + -0.03377101942896843, + -0.030418269336223602, + 0.003273148089647293, + -0.02464928664267063, + -0.0011630445951595902, + 0.04394166171550751, + 0.020416202023625374, + 0.021034307777881622, + 0.0008352614822797477, + 0.05083446949720383, + -0.03319037705659866, + -0.009290308691561222, + -0.02464928664267063, + 0.04083240404725075, + -0.049785565584897995, + -0.031298600137233734, + 0.02560454048216343, + -0.01760663278400898, + 0.014263246208429337, + -0.01757853850722313, + -0.00598438223823905, + -0.005591042805463076, + 0.02594168856739998, + -0.012867826968431473, + -0.0477626770734787, + 0.033902134746313095, + -0.025061357766389847, + -0.05274497717618942, + 0.031523365527391434, + 0.03970857709646225, + 0.01596771739423275, + 0.013888636603951454, + 0.03877205401659012, + -0.009739839471876621, + -0.009786665439605713, + -0.04761283099651337, + -0.008143068291246891, + 0.032684653997421265, + 0.03822886943817139, + 0.009163877926766872, + -0.028170613572001457, + 0.005843903869390488, + 0.020753350108861923, + -0.007281466852873564, + -0.014216420240700245, + -0.014057210646569729, + -0.006424548104405403, + -0.011172720231115818, + -0.03332148864865303, + 0.03974603861570358, + 0.013757524080574512, + -0.015012464486062527, + 0.0028165930416435003, + 0.01327053178101778, + -0.019910480827093124, + 0.006625900510698557, + -0.022626396268606186, + -0.011912573128938675, + -0.02537977509200573, + -0.024574365466833115, + 0.007300197146832943, + -0.023169580847024918, + -0.047425527125597, + 0.0003184178494848311, + 0.02537977509200573, + -0.16213087737560272, + 0.02176479622721672, + 0.0014211736852303147, + 0.0010805134661495686, + -0.021952100098133087, + -0.02803950011730194, + -0.036280903965234756, + -0.023281963542103767, + 0.028245534747838974, + -0.02571692317724228, + 0.011734633706510067, + -0.03586883470416069, + 0.035793911665678024, + -0.030736686661839485, + -0.02288862317800522, + 0.02682201936841011, + 0.012006225995719433, + 0.04476580023765564, + -0.06004985794425011, + 0.09335261583328247, + -0.005075954832136631, + -0.030418269336223602, + -0.004537454340606928, + 0.005057224538177252, + -0.048624277114868164, + -0.0246118251234293, + 0.031017644330859184, + 0.02798330970108509, + -0.03510088473558426, + 0.03560660779476166, + -0.02120288275182247, + 0.031298600137233734, + 0.02449944242835045, + 0.05858888104557991, + 0.03335895016789436, + -0.01662328466773033, + 0.011725268326699734, + -0.04349213093519211, + -0.015340247191488743, + -0.00708011444658041, + -0.05031001940369606, + 0.01857125200331211, + -0.009627456776797771, + -0.015265326015651226, + 0.03616851940751076, + 0.06383340805768967, + -0.027833465486764908, + 0.009430787526071072, + -0.037685688585042953, + 0.02260766737163067, + -0.012202895246446133, + 0.021614952012896538, + -0.009009351953864098, + -0.009674282744526863, + -0.02116542123258114, + -0.017953146249055862, + -0.01185638178139925, + -0.013869906775653362, + -0.011453676968812943, + 0.016951067373156548, + -0.01976063661277294, + -0.012540043331682682, + -0.0244432520121336, + -0.042068615555763245, + -0.018955226987600327, + 0.011697173118591309, + -0.06926524639129639, + 0.0025239295791834593, + -0.015920892357826233, + 0.009945875033736229, + -0.035119615495204926, + 0.027009325101971626, + 0.02015397697687149, + -0.009777300991117954, + -0.015977082774043083, + 0.05308212712407112, + 0.06327150017023087, + 0.021558761596679688, + 0.017091546207666397, + -0.01766282506287098, + -0.013336088508367538, + 0.005932873580604792, + -0.0006385916494764388, + -0.015649300068616867, + 0.015049925073981285, + 0.03538184240460396, + -0.005164924543350935, + 0.003172471886500716, + -0.03903428092598915, + -0.01271798275411129, + -0.09005605429410934, + -0.008620694279670715, + 0.045664865523576736, + 0.003343387274071574, + -0.0031935435254126787, + 0.08196450024843216, + -0.023019736632704735, + -0.03611232712864876, + -0.037123773247003555, + -0.014216420240700245, + -0.009983335621654987, + 0.02026635967195034, + 0.00901871733367443, + 0.027964578941464424, + 0.009561900049448013, + 0.0246118251234293, + -0.01416022889316082, + -0.039558734744787216, + 0.006003112997859716, + 0.024087373167276382, + -0.036131057888269424, + -0.04008318483829498, + -0.04064510017633438, + -0.01315814908593893, + -0.08001653105020523, + -0.01819664239883423, + -0.020228898152709007, + 0.014759602956473827, + -0.05536724254488945, + -0.021577490493655205, + 0.035962484776973724, + 0.008822047151625156, + 0.01771901547908783, + -0.027908386662602425, + 0.019854288548231125, + 0.01024556253105402, + 0.029800163581967354, + 0.002376427175477147, + -0.007960446178913116, + 0.005197702907025814, + -0.019236182793974876, + 0.000641518272459507, + 0.04941095784306526, + 0.059113334864377975, + -0.0018589982064440846, + -0.006073351949453354, + -0.012024955824017525, + 0.013420375064015388, + -0.05192083865404129, + -0.03547549247741699, + -0.011425580829381943, + -0.008737759664654732, + 0.024836590513586998, + -0.014075941406190395, + -0.008587916381657124, + 0.013195609673857689, + 0.032628461718559265, + 0.003600931027904153, + -0.021633682772517204, + -0.0024490077048540115, + 0.018814748153090477, + -0.010835572145879269, + 0.036730434745550156, + 0.056116461753845215, + -0.047238223254680634, + 0.013729427941143513, + -0.006794474553316832, + 0.02987508475780487, + 0.008246085606515408, + 0.032347507774829865, + -0.04191877320408821, + 0.03465135395526886, + -0.061061304062604904, + -0.030792877078056335, + 0.013111323118209839, + 0.03854728862643242, + 0.03127986937761307, + 0.006546296179294586, + -0.062447357922792435, + 0.021970830857753754, + 0.021408917382359505, + 0.04547755792737007, + -0.034070707857608795, + -0.003069454338401556, + 0.006246608681976795, + -0.03761076554656029, + 0.017119640484452248, + 0.03736726939678192, + 0.018608713522553444, + -0.030774148181080818, + 0.021408917382359505, + -0.015321517363190651, + 0.019704444333910942, + 0.009730474092066288, + 0.02082827314734459, + 0.017110276967287064, + -0.012146703898906708, + 0.018589982762932777, + -0.006349626462906599, + 0.01249321736395359, + 0.025248661637306213, + -0.02309465780854225, + -0.053681500256061554, + -0.03335895016789436, + 0.04158162325620651, + -0.02916332706809044, + -0.02815188281238079, + 0.026110263541340828, + -0.014825159683823586, + 0.06844110786914825, + -0.04536517709493637, + -0.03854728862643242, + 0.013261166401207447, + -0.049785565584897995, + 0.008058780804276466, + 0.020453663542866707, + -0.010535884648561478, + 0.019517140462994576, + -0.04693853482604027, + 0.016342326998710632, + -0.020640967413783073, + -0.01554628275334835, + 0.017896955832839012, + 0.012184164486825466, + -0.02148383855819702, + 0.01065763272345066, + 0.0030835021752864122, + 0.019348565489053726, + -0.02071589045226574, + 0.05364403873682022, + -0.017054084688425064, + -0.004036414436995983, + 0.026335028931498528, + -0.03764822706580162, + -0.0230384673923254, + 0.00017208610370289534, + -0.0007825820939615369, + 0.0464140810072422, + 0.02110922895371914, + 0.048849042505025864, + 0.021633682772517204, + 0.024349600076675415, + -0.02910713665187359, + 0.022982275113463402, + -0.01966698467731476, + -0.011921938508749008, + 0.027065515518188477, + 0.017232025042176247, + -0.03560660779476166, + -0.01119145005941391, + 0.0029125867877155542, + -0.019123800098896027, + -0.04903634637594223, + -0.02921951934695244 + ], + "key": "Test Memory Example", + "status": "activated" + } + }, + { + "id": "3178a22f-4edb-46e1-89fd-a08009b131a1", + "memory": "On January 13, 2026, at 8:51 PM, the user created a test memory for a demonstration purpose.", + "metadata": { + "sources": [ + "\"{\\\"type\\\": \\\"chat\\\", \\\"role\\\": \\\"user\\\", \\\"chat_time\\\": \\\"08:51 PM on 13 January, 2026\\\", \\\"content\\\": \\\"This is a test memory for dump example\\\", \\\"lang\\\": \\\"en\\\"}\"", + "\"{\\\"type\\\": \\\"chat\\\", \\\"role\\\": \\\"user\\\", \\\"chat_time\\\": \\\"08:51 PM on 13 January, 2026\\\", \\\"content\\\": \\\"Another memory to demonstrate persistence\\\", \\\"lang\\\": \\\"en\\\"}\"" + ], + "confidence": 0.99, + "usage": [], + "created_at": "2026-01-13T20:51:42.359643000+00:00", + "session_id": "default_session", + "memory_type": "WorkingMemory", + "type": "fact", + "tags": [ + "test", + "memory", + "demonstration" + ], + "updated_at": "2026-01-13T20:51:42.396195000+00:00", + "user_id": "example_user", + "background": "On January 13, 2026, at 8:51 PM, the user engaged in a memory demonstration by creating a test memory example and another memory to illustrate the concept of persistence. These actions reflect the user's interest in exploring and demonstrating memory-related concepts.", + "file_ids": [], + "embedding": [ + -0.023225771263241768, + 0.009711744263768196, + 0.01401975005865097, + -0.023113388568162918, + -0.044953107833862305, + -0.025454696267843246, + 0.004780950024724007, + 0.012184164486825466, + 0.003397237276658416, + 0.006766378879547119, + 0.033059265464544296, + -0.0018625102238729596, + -0.00135678774677217, + -0.0159396231174469, + -0.0017864176770672202, + 0.03302180394530296, + 0.05941302329301834, + -0.013467201963067055, + -0.011341294273734093, + -0.03324656933546066, + 0.01599581353366375, + -0.05263259634375572, + 0.07222465425729752, + 0.022851161658763885, + -0.007220592815428972, + 0.0017969536129385233, + -0.012418295256793499, + 0.013214340433478355, + 0.08645980805158615, + 0.0018215373856946826, + 0.02071589045226574, + -0.028620144352316856, + -0.015490091405808926, + -0.025454696267843246, + -0.04469088092446327, + -0.02105303853750229, + 0.027945848181843758, + 3.424162423470989e-05, + -0.028526492416858673, + -0.019479678943753242, + -0.008882921189069748, + 0.012596234679222107, + 0.015480726025998592, + 0.006949000991880894, + 0.015499456785619259, + 0.014825159683823586, + -0.012549408711493015, + -0.02382514625787735, + -0.025342313572764397, + -0.013364183716475964, + 0.002884491113945842, + 0.03133606165647507, + 0.04525279253721237, + -0.037179965525865555, + 0.027009325101971626, + 0.021689873188734055, + -0.05034748092293739, + -0.043342288583517075, + -0.019816827028989792, + 0.007492184638977051, + -0.03575645014643669, + 0.007862110622227192, + 0.0021774161141365767, + 0.01226845197379589, + -0.013504662550985813, + 0.03937143087387085, + 0.0068834442645311356, + 0.0019292375072836876, + -0.00272059952840209, + -0.023169580847024918, + 0.0013368866639211774, + 0.04581470787525177, + -0.05514247715473175, + -0.01210924331098795, + -0.0809905156493187, + 0.01955460198223591, + 0.0017419328214600682, + -0.015677396208047867, + -0.009426104836165905, + 0.021408917382359505, + 0.03892189636826515, + -0.022476553916931152, + 0.01390736736357212, + -0.0002572511730249971, + 0.047350604087114334, + 0.01260560005903244, + -0.06349626183509827, + 0.057502515614032745, + 0.040345411747694016, + -0.04274291172623634, + 0.028788719326257706, + 0.0003658293280750513, + 0.010807476006448269, + -0.02504262700676918, + 0.004296299535781145, + 0.010404771193861961, + -0.05158368870615959, + 0.003305926453322172, + 0.029069675132632256, + 0.011032241396605968, + 0.05813935026526451, + 0.022532744333148003, + 0.02037874236702919, + -0.017878225073218346, + -0.0010945613030344248, + -0.028788719326257706, + 0.02927570976316929, + 0.014375628903508186, + -0.009627456776797771, + 0.0005191849777474999, + 0.001597942435182631, + 0.04281783476471901, + 0.04918619245290756, + 0.01210924331098795, + 0.04203115403652191, + 0.013766888529062271, + 0.013485931791365147, + -0.01013317983597517, + -0.05083446949720383, + -0.014778333716094494, + 0.017484884709119797, + 0.014703411608934402, + 0.038341253995895386, + -0.03693646937608719, + -0.06076161563396454, + -0.014787699095904827, + -0.01294274814426899, + -0.023019736632704735, + -0.0010770015651360154, + -0.004511700011789799, + 0.016042640432715416, + -0.000730487983673811, + 0.00883141253143549, + -0.008058780804276466, + -0.026971863582730293, + -0.021614952012896538, + 0.011247641406953335, + 0.05281990021467209, + -0.010498423129320145, + -0.03399578854441643, + 0.020341280847787857, + -0.003252076217904687, + -0.008681568317115307, + -0.026859480887651443, + 0.015930257737636566, + -0.02498643472790718, + 0.005469294730573893, + -0.003645415883511305, + 0.02721535973250866, + 0.001160117913968861, + -0.01229654811322689, + 0.019891750067472458, + -0.02288862317800522, + -0.04469088092446327, + -0.022851161658763885, + -0.003551763715222478, + 0.01916126161813736, + 0.06192290410399437, + -0.01468468178063631, + 0.012624330818653107, + -0.0015429217601194978, + -0.048511892557144165, + -0.06076161563396454, + 0.0465264655649662, + 0.02032255008816719, + 0.02860141359269619, + 0.013457836583256721, + 0.025248661637306213, + -0.0460769347846508, + 0.03639328479766846, + 0.002751036547124386, + -0.008306959643959999, + -0.008747125044465065, + -0.008157115429639816, + -0.008611329831182957, + 0.033845942467451096, + 0.038734592497348785, + -0.000227984826778993, + -0.01032984908670187, + -0.0321040116250515, + -0.05158368870615959, + 0.0031490586698055267, + 0.030287155881524086, + 0.013794984668493271, + -0.04135685786604881, + -0.003430015640333295, + -0.012390200048685074, + 0.004427412524819374, + 0.015049925073981285, + 0.01107906736433506, + 0.008161798119544983, + -0.03180432319641113, + -0.000778484798502177, + 0.00315842404961586, + -0.005722155794501305, + -0.027908386662602425, + -0.0014211736852303147, + 0.0008768197149038315, + -0.02448071353137493, + 0.005408420693129301, + -0.011060337536036968, + 0.035737719386816025, + 0.015668030828237534, + 0.02127780392765999, + -0.01746615581214428, + -0.013823079876601696, + -0.01416022889316082, + 0.0026854798197746277, + 0.0031677891965955496, + -0.0812152773141861, + -0.024836590513586998, + 0.0003201738290954381, + -0.016014544293284416, + -0.011032241396605968, + 0.02238290011882782, + -0.013261166401207447, + -0.0022523379884660244, + -0.005698742810636759, + 0.012764809653162956, + -0.015499456785619259, + -0.05371896177530289, + 0.04941095784306526, + -0.017447425052523613, + 0.030343346297740936, + 0.030811607837677002, + -0.01826220005750656, + 0.008536407724022865, + 0.05016017332673073, + 0.02848903089761734, + -0.0326659232378006, + 0.026110263541340828, + 0.018721096217632294, + 0.012015590444207191, + -0.007585836574435234, + 0.03513834625482559, + -0.020847003906965256, + -0.04154416173696518, + 0.0050150807946920395, + 0.02605407126247883, + 0.0016857414739206433, + -0.01627677120268345, + 0.004181575495749712, + 0.01961079239845276, + 0.018365217372775078, + -0.04154416173696518, + -0.026241375133395195, + 0.0030554065015167, + 0.01035794522613287, + 0.02227051742374897, + 0.01488135103136301, + 0.01074191927909851, + 0.007899572141468525, + -0.05918825790286064, + -0.001568676088936627, + 0.01715710200369358, + 0.009051495231688023, + 0.029032215476036072, + -0.006073351949453354, + 0.03877205401659012, + -0.0013439105823636055, + 0.04064510017633438, + -0.03774188086390495, + 0.025810575112700462, + 0.03341514244675636, + 0.022944815456867218, + -0.019854288548231125, + -0.019535871222615242, + -0.013411009684205055, + 0.014591028913855553, + -0.008742442354559898, + 0.02993127703666687, + -0.05525486171245575, + -0.02365657314658165, + 0.03487611934542656, + -0.015705492347478867, + 0.008349102921783924, + -0.02970651164650917, + 5.427992618933786e-06, + -0.020584776997566223, + 0.038846977055072784, + 0.016632650047540665, + -0.026409950107336044, + 0.031298600137233734, + -0.0122216260060668, + 0.0071503533981740475, + -0.017335042357444763, + -0.027796003967523575, + -0.027140438556671143, + 0.03952127322554588, + -0.019704444333910942, + -0.019460948184132576, + 0.01024556253105402, + 0.06357118487358093, + -0.04360451176762581, + -0.018206007778644562, + -0.021952100098133087, + 0.00883141253143549, + -0.15463869273662567, + -0.008232037536799908, + -0.013879271224141121, + -0.0018063188763335347, + -0.014413089491426945, + 0.016632650047540665, + -0.04682615399360657, + -0.0486617386341095, + -0.010479693301022053, + 0.08968144655227661, + -0.005267942324280739, + -0.04244322329759598, + 0.017306946218013763, + 0.026353757828474045, + -0.025866767391562462, + -0.005651916842907667, + -0.03865967318415642, + -0.027908386662602425, + -0.024312138557434082, + -0.011931303888559341, + 0.010030161589384079, + -0.02148383855819702, + 0.05424341559410095, + -0.01921745389699936, + -0.002258191118016839, + -0.00631684809923172, + 0.04941095784306526, + 0.009739839471876621, + -0.020734621211886406, + 0.002176245441660285, + -0.014169593341648579, + 0.04128193482756615, + -0.006579074542969465, + -0.005895412527024746, + -0.008999986574053764, + 0.05971270799636841, + -0.0005150876822881401, + 0.007473453879356384, + 0.011453676968812943, + -0.020416202023625374, + 0.015396438539028168, + 0.066230908036232, + -0.03858475014567375, + 0.045964550226926804, + 0.03455770015716553, + 0.0163891538977623, + -0.004703687038272619, + -0.012184164486825466, + -0.0476502925157547, + -0.011369389481842518, + -0.007332975510507822, + -0.020022863522171974, + -0.015096751973032951, + -0.036505669355392456, + -0.020528584718704224, + -0.013401645235717297, + -0.054955173283815384, + 0.017035353928804398, + -0.08526106178760529, + 0.021614952012896538, + -0.0310363732278347, + -0.05192083865404129, + -0.010760650038719177, + 0.040345411747694016, + -0.02970651164650917, + 0.00763734569773078, + 0.04030795395374298, + -0.06394579261541367, + -0.0025051990523934364, + -0.03637455403804779, + 0.04847443476319313, + -0.018449503928422928, + 0.012736713513731956, + 0.001112121157348156, + 0.021352725103497505, + 0.06166067719459534, + 0.002046302892267704, + -0.03247861936688423, + 0.011013510636985302, + -0.08885730803012848, + -0.027796003967523575, + 0.053681500256061554, + -0.015452630817890167, + -0.03388340398669243, + -0.01715710200369358, + 0.00902808178216219, + -0.01121954619884491, + 0.008096241392195225, + 0.04113209247589111, + 0.20933163166046143, + -0.0004925525863654912, + -0.0620727464556694, + -0.013055131770670414, + 0.03787299245595932, + -0.010076988488435745, + -0.053606580942869186, + 0.01327053178101778, + -0.04233084246516228, + -0.04536517709493637, + 0.020285088568925858, + -0.0007509744027629495, + 0.01352339331060648, + 0.010208101011812687, + 0.0115192336961627, + -0.0243870597332716, + -0.01263369619846344, + -0.007501549553126097, + 0.08908206969499588, + 0.016585823148489, + 0.035793911665678024, + -0.014338168315589428, + -0.001372006256133318, + -0.001101585221476853, + -0.04525279253721237, + -0.02648487128317356, + -0.00762329762801528, + 0.07971683889627457, + -0.012708617374300957, + 0.056565992534160614, + -0.018917765468358994, + 0.011697173118591309, + 0.007997906766831875, + 0.004799680784344673, + 0.023469267413020134, + 0.009664918296039104, + 0.007894889451563358, + -0.04281783476471901, + -0.017569173127412796, + 0.02921951934695244, + 0.040794942528009415, + -0.023787684738636017, + 0.013841810636222363, + 0.04124447703361511, + -0.002535636071115732, + 0.01271798275411129, + 0.011978129856288433, + -0.02232670970261097, + -0.04135685786604881, + -0.002619923325255513, + -0.03480119630694389, + 0.016838684678077698, + 0.0310363732278347, + -0.016005178913474083, + 0.013682601973414421, + -0.006705504842102528, + -0.011097798123955727, + -0.02687821164727211, + 0.002102494239807129, + 0.03644947707653046, + 0.03888443857431412, + 0.0315420962870121, + -0.028751257807016373, + 0.051995761692523956, + -0.014937542378902435, + 0.0018285613041371107, + 0.0020779105834662914, + -0.05240783095359802, + 0.03425801172852516, + 0.026953132823109627, + 0.06267212331295013, + 0.010020796209573746, + 0.03259100392460823, + 0.0234318058937788, + 0.008007272146642208, + 0.012689887546002865, + 0.03985841944813728, + 0.05488025024533272, + 0.021858448162674904, + 0.07214973866939545, + 0.0159115269780159, + -0.020453663542866707, + -0.04027049243450165, + 0.008372516371309757, + 0.005876682233065367, + -0.05701552331447601, + -0.024143563583493233, + 0.052445292472839355, + -0.024518173187971115, + -0.03463262319564819, + -0.0329468809068203, + 0.045739784836769104, + -0.025510888546705246, + -0.027833465486764908, + 0.048174746334552765, + 0.0156586654484272, + 0.04124447703361511, + -0.0003596833848860115, + -0.029743971303105354, + 0.0014972662320360541, + -0.05083446949720383, + -0.02843284048140049, + 0.013663871213793755, + -0.0015663348603993654, + -0.010320483706891537, + -0.03315291553735733, + -0.028414109721779823, + -0.003565811552107334, + 0.008011954836547375, + 0.012764809653162956, + -0.0057455687783658504, + 0.03433293476700783, + -0.07361071556806564, + -0.04446611553430557, + -0.0036758529022336006, + 0.003003897611051798, + -0.014619125053286552, + 0.009084274061024189, + -0.009491661563515663, + 0.047425527125597, + 0.009936509653925896, + 0.00795576348900795, + 0.010620171204209328, + 0.05540470406413078, + 0.009861587546765804, + -0.022364171221852303, + -0.0022628738079220057, + -0.012006225995719433, + 0.004530430305749178, + -0.01372006256133318, + 0.026615984737873077, + -0.036187250167131424, + 0.04094478860497475, + -0.03877205401659012, + 0.02732774242758751, + 0.014862621203064919, + 0.021858448162674904, + 0.01667010970413685, + -0.01502182986587286, + 0.033621177077293396, + 0.0031748132314532995, + -0.09402691572904587, + -0.000989787746220827, + 0.027346473187208176, + 0.025585809722542763, + 0.01667947508394718, + 0.024181025102734566, + -0.02427467703819275, + -0.01249321736395359, + 0.02448071353137493, + 0.03933396935462952, + 0.01710091158747673, + 0.02449944242835045, + -0.05960032716393471, + -0.03335895016789436, + -0.039071742445230484, + -0.04547755792737007, + 0.017475519329309464, + 0.03644947707653046, + -0.04236830398440361, + -0.006752331275492907, + 0.031841784715652466, + 0.03397705778479576, + 0.08226418495178223, + 0.0621851310133934, + -0.04214353859424591, + -0.011397485621273518, + 0.030006198212504387, + 0.0031162805389612913, + -0.023356884717941284, + 0.006532248109579086, + -0.005928190890699625, + -0.016407884657382965, + 0.021184151992201805, + 0.0018777287332341075, + -0.047425527125597, + -0.04903634637594223, + 0.026840750128030777, + -0.021502569317817688, + -0.01077001541852951, + -0.008461485616862774, + -0.02521120011806488, + -0.037685688585042953, + -0.011650347150862217, + 0.02054731547832489, + 0.021090500056743622, + 0.01893649622797966, + -0.028357917442917824, + 0.012689887546002865, + -0.001998306019231677, + -0.08511121571063995, + 0.06967731565237045, + 0.030399538576602936, + -0.015471360646188259, + 0.04203115403652191, + -0.043791819363832474, + 0.02854522317647934, + 0.010301753878593445, + 0.011388120241463184, + 0.022233057767152786, + -0.012661791406571865, + 0.0016037956811487675, + -0.019685715436935425, + -0.01944221928715706, + -0.019292375072836876, + -0.012905287556350231, + 0.009908413514494896, + 0.0161643885076046, + -0.0068834442645311356, + 0.02015397697687149, + -0.018589982762932777, + -0.019966671243309975, + 0.0020123538561165333, + 0.014328802935779095, + -0.050722088664770126, + -0.014890716411173344, + 0.07177512347698212, + 0.044503577053546906, + -0.011921938508749008, + 0.010320483706891537, + -0.014478646218776703, + 0.01338291447609663, + 0.010526519268751144, + -0.01812172122299671, + -0.02142764814198017, + 0.01088239811360836, + 0.00478329136967659, + -0.05225798487663269, + 0.02693440392613411, + 0.022682588547468185, + -0.005764299537986517, + -0.02238290011882782, + -0.024087373167276382, + -0.019311105832457542, + 0.05296974256634712, + -0.0651070848107338, + 0.020809542387723923, + 0.015780413523316383, + -0.024780400097370148, + -0.0027042103465646505, + -0.02914459817111492, + 0.014207054860889912, + 0.020079053938388824, + 0.04768775403499603, + 0.0315982885658741, + 0.03526945784687996, + 0.013439105823636055, + 0.012886556796729565, + 0.007670123595744371, + 0.023974990472197533, + 0.017597267404198647, + 0.007866793312132359, + -0.048174746334552765, + -0.0489988848567009, + 0.040457796305418015, + 0.08016636967658997, + 0.013682601973414421, + -0.018046800047159195, + -0.06683028489351273, + -0.034351665526628494, + -0.006799157243221998, + 0.0028048865497112274, + -0.020341280847787857, + 0.01605200581252575, + -0.04637662321329117, + -0.06993953883647919, + 0.017569173127412796, + -0.06008731946349144, + -0.006916222628206015, + -0.03309672325849533, + -0.027758542448282242, + 0.004675591364502907, + -0.00883141253143549, + -0.03899681940674782, + 0.07012684643268585, + -0.0009002327569760382, + -0.02620391547679901, + -0.025023896247148514, + 0.06802903115749359, + 0.030193503946065903, + -0.00706606637686491, + 0.052332907915115356, + 0.02976270206272602, + 0.03502596169710159, + -0.012886556796729565, + -3.263197504566051e-05, + 0.019179992377758026, + -0.02637248858809471, + 0.017278850078582764, + -0.06031208485364914, + -0.012249721214175224, + 0.00553953368216753, + -0.029182057827711105, + 0.040907327085733414, + -0.010788745246827602, + 0.01755044236779213, + -0.042293380945920944, + 0.02043493278324604, + -0.0234318058937788, + 0.02427467703819275, + -0.030455728992819786, + 0.011584790423512459, + -0.045515019446611404, + -0.015181038528680801, + 0.0005320621421560645, + 0.00795576348900795, + -0.01710091158747673, + 0.017082180827856064, + 0.029687780886888504, + 0.003418309148401022, + 0.020734621211886406, + -0.07994160801172256, + 0.007215910125523806, + 0.022457823157310486, + -0.0006052280077710748, + -0.015817875042557716, + 0.02682201936841011, + -0.058176811784505844, + 0.05413103103637695, + -0.004389951936900616, + 0.03639328479766846, + 0.011172720231115818, + -0.0489988848567009, + -0.005675329826772213, + -0.031111296266317368, + -0.0019268961623311043, + -0.0032825132366269827, + -0.017981242388486862, + 0.036955200135707855, + -0.03377101942896843, + -0.030418269336223602, + 0.003273148089647293, + -0.02464928664267063, + -0.0011630445951595902, + 0.04394166171550751, + 0.020416202023625374, + 0.021034307777881622, + 0.0008352614822797477, + 0.05083446949720383, + -0.03319037705659866, + -0.009290308691561222, + -0.02464928664267063, + 0.04083240404725075, + -0.049785565584897995, + -0.031298600137233734, + 0.02560454048216343, + -0.01760663278400898, + 0.014263246208429337, + -0.01757853850722313, + -0.00598438223823905, + -0.005591042805463076, + 0.02594168856739998, + -0.012867826968431473, + -0.0477626770734787, + 0.033902134746313095, + -0.025061357766389847, + -0.05274497717618942, + 0.031523365527391434, + 0.03970857709646225, + 0.01596771739423275, + 0.013888636603951454, + 0.03877205401659012, + -0.009739839471876621, + -0.009786665439605713, + -0.04761283099651337, + -0.008143068291246891, + 0.032684653997421265, + 0.03822886943817139, + 0.009163877926766872, + -0.028170613572001457, + 0.005843903869390488, + 0.020753350108861923, + -0.007281466852873564, + -0.014216420240700245, + -0.014057210646569729, + -0.006424548104405403, + -0.011172720231115818, + -0.03332148864865303, + 0.03974603861570358, + 0.013757524080574512, + -0.015012464486062527, + 0.0028165930416435003, + 0.01327053178101778, + -0.019910480827093124, + 0.006625900510698557, + -0.022626396268606186, + -0.011912573128938675, + -0.02537977509200573, + -0.024574365466833115, + 0.007300197146832943, + -0.023169580847024918, + -0.047425527125597, + 0.0003184178494848311, + 0.02537977509200573, + -0.16213087737560272, + 0.02176479622721672, + 0.0014211736852303147, + 0.0010805134661495686, + -0.021952100098133087, + -0.02803950011730194, + -0.036280903965234756, + -0.023281963542103767, + 0.028245534747838974, + -0.02571692317724228, + 0.011734633706510067, + -0.03586883470416069, + 0.035793911665678024, + -0.030736686661839485, + -0.02288862317800522, + 0.02682201936841011, + 0.012006225995719433, + 0.04476580023765564, + -0.06004985794425011, + 0.09335261583328247, + -0.005075954832136631, + -0.030418269336223602, + -0.004537454340606928, + 0.005057224538177252, + -0.048624277114868164, + -0.0246118251234293, + 0.031017644330859184, + 0.02798330970108509, + -0.03510088473558426, + 0.03560660779476166, + -0.02120288275182247, + 0.031298600137233734, + 0.02449944242835045, + 0.05858888104557991, + 0.03335895016789436, + -0.01662328466773033, + 0.011725268326699734, + -0.04349213093519211, + -0.015340247191488743, + -0.00708011444658041, + -0.05031001940369606, + 0.01857125200331211, + -0.009627456776797771, + -0.015265326015651226, + 0.03616851940751076, + 0.06383340805768967, + -0.027833465486764908, + 0.009430787526071072, + -0.037685688585042953, + 0.02260766737163067, + -0.012202895246446133, + 0.021614952012896538, + -0.009009351953864098, + -0.009674282744526863, + -0.02116542123258114, + -0.017953146249055862, + -0.01185638178139925, + -0.013869906775653362, + -0.011453676968812943, + 0.016951067373156548, + -0.01976063661277294, + -0.012540043331682682, + -0.0244432520121336, + -0.042068615555763245, + -0.018955226987600327, + 0.011697173118591309, + -0.06926524639129639, + 0.0025239295791834593, + -0.015920892357826233, + 0.009945875033736229, + -0.035119615495204926, + 0.027009325101971626, + 0.02015397697687149, + -0.009777300991117954, + -0.015977082774043083, + 0.05308212712407112, + 0.06327150017023087, + 0.021558761596679688, + 0.017091546207666397, + -0.01766282506287098, + -0.013336088508367538, + 0.005932873580604792, + -0.0006385916494764388, + -0.015649300068616867, + 0.015049925073981285, + 0.03538184240460396, + -0.005164924543350935, + 0.003172471886500716, + -0.03903428092598915, + -0.01271798275411129, + -0.09005605429410934, + -0.008620694279670715, + 0.045664865523576736, + 0.003343387274071574, + -0.0031935435254126787, + 0.08196450024843216, + -0.023019736632704735, + -0.03611232712864876, + -0.037123773247003555, + -0.014216420240700245, + -0.009983335621654987, + 0.02026635967195034, + 0.00901871733367443, + 0.027964578941464424, + 0.009561900049448013, + 0.0246118251234293, + -0.01416022889316082, + -0.039558734744787216, + 0.006003112997859716, + 0.024087373167276382, + -0.036131057888269424, + -0.04008318483829498, + -0.04064510017633438, + -0.01315814908593893, + -0.08001653105020523, + -0.01819664239883423, + -0.020228898152709007, + 0.014759602956473827, + -0.05536724254488945, + -0.021577490493655205, + 0.035962484776973724, + 0.008822047151625156, + 0.01771901547908783, + -0.027908386662602425, + 0.019854288548231125, + 0.01024556253105402, + 0.029800163581967354, + 0.002376427175477147, + -0.007960446178913116, + 0.005197702907025814, + -0.019236182793974876, + 0.000641518272459507, + 0.04941095784306526, + 0.059113334864377975, + -0.0018589982064440846, + -0.006073351949453354, + -0.012024955824017525, + 0.013420375064015388, + -0.05192083865404129, + -0.03547549247741699, + -0.011425580829381943, + -0.008737759664654732, + 0.024836590513586998, + -0.014075941406190395, + -0.008587916381657124, + 0.013195609673857689, + 0.032628461718559265, + 0.003600931027904153, + -0.021633682772517204, + -0.0024490077048540115, + 0.018814748153090477, + -0.010835572145879269, + 0.036730434745550156, + 0.056116461753845215, + -0.047238223254680634, + 0.013729427941143513, + -0.006794474553316832, + 0.02987508475780487, + 0.008246085606515408, + 0.032347507774829865, + -0.04191877320408821, + 0.03465135395526886, + -0.061061304062604904, + -0.030792877078056335, + 0.013111323118209839, + 0.03854728862643242, + 0.03127986937761307, + 0.006546296179294586, + -0.062447357922792435, + 0.021970830857753754, + 0.021408917382359505, + 0.04547755792737007, + -0.034070707857608795, + -0.003069454338401556, + 0.006246608681976795, + -0.03761076554656029, + 0.017119640484452248, + 0.03736726939678192, + 0.018608713522553444, + -0.030774148181080818, + 0.021408917382359505, + -0.015321517363190651, + 0.019704444333910942, + 0.009730474092066288, + 0.02082827314734459, + 0.017110276967287064, + -0.012146703898906708, + 0.018589982762932777, + -0.006349626462906599, + 0.01249321736395359, + 0.025248661637306213, + -0.02309465780854225, + -0.053681500256061554, + -0.03335895016789436, + 0.04158162325620651, + -0.02916332706809044, + -0.02815188281238079, + 0.026110263541340828, + -0.014825159683823586, + 0.06844110786914825, + -0.04536517709493637, + -0.03854728862643242, + 0.013261166401207447, + -0.049785565584897995, + 0.008058780804276466, + 0.020453663542866707, + -0.010535884648561478, + 0.019517140462994576, + -0.04693853482604027, + 0.016342326998710632, + -0.020640967413783073, + -0.01554628275334835, + 0.017896955832839012, + 0.012184164486825466, + -0.02148383855819702, + 0.01065763272345066, + 0.0030835021752864122, + 0.019348565489053726, + -0.02071589045226574, + 0.05364403873682022, + -0.017054084688425064, + -0.004036414436995983, + 0.026335028931498528, + -0.03764822706580162, + -0.0230384673923254, + 0.00017208610370289534, + -0.0007825820939615369, + 0.0464140810072422, + 0.02110922895371914, + 0.048849042505025864, + 0.021633682772517204, + 0.024349600076675415, + -0.02910713665187359, + 0.022982275113463402, + -0.01966698467731476, + -0.011921938508749008, + 0.027065515518188477, + 0.017232025042176247, + -0.03560660779476166, + -0.01119145005941391, + 0.0029125867877155542, + -0.019123800098896027, + -0.04903634637594223, + -0.02921951934695244 + ], + "key": "Test Memory Example", + "status": "activated" + } + } + ], + "edges": [], + "total_nodes": 4, + "total_edges": 0 +} \ No newline at end of file diff --git a/examples/data/mem_scheduler/scene_data.json b/examples/data/mem_scheduler/scene_data.json deleted file mode 100644 index 94f95c9ba..000000000 --- a/examples/data/mem_scheduler/scene_data.json +++ /dev/null @@ -1,118 +0,0 @@ -[ - [ - { - "role": "user", - "content": "I'm feeling a bit down today." - }, - { - "role": "assistant", - "content": "I'm sorry to hear that. Do you want to talk about what's been going on?" - }, - { - "role": "user", - "content": "It's just been a tough couple of days, you know? Everything feels a bit overwhelming, and I just can't seem to shake it off." - }, - { - "role": "assistant", - "content": "It sounds like you're going through a lot right now. Sometimes it helps to talk things out. Is there something specific that's been weighing on you, or is it more of a general feeling?" - }, - { - "role": "user", - "content": "It's a mix, I guess. Work's been really stressful, and on top of that, I've been feeling kinda disconnected from the people around me." - }, - { - "role": "assistant", - "content": "That can be really draining, especially when you're feeling isolated on top of the stress. Do you think there's something from your past that's contributing to how you're feeling now? Sometimes our emotions are tied to older experiences." - }, - { - "role": "user", - "content": "Yeah, now that you mention it... I've been thinking a lot about my childhood lately. I didn't have the easiest upbringing. My parents were always busy, and I often felt like I had to figure things out on my own. I remember one winter, I got really sick with a high fever. My parents were both working, and I didn't want to bother them. I just lay in bed, crying by myself, not knowing what to do." - }, - { - "role": "assistant", - "content": "That must have been very lonely. Growing up without that support can really make you feel like you have to handle everything on your own. How do you think that experience has affected you now?" - }, - { - "role": "user", - "content": "Yeah, it was really lonely. I remember spending a lot of time doing things on my own, like drawing, because it was my only way to escape. I used to draw animals, and I created my own little world on paper. It was peaceful in a way, just me and my imagination, with no outside distractions." - }, - { - "role": "assistant", - "content": "Drawing sounds like a beautiful way to express yourself and find some peace in that solitude. It must have been a very quiet and private world for you. Do you think that sense of loneliness from back then has shaped how you connect with others today?" - }, - { - "role": "user", - "content": "I think it has. I've always found it hard to let others in. I feel like I have to handle everything on my own. If I rely on someone else, I feel vulnerable, and I start to feel that loneliness again." - }, - { - "role": "assistant", - "content": "It sounds like this was a protective mechanism that helped you cope with your feelings of loneliness back then. But as you've grown, it seems like this pattern might not be serving you as well anymore. Have you thought about trying to break this cycle, or finding new ways to connect with others?" - }, - { - "role": "user", - "content": "I've thought about it, though it's hard. But I also know I can't keep going like this forever. Maybe it's time to seek some help." - }, - { - "role": "assistant", - "content": "It's a big step, but a positive one. Exploring your emotions and past experiences could really help you understand how they've shaped your present. Therapy could be a great place to start, as it would help you connect with both yourself and others. Would you be open to trying it?" - }, - { - "role": "user", - "content": "I'm open to trying. Thank you for encouraging me. It feels like it might be a good starting point for me." - }, - { - "role": "assistant", - "content": "I'm really glad to hear that. Taking that first step is always difficult, but you've made an important decision. If you need help finding the right therapist, or just want to talk, I'm here for you." - } - ], - [ - { - "role": "user", - "content": "This week I've been feeling a bit off, especially when playing tennis. My body just doesn't feel right." - }, - { - "role": "assistant", - "content": "It sounds like you've been having some physical discomfort lately. Where specifically are you feeling unwell? Is there something particular about tennis that's been difficult?" - }, - { - "role": "user", - "content": "I just feel a bit fatigued, and every time I play tennis, my movements feel off. I used to play pretty well, but lately, I've been feeling weak and my movements aren't as coordinated as they used to be." - }, - { - "role": "assistant", - "content": "Physical discomfort can definitely affect performance, especially in a sport like tennis that requires quick reactions and coordination. Have you noticed anything specific that might be causing these changes? Could it be overtraining, or is there another physical issue making you feel off?" - }, - { - "role": "user", - "content": "I think it might be due to stress and lack of sleep recently. I just feel drained all the time. Plus, I've been frustrated with my tennis performance, which makes me feel even worse, like I'm stuck in a vicious cycle." - }, - { - "role": "assistant", - "content": "Stress and poor sleep quality can definitely affect your physical performance, especially in sports. When you're also feeling down about your performance, it's easy to fall into a negative cycle. Have you tried anything to relieve some of the stress, or to improve your sleep?" - }, - { - "role": "user", - "content": "I've tried some relaxation techniques, like deep breathing and simple meditation, but it doesn't seem to help much. I still can't focus during tennis, and I feel like my mind is distracted by other things." - }, - { - "role": "assistant", - "content": "Relaxation techniques like deep breathing and meditation can help reduce stress, but sometimes we need a more structured approach to improve both our physical and mental state. For example, you might want to focus on specific aspects of your tennis training, or adjust your rest and recovery time. Have you thought about setting smaller goals for yourself? Like focusing on a specific tennis move each day, rather than expecting perfection right away? That might help you gradually regain confidence." - }, - { - "role": "user", - "content": "That's a good idea. I think I've been expecting too much from myself, which just makes me more anxious. Maybe I can start with small practice sessions and gradually find my rhythm again." - }, - { - "role": "assistant", - "content": "Exactly! Setting small and clear goals will help reduce pressure and give you a sense of accomplishment. And don't forget that rest is just as important—taking time to recover will help you feel better physically and mentally. Both physical and mental recovery are key to improving your performance and health." - }, - { - "role": "user", - "content": "Thanks, I feel like I need to rethink a few things. Maybe I really need to be kinder to myself and take things step by step." - }, - { - "role": "assistant", - "content": "You're on the right track. Rethinking things and giving yourself space and patience is so important. You've already taken the first step, now just keep moving forward, one step at a time. If you need anything, I'm always here to help." - } - ] -] diff --git a/examples/mem_api/pipeline_test.py b/examples/mem_api/pipeline_test.py deleted file mode 100644 index cd7b3bee3..000000000 --- a/examples/mem_api/pipeline_test.py +++ /dev/null @@ -1,178 +0,0 @@ -""" -Pipeline test script for MemOS Server API functions. -This script directly tests add and search functionalities without going through the API layer. -If you want to start server_api set .env to MemOS/.env and run: -uvicorn memos.api.server_api:app --host 0.0.0.0 --port 8002 --workers 4 -""" - -from typing import Any - -from dotenv import load_dotenv - -# Import directly from server_router to reuse initialized components -from memos.api.routers.server_router import ( - _create_naive_mem_cube, - mem_reader, -) -from memos.log import get_logger - - -# Load environment variables -load_dotenv() - -logger = get_logger(__name__) - - -def test_add_memories( - messages: list[dict[str, str]], - user_id: str, - mem_cube_id: str, - session_id: str = "default_session", -) -> list[str]: - """ - Test adding memories to the system. - - Args: - messages: List of message dictionaries with 'role' and 'content' - user_id: User identifier - mem_cube_id: Memory cube identifier - session_id: Session identifier - - Returns: - List of memory IDs that were added - """ - logger.info(f"Testing add memories for user: {user_id}, mem_cube: {mem_cube_id}") - - # Create NaiveMemCube using server_router function - naive_mem_cube = _create_naive_mem_cube() - - # Extract memories from messages using server_router's mem_reader - memories = mem_reader.get_memory( - [messages], - type="chat", - info={ - "user_id": user_id, - "session_id": session_id, - }, - ) - - # Flatten memory list - flattened_memories = [mm for m in memories for mm in m] - - # Add memories to the system - mem_id_list: list[str] = naive_mem_cube.text_mem.add( - flattened_memories, - user_name=mem_cube_id, - ) - - logger.info(f"Added {len(mem_id_list)} memories: {mem_id_list}") - - # Print details of added memories - for memory_id, memory in zip(mem_id_list, flattened_memories, strict=False): - logger.info(f" - ID: {memory_id}") - logger.info(f" Memory: {memory.memory}") - logger.info(f" Type: {memory.metadata.memory_type}") - - return mem_id_list - - -def test_search_memories( - query: str, - user_id: str, - mem_cube_id: str, - session_id: str = "default_session", - top_k: int = 5, - mode: str = "fast", - internet_search: bool = False, - moscube: bool = False, - chat_history: list | None = None, -) -> list[Any]: - """ - Test searching memories from the system. - - Args: - query: Search query text - user_id: User identifier - mem_cube_id: Memory cube identifier - session_id: Session identifier - top_k: Number of top results to return - mode: Search mode - internet_search: Whether to enable internet search - moscube: Whether to enable moscube search - chat_history: Chat history for context - - Returns: - List of search results - """ - - # Create NaiveMemCube using server_router function - naive_mem_cube = _create_naive_mem_cube() - - # Prepare search filter - search_filter = {"session_id": session_id} if session_id != "default_session" else None - - search_results = naive_mem_cube.text_mem.search( - query=query, - user_name=mem_cube_id, - top_k=top_k, - mode=mode, - manual_close_internet=not internet_search, - moscube=moscube, - search_filter=search_filter, - info={ - "user_id": user_id, - "session_id": session_id, - "chat_history": chat_history or [], - }, - ) - - # Print search results - for idx, result in enumerate(search_results, 1): - logger.info(f"\n Result {idx}:") - logger.info(f" ID: {result.id}") - logger.info(f" Memory: {result.memory}") - logger.info(f" Score: {getattr(result, 'score', 'N/A')}") - logger.info(f" Type: {result.metadata.memory_type}") - - return search_results - - -def main(): - # Test parameters - user_id = "test_user_123" - mem_cube_id = "test_cube_123" - session_id = "test_session_001" - - test_messages = [ - {"role": "user", "content": "Where should I go for Christmas?"}, - { - "role": "assistant", - "content": "There are many places to visit during Christmas, such as the Bund and Disneyland in Shanghai.", - }, - {"role": "user", "content": "What about New Year's Eve?"}, - { - "role": "assistant", - "content": "For New Year's Eve, you could visit Times Square in New York or watch fireworks at the Sydney Opera House.", - }, - ] - - memory_ids = test_add_memories( - messages=test_messages, user_id=user_id, mem_cube_id=mem_cube_id, session_id=session_id - ) - - logger.info(f"\nSuccessfully added {len(memory_ids)} memories!") - - search_queries = [ - "How to enjoy Christmas?", - "Where to celebrate New Year?", - "What are good places to visit during holidays?", - ] - - for query in search_queries: - logger.info("\n" + "-" * 80) - results = test_search_memories(query=query, user_id=user_id, mem_cube_id=mem_cube_id) - print(f"Query: '{query}' returned {len(results)} results") - - -if __name__ == "__main__": - main() diff --git a/examples/mem_chat/chat_w_generated_cube_explicit_memory_only.py b/examples/mem_chat/chat_w_generated_cube_explicit_memory_only.py index a1cce6e8a..2c72fd192 100644 --- a/examples/mem_chat/chat_w_generated_cube_explicit_memory_only.py +++ b/examples/mem_chat/chat_w_generated_cube_explicit_memory_only.py @@ -1,127 +1,148 @@ +import os +import sys + + +# Add project root to python path to ensure src modules can be imported +sys.path.append(os.path.abspath(os.path.join(os.path.dirname(__file__), "../../../src"))) + from memos.configs.mem_chat import MemChatConfigFactory from memos.configs.mem_cube import GeneralMemCubeConfig from memos.mem_chat.factory import MemChatFactory from memos.mem_cube.general import GeneralMemCube -mem_chat_config = MemChatConfigFactory.model_validate( - { - "backend": "simple", - "config": { - "user_id": "user_123", - "chat_llm": { - "backend": "ollama", - "config": { - "model_name_or_path": "qwen3:1.7b", - "temperature": 0.0, - "remove_think_prefix": True, - "max_tokens": 4096, - }, - }, - "max_turns_window": 20, - "top_k": 5, - "enable_textual_memory": True, - "enable_activation_memory": False, - "enable_parametric_memory": False, - }, - } -) -mem_chat = MemChatFactory.from_config(mem_chat_config) - -# Initialize_memory_cube -config = GeneralMemCubeConfig.model_validate( - { - "user_id": "user03alice", - "cube_id": "user03alice/mem_cube_tree", - "text_mem": { - "backend": "tree_text", - "config": { - "extractor_llm": { - "backend": "ollama", - "config": { - "model_name_or_path": "qwen3:1.7b", - "temperature": 0.0, - "remove_think_prefix": True, - "max_tokens": 8192, - }, - }, - "dispatcher_llm": { - "backend": "ollama", - "config": { - "model_name_or_path": "qwen3:1.7b", - "temperature": 0.0, - "remove_think_prefix": True, - "max_tokens": 8192, - }, - }, - "graph_db": { - "backend": "neo4j", - "config": { - "uri": "bolt://localhost:7687", - "user": "neo4j", - "password": "12345678", - "db_name": "user03alice_mem_cube_3", - "auto_create": True, - }, - }, - "embedder": { - "backend": "ollama", - "config": { - "model_name_or_path": "nomic-embed-text:latest", - }, - }, - }, - }, - "act_mem": { - "backend": "kv_cache", +def get_mem_chat_config() -> MemChatConfigFactory: + """ + Generates the configuration object for MemChat. + + MemChat is the top-level component for user interaction, responsible for managing the conversation flow, + invoking the LLM, and interacting with the memory module. + The configuration includes: + - user_id: User identifier + - chat_llm: LLM configuration used for chat (uses OpenAI compatible interface here) + - max_turns_window: Size of the conversation history window + - enable_textual_memory: Whether to enable textual memory (Explicit Memory) + """ + return MemChatConfigFactory.model_validate( + { + "backend": "simple", "config": { - "memory_filename": "activation_memory.pickle", - "extractor_llm": { - "backend": "huggingface", + "user_id": "user_123", + "chat_llm": { + "backend": "openai", "config": { - "model_name_or_path": "Qwen/Qwen3-1.7B", + # Prioritize getting sensitive information and model configuration from environment variables + "model_name_or_path": os.getenv("MOS_CHAT_MODEL", "gpt-4o"), "temperature": 0.8, "max_tokens": 1024, "top_p": 0.9, "top_k": 50, - "add_generation_prompt": True, - "remove_think_prefix": False, + "api_key": os.getenv("OPENAI_API_KEY"), + "api_base": os.getenv("OPENAI_API_BASE"), }, }, + "max_turns_window": 20, + "top_k": 5, + # Enable textual memory functionality, allowing the system to retrieve and store explicit memories + "enable_textual_memory": True, + # This example demonstrates only explicit memory, so activation memory and parametric memory are disabled + "enable_activation_memory": False, + "enable_parametric_memory": False, }, - }, - "para_mem": { - "backend": "lora", - "config": { - "memory_filename": "parametric_memory.adapter", - "extractor_llm": { - "backend": "huggingface", - "config": { - "model_name_or_path": "Qwen/Qwen3-1.7B", - "temperature": 0.8, - "max_tokens": 1024, - "top_p": 0.9, - "top_k": 50, - "add_generation_prompt": True, - "remove_think_prefix": False, + } + ) + + +def get_mem_cube_config() -> GeneralMemCubeConfig: + """ + Generates the configuration object for GeneralMemCube. + + MemCube (Memory Cube) is the core storage and management unit for memory. + GeneralMemCube is a general implementation of the memory cube, supporting extraction, vectorized storage, and retrieval of textual memory. + The configuration includes: + - user_id / cube_id: Identifiers for the user and the cube to which the memory belongs + - text_mem: Specific configuration for textual memory + - extractor_llm: LLM used to extract memory fragments from the conversation + - vector_db: Database used to store memory vectors (uses Qdrant here) + - embedder: Model used to generate text vectors (uses OpenAI compatible interface here) + """ + return GeneralMemCubeConfig.model_validate( + { + "user_id": "user03alice", + "cube_id": "user03alice/mem_cube_tree", + "text_mem": { + "backend": "general_text", + "config": { + "cube_id": "user03alice/mem_cube_general", + "memory_filename": "textual_memory.json", + "extractor_llm": { + "backend": "openai", + "config": { + "model_name_or_path": os.getenv("MOS_CHAT_MODEL", "gpt-4o"), + "temperature": 0.8, + "max_tokens": 1024, + "top_p": 0.9, + "top_k": 50, + "api_key": os.getenv("OPENAI_API_KEY"), + "api_base": os.getenv("OPENAI_API_BASE"), + }, + }, + "vector_db": { + "backend": "qdrant", + "config": { + "collection_name": "user03alice_mem_cube_general", + "vector_dimension": 1024, + "distance_metric": "cosine", + }, + }, + "embedder": { + "backend": os.getenv("MOS_EMBEDDER_BACKEND", "universal_api"), + "config": { + "provider": "openai", + "api_key": os.getenv("MOS_EMBEDDER_API_KEY", "EMPTY"), + "model_name_or_path": os.getenv("MOS_EMBEDDER_MODEL", "bge-m3"), + "base_url": os.getenv("MOS_EMBEDDER_API_BASE"), + }, }, }, }, - }, - }, -) + } + ) + + +def main(): + """ + Main program entry point: + 1. Initialize MemChat (Conversation Controller) + 2. Initialize MemCube (Memory Storage) + 3. Mount MemCube to MemChat + 4. Start the chat loop + 5. Save memory after the chat ends + """ + print("Initializing MemChat...") + mem_chat_config = get_mem_chat_config() + mem_chat = MemChatFactory.from_config(mem_chat_config) + + print("Initializing MemCube...") + mem_cube_config = get_mem_cube_config() + mem_cube = GeneralMemCube(mem_cube_config) -# Initialize the MemCube with the configuration -mem_cube = GeneralMemCube(config) + # Mount the initialized memory cube onto the chat system + # This allows MemChat to perform memory retrieval (search) and organization (organize) via mem_cube during the conversation + mem_chat.mem_cube = mem_cube -# TODO: Read memory and prepare data -# Hope to read user docs and save in a file + print("Starting Chat Session...") + try: + mem_chat.run() + except KeyboardInterrupt: + print("\nChat session interrupted.") + finally: + # Ensure memory is persisted to disk before the program exits + # The dump method saves the in-memory memory state to the specified path + print("Saving memory cube...") + mem_chat.mem_cube.dump("new_cube_path") + print("Memory cube saved to 'new_cube_path'.") -# TODO: Organize MemoryCube -# Call Tree.add() -# save in memory cube -# chat and search and organize -mem_chat.mem_cube = mem_cube -mem_chat.run() -mem_chat.mem_cube.dump("new_cube_path") +if __name__ == "__main__": + main() diff --git a/examples/mem_chat/simple_mem_chat.py b/examples/mem_chat/simple_mem_chat.py deleted file mode 100644 index a6b4f5cbe..000000000 --- a/examples/mem_chat/simple_mem_chat.py +++ /dev/null @@ -1,33 +0,0 @@ -from memos.configs.mem_chat import MemChatConfigFactory -from memos.mem_chat.factory import MemChatFactory -from memos.mem_cube.general import GeneralMemCube - - -config = MemChatConfigFactory.model_validate( - { - "backend": "simple", - "config": { - "user_id": "user_123", - "chat_llm": { - "backend": "huggingface", - "config": { - "model_name_or_path": "Qwen/Qwen3-1.7B", - "temperature": 0.1, - "remove_think_prefix": True, - "max_tokens": 4096, - }, - }, - "max_turns_window": 20, - "top_k": 5, - "enable_textual_memory": True, - "enable_activation_memory": False, - "enable_parametric_memory": False, - }, - } -) -mem_chat = MemChatFactory.from_config(config) -mem_chat.mem_cube = GeneralMemCube.init_from_dir("examples/data/mem_cube_2") - -mem_chat.run() - -mem_chat.mem_cube.dump("tmp/mem_cube") diff --git a/examples/mem_chunk/markdown_chunk.py b/examples/mem_chunk/markdown_chunk.py deleted file mode 100644 index ce7d2b9ae..000000000 --- a/examples/mem_chunk/markdown_chunk.py +++ /dev/null @@ -1,33 +0,0 @@ -from memos.chunkers import ChunkerFactory -from memos.configs.chunker import ChunkerConfigFactory - - -config = ChunkerConfigFactory.model_validate( - { - "backend": "markdown", - "config": { - "chunk_size": 1000, - "chunk_overlap": 100, - "recursive": True, - }, - } -) - -chunker = ChunkerFactory.from_config(config) - -text = """ -# Header 1 -This is the first sentence. This is the second sentence. -And here's a third one with some additional context. - -# Header 2 -This is the fourth sentence. This is the fifth sentence. -And here's a sixth one with some additional context. - -# Header 3 -This is the seventh sentence. This is the eighth sentence. -And here's a ninth one with some additional context. -""" -chunks = chunker.chunk(text) -for chunk in chunks: - print("doc:", chunk) diff --git a/examples/mem_cube/_deprecated/README.md b/examples/mem_cube/_deprecated/README.md new file mode 100644 index 000000000..da02eebef --- /dev/null +++ b/examples/mem_cube/_deprecated/README.md @@ -0,0 +1,81 @@ +# Deprecated Examples + +⚠️ **These examples are deprecated and no longer maintained.** + +## Why deprecated? + +These examples demonstrate old APIs that directly access MemCube internals (e.g., `mem_cube.text_mem.get_all()`), which is no longer the recommended approach. + +## Current Best Practice + +**Use `SingleCubeView` / `CompositeCubeView` for all add/search operations.** + +The new View architecture provides: +- ✅ Unified API interface +- ✅ Multi-cube support +- ✅ Better integration with MemOS Server +- ✅ Consistent result format with `cube_id` tracking + +## Updated Examples + +See the following files in the parent directory: +- **`../load_cube.py`** - Load MemCube and operate via SingleCubeView +- **`../dump_cube.py`** - Persist MemCube to disk + +## Migration Guide + +### Old approach (deprecated): +```python +mem_cube = GeneralMemCube.init_from_dir("examples/data/mem_cube_2") +items = mem_cube.text_mem.get_all() # ❌ Direct access +for item in items: + print(item) +``` + +### New approach (recommended): +```python +import json +from memos.api.handlers import init_server +from memos.api.product_models import APISearchRequest +from memos.multi_mem_cube.single_cube import SingleCubeView +from memos.log import get_logger + +logger = get_logger(__name__) + +# Initialize server (uses .env configuration) +components = init_server() +naive = components["naive_mem_cube"] + +# Create View +view = SingleCubeView( + cube_id="my_cube", + naive_mem_cube=naive, + mem_reader=components["mem_reader"], + mem_scheduler=components["mem_scheduler"], + logger=logger, + searcher=components["searcher"], + feedback_server=components["feedback_server"], +) + +# Load data from exported JSON +with open("examples/data/mem_cube_tree/textual_memory.json") as f: + json_data = json.load(f) +naive.text_mem.graph_store.import_graph(json_data, user_name="my_cube") + +# Use View API for search +results = view.search_memories(APISearchRequest( + user_id="user", + readable_cube_ids=["my_cube"], + query="your query here", +)) +for group in results.get("text_mem", []): + for mem in group.get("memories", []): + print(mem.get("metadata", {}).get("memory", "N/A")) +``` + +> **Note on Embeddings**: The sample data uses **bge-m3** model with **1024 dimensions**. +> Ensure your environment uses the same embedding configuration for accurate search. + +--- + +For more information, see the [MemCube documentation](https://memos-doc.memoryos.ai/open_source/modules/mem_cube). diff --git a/examples/mem_cube/load_from_folder.py b/examples/mem_cube/_deprecated/load_from_folder.py similarity index 100% rename from examples/mem_cube/load_from_folder.py rename to examples/mem_cube/_deprecated/load_from_folder.py diff --git a/examples/mem_cube/load_from_remote.py b/examples/mem_cube/_deprecated/load_from_remote.py similarity index 100% rename from examples/mem_cube/load_from_remote.py rename to examples/mem_cube/_deprecated/load_from_remote.py diff --git a/examples/mem_cube/load_lazily.py b/examples/mem_cube/_deprecated/load_lazily.py similarity index 100% rename from examples/mem_cube/load_lazily.py rename to examples/mem_cube/_deprecated/load_lazily.py diff --git a/examples/mem_cube/dump_cube.py b/examples/mem_cube/dump_cube.py new file mode 100644 index 000000000..5a2967ff9 --- /dev/null +++ b/examples/mem_cube/dump_cube.py @@ -0,0 +1,137 @@ +""" +MemCube dump example using SingleCubeView. + +Demonstrates: +1. Initialize server and create SingleCubeView with NEW cube_id +2. Add memories via View +3. Dump ONLY this cube's data to directory + +Requirements: + - MemOS service environment (.env configured) + - Neo4j graph database (set NEO4J_BACKEND=neo4j in .env) + +Note on Embeddings: + This example exports embeddings along with memory data. + The sample data uses: bge-m3 model, 1024 dimensions. + If your environment uses a different embedding model or dimension, + you may need to re-embed the data after import, or the semantic + search results may be inaccurate or fail. +""" + +import contextlib +import json +import os +import shutil + +from memos.api.handlers import init_server +from memos.api.product_models import APIADDRequest +from memos.log import get_logger +from memos.multi_mem_cube.single_cube import SingleCubeView + + +logger = get_logger(__name__) + +# NEW cube_id to avoid dumping existing data +EXAMPLE_CUBE_ID = "example_dump_cube" +EXAMPLE_USER_ID = "example_user" + +# ============================================================================= +# Step 1: Initialize server +# ============================================================================= +print("=" * 60) +print("Step 1: Initialize server") +print("=" * 60) + +components = init_server() +print("✓ Server initialized") + +# ============================================================================= +# Step 2: Create SingleCubeView with NEW cube_id +# ============================================================================= +print("\n" + "=" * 60) +print(f"Step 2: Create SingleCubeView (cube_id={EXAMPLE_CUBE_ID})") +print("=" * 60) + +naive = components["naive_mem_cube"] +view = SingleCubeView( + cube_id=EXAMPLE_CUBE_ID, # NEW cube_id + naive_mem_cube=naive, + mem_reader=components["mem_reader"], + mem_scheduler=components["mem_scheduler"], + logger=logger, + searcher=components["searcher"], + feedback_server=components["feedback_server"], +) +print("✓ SingleCubeView created") + +# ============================================================================= +# Step 3: Add memories via View +# ============================================================================= +print("\n" + "=" * 60) +print("Step 3: Add memories via SingleCubeView") +print("=" * 60) + +result = view.add_memories( + APIADDRequest( + user_id=EXAMPLE_USER_ID, + writable_cube_ids=[EXAMPLE_CUBE_ID], + messages=[ + {"role": "user", "content": "This is a test memory for dump example"}, + {"role": "user", "content": "Another memory to demonstrate persistence"}, + ], + async_mode="sync", + ) +) +print(f"✓ Added {len(result)} memories") + +# ============================================================================= +# Step 4: Dump ONLY this cube's data +# ============================================================================= +print("\n" + "=" * 60) +print("Step 4: Dump cube data (filtered by cube_id)") +print("=" * 60) + +output_dir = "tmp/mem_cube_dump" +if os.path.exists(output_dir): + shutil.rmtree(output_dir) +os.makedirs(output_dir, exist_ok=True) + +# Export only this cube's data using user_name filter +text_mem = naive.text_mem +json_data = text_mem.graph_store.export_graph( + include_embedding=True, # Include embeddings for semantic search + user_name=EXAMPLE_CUBE_ID, # Filter by cube_id +) + +# Fix embedding format: parse string to list for import compatibility +# (export_graph stores embedding as string in metadata, but add_node expects list) +for node in json_data.get("nodes", []): + metadata = node.get("metadata", {}) + if "embedding" in metadata and isinstance(metadata["embedding"], str): + with contextlib.suppress(json.JSONDecodeError): + metadata["embedding"] = json.loads(metadata["embedding"]) + +print(f"✓ Exported {len(json_data.get('nodes', []))} nodes") + +# Save to file +memory_file = os.path.join(output_dir, "textual_memory.json") +with open(memory_file, "w", encoding="utf-8") as f: + json.dump(json_data, f, indent=2, ensure_ascii=False) +print(f"✓ Saved to: {memory_file}") + +# Save config (user can modify sensitive fields before sharing) +config = components["default_cube_config"].model_copy(deep=True) +config.user_id = EXAMPLE_USER_ID +config.cube_id = EXAMPLE_CUBE_ID +config_file = os.path.join(output_dir, "config.json") +config.to_json_file(config_file) +print(f"✓ Config saved to: {config_file}") + +# ============================================================================= +# Done +# ============================================================================= +print("\n" + "=" * 60) +print("✅ Example completed!") +print("=" * 60) +print(f"\nDumped to: {output_dir}") +print("Run load_cube.py to load this data") diff --git a/examples/mem_cube/load_cube.py b/examples/mem_cube/load_cube.py new file mode 100644 index 000000000..c7552a4a2 --- /dev/null +++ b/examples/mem_cube/load_cube.py @@ -0,0 +1,137 @@ +""" +MemCube load example using SingleCubeView. + +Demonstrates: +1. Initialize server and create SingleCubeView +2. Load memories from dump via graph_store.import_graph() +3. Display loaded memories +4. Search loaded memories (semantic search) + +Requirements: + - MemOS service environment (.env configured) + - Neo4j graph database (set NEO4J_BACKEND=neo4j in .env) + +Note on Embeddings: + The sample data (examples/data/mem_cube_tree) uses: bge-m3 model, 1024 dimensions. + For semantic search to work correctly, your environment must use the same + embedding model and dimension. If different, search results may be inaccurate. +""" + +import json +import os + +from memos.api.handlers import init_server +from memos.api.product_models import APISearchRequest +from memos.log import get_logger +from memos.multi_mem_cube.single_cube import SingleCubeView + + +logger = get_logger(__name__) + +EXAMPLE_CUBE_ID = "example_dump_cube" +EXAMPLE_USER_ID = "example_user" + +# ============================================================================= +# Step 1: Initialize server +# ============================================================================= +print("=" * 60) +print("Step 1: Initialize server") +print("=" * 60) + +components = init_server() +print("✓ Server initialized") + +# ============================================================================= +# Step 2: Create SingleCubeView +# ============================================================================= +print("\n" + "=" * 60) +print(f"Step 2: Create SingleCubeView (cube_id={EXAMPLE_CUBE_ID})") +print("=" * 60) + +naive = components["naive_mem_cube"] +view = SingleCubeView( + cube_id=EXAMPLE_CUBE_ID, + naive_mem_cube=naive, + mem_reader=components["mem_reader"], + mem_scheduler=components["mem_scheduler"], + logger=logger, + searcher=components["searcher"], + feedback_server=components["feedback_server"], +) +print("✓ SingleCubeView created") + +# ============================================================================= +# Step 3: Load memories from dump +# ============================================================================= +print("\n" + "=" * 60) +print("Step 3: Load memories from dump") +print("=" * 60) + +load_dir = "examples/data/mem_cube_tree" +memory_file = os.path.join(load_dir, "textual_memory.json") + +if not os.path.exists(memory_file): + print(f"❌ File not found: {memory_file}") + print(" Run dump_cube.py first to create data!") + exit(1) + +with open(memory_file, encoding="utf-8") as f: + json_data = json.load(f) + +# Import graph data into graph_store +text_mem = naive.text_mem +text_mem.graph_store.import_graph(json_data, user_name=EXAMPLE_CUBE_ID) + +nodes = json_data.get("nodes", []) +edges = json_data.get("edges", []) +print(f"✓ Imported {len(nodes)} nodes, {len(edges)} edges") + +# ============================================================================= +# Step 4: Display loaded memories +# ============================================================================= +print("\n" + "=" * 60) +print("Step 4: Display loaded memories") +print("=" * 60) + +print(f"\nLoaded {len(nodes)} memories:") +for i, node in enumerate(nodes, 1): + metadata = node.get("metadata", {}) + memory_text = node.get("memory", "N/A") + mem_type = metadata.get("memory_type", "unknown") + print(f"\n [{i}] Type: {mem_type}") + print(f" Content: {memory_text[:70]}...") + +# ============================================================================= +# Step 5: Search loaded memories +# ============================================================================= +print("\n" + "=" * 60) +print("Step 5: Search loaded memories") +print("=" * 60) + +query = "test memory dump persistence demonstration" +print(f'Query: "{query}"') + +search_result = view.search_memories( + APISearchRequest( + user_id=EXAMPLE_USER_ID, + readable_cube_ids=[EXAMPLE_CUBE_ID], + query=query, + ) +) + +text_mem_results = search_result.get("text_mem", []) +memories = [] +for group in text_mem_results: + memories.extend(group.get("memories", [])) + +print(f"\n✓ Found {len(memories)} relevant memories:") +for i, mem in enumerate(memories[:3], 1): + content = mem.get("metadata", {}).get("memory", "N/A")[:70] + print(f" [{i}] {content}...") + +# ============================================================================= +# Done +# ============================================================================= +print("\n" + "=" * 60) +print("✅ Example completed!") +print("=" * 60) diff --git a/examples/mem_feedback/example_feedback.py b/examples/mem_feedback/example_feedback.py new file mode 100644 index 000000000..8f4446863 --- /dev/null +++ b/examples/mem_feedback/example_feedback.py @@ -0,0 +1,248 @@ +import json +import os +import sys + + +# Add project root to python path to ensure src modules can be imported +sys.path.append(os.path.abspath(os.path.join(os.path.dirname(__file__), "../../../src"))) + + +def init_components(): + """ + Initialize MemOS core components. + + This function is responsible for building and configuring all basic components required for MemOS operation, including: + 1. LLM (Large Language Model): Model responsible for natural language understanding and generation (e.g., GPT-4o). + 2. Embedder: Responsible for converting text into vector representations for semantic search and similarity calculation. + 3. GraphDB (Neo4j): Graph database for persistent storage of memory nodes and their relationships. + 4. MemoryManager: Memory manager responsible for memory CRUD operations. + 5. MemReader: Memory reader for parsing and processing input text. + 6. Reranker: Reranker for refining the sorting of retrieval results. + 7. Searcher: Searcher that integrates retrieval and reranking logic. + 8. FeedbackServer (SimpleMemFeedback): Feedback service core, responsible for processing user feedback and updating memory. + + Returns: + tuple: (feedback_server, memory_manager, embedder) + """ + # Lazy import to avoid E402 (module level import not at top of file) + from memos.configs.embedder import EmbedderConfigFactory + from memos.configs.graph_db import GraphDBConfigFactory + from memos.configs.llm import LLMConfigFactory + from memos.configs.mem_reader import MemReaderConfigFactory + from memos.configs.reranker import RerankerConfigFactory + from memos.embedders.factory import EmbedderFactory + from memos.graph_dbs.factory import GraphStoreFactory + from memos.llms.factory import LLMFactory + from memos.mem_feedback.simple_feedback import SimpleMemFeedback + from memos.mem_reader.factory import MemReaderFactory + from memos.memories.textual.tree_text_memory.organize.manager import MemoryManager + from memos.memories.textual.tree_text_memory.retrieve.searcher import Searcher + from memos.reranker.factory import RerankerFactory + + print("Initializing MemOS Components...") + + # 1. LLM: Configure Large Language Model, using OpenAI compatible interface + llm_config = LLMConfigFactory.model_validate( + { + "backend": "openai", + "config": { + "model_name_or_path": os.getenv("MOS_CHAT_MODEL", "gpt-4o"), + "temperature": 0.8, + "max_tokens": 1024, + "top_p": 0.9, + "top_k": 50, + "api_key": os.getenv("OPENAI_API_KEY"), + "api_base": os.getenv("OPENAI_API_BASE"), + }, + } + ) + llm = LLMFactory.from_config(llm_config) + + # 2. Embedder: Configure embedding model for generating text vectors + embedder_config = EmbedderConfigFactory.model_validate( + { + "backend": os.getenv("MOS_EMBEDDER_BACKEND", "universal_api"), + "config": { + "provider": "openai", + "api_key": os.getenv("MOS_EMBEDDER_API_KEY", "EMPTY"), + "model_name_or_path": os.getenv("MOS_EMBEDDER_MODEL", "bge-m3"), + "base_url": os.getenv("MOS_EMBEDDER_API_BASE"), + }, + } + ) + embedder = EmbedderFactory.from_config(embedder_config) + + # 3. GraphDB: Configure Neo4j graph database connection + graph_db = GraphStoreFactory.from_config( + GraphDBConfigFactory.model_validate( + { + "backend": "neo4j", + "config": { + "uri": os.getenv("NEO4J_URI", "neo4j://127.0.0.1:7687"), + "user": os.getenv("NEO4J_USER", "neo4j"), + "password": os.getenv("NEO4J_PASSWORD", "12345678"), + "db_name": os.getenv("NEO4J_DB_NAME", "neo4j"), + "user_name": "zhs", + "auto_create": True, + "use_multi_db": False, + "embedding_dimension": int(os.getenv("EMBEDDING_DIMENSION", "1024")), + }, + } + ) + ) + + # Clear test data for specific user to ensure a clean environment for each run + graph_db.clear(user_name="cube_id_001_0115") + + # 4. MemoryManager: Core memory management, coordinating storage and retrieval + memory_manager = MemoryManager(graph_db, embedder, llm, is_reorganize=False) + + # 5. MemReader: Configure memory reader, including chunking strategy + mem_reader = MemReaderFactory.from_config( + MemReaderConfigFactory.model_validate( + { + "backend": "simple_struct", + "config": { + "llm": llm_config.model_dump(), + "embedder": embedder_config.model_dump(), + "chunker": { + "backend": "sentence", + "config": { + "tokenizer_or_token_counter": "gpt2", + "chunk_size": 512, + "chunk_overlap": 128, + "min_sentences_per_chunk": 1, + }, + }, + }, + } + ) + ) + + # 6. Reranker: Configure reranker to improve retrieval relevance + mem_reranker = RerankerFactory.from_config( + RerankerConfigFactory.model_validate( + { + "backend": os.getenv("MOS_RERANKER_BACKEND", "cosine_local"), + "config": { + "level_weights": {"topic": 1.0, "concept": 1.0, "fact": 1.0}, + "level_field": "background", + }, + } + ) + ) + + # 7. Searcher: Comprehensive searcher + searcher = Searcher(llm, graph_db, embedder, mem_reranker) + + # 8. Feedback Server: Initialize feedback service, the core of this example + feedback_server = SimpleMemFeedback( + llm=llm, + embedder=embedder, + graph_store=graph_db, + memory_manager=memory_manager, + mem_reader=mem_reader, + searcher=searcher, + reranker=mem_reranker, + pref_mem=None, + ) + + return feedback_server, memory_manager, embedder + + +def main(): + """ + Main program flow: + 1. Initialize components. + 2. Simulate a conversation scenario and existing (possibly incorrect) memory. + 3. Receive user feedback (correct memory). + 4. Process feedback and update memory store. + 5. Display processing results. + """ + # Load dotenv in main to avoid affecting module import order + from dotenv import load_dotenv + + load_dotenv() + + # Lazy import to avoid E402 + from memos.mem_feedback.utils import make_mem_item + + feedback_server, memory_manager, embedder = init_components() + print("-" * 50) + print("Initialization Done. Processing Feedback...") + print("-" * 50) + + # 1. Simulate Chat History + # Simulate a conversation between user and assistant, where the assistant's response contains a statement about user preferences. + history = [ + {"role": "user", "content": "我喜欢什么水果,不喜欢什么水果"}, + {"role": "assistant", "content": "你喜欢苹果,不喜欢香蕉"}, + ] + + # 2. Simulate Initial Memory + # We manually add a memory to the database, representing what the system currently believes to be a "fact". + # This memory content is "你喜欢苹果,不喜欢香蕉", which we will later correct via feedback. + mem_text = "你喜欢苹果,不喜欢香蕉" + memory_manager.add( + [ + make_mem_item( + mem_text, + user_id="user_id_001", + user_name="cube_id_001_0115", + session_id="session_id", + tags=["fact"], + key="food_preference", + sources=[{"type": "chat"}], + background="init from chat history", + embedding=embedder.embed([mem_text])[ + 0 + ], # Generate embedding for subsequent retrieval + info={ + "user_id": "user_id_001", + "user_name": "cube_id_001_0115", + "session_id": "session_id", + }, + ) + ], + user_name="cube_id_001_0115", + mode="sync", + ) + + # 3. Feedback Input + # The user points out the previous memory is incorrect and provides the correct information. + feedback_content = "错了,实际上我喜欢的是山竹" + + print("\nChat History:") + print(json.dumps(history, ensure_ascii=False, indent=2)) + print("\nFeedback Input:") + print(feedback_content) + + # 4. Process Feedback + # Core step: Call feedback_server to process user correction information. + # The system analyzes feedback content, retrieves relevant memories, and generates update operations (e.g., add, modify, or archive old memories). + res = feedback_server.process_feedback( + user_id="user_id_001", + user_name="cube_id_001_0115", + session_id="session_id", + chat_history=history, + feedback_content=feedback_content, + feedback_time="", + async_mode="sync", + corrected_answer="", + task_id="task_id", + info={}, + ) + + # 5. Feedback Result + print("\n" + "=" * 50) + print("Feedback Result") + print("=" * 50) + + """ + Print feedback processing results, including added or updated memory operations (add/update) + """ + print(json.dumps(res, ensure_ascii=False, indent=4, default=str)) + + +if __name__ == "__main__": + main() diff --git a/examples/mem_mcp/simple_fastmcp_client.py b/examples/mem_mcp/simple_fastmcp_client.py index 1981e3066..e5c186b3b 100644 --- a/examples/mem_mcp/simple_fastmcp_client.py +++ b/examples/mem_mcp/simple_fastmcp_client.py @@ -12,66 +12,43 @@ async def main(): print("Working FastMCP Client") print("=" * 40) - async with Client("http://127.0.0.1:8000/mcp") as client: - print("Connected to MOS MCP server!") + # Connect to MCP server via HTTP + # FastMCP HTTP endpoint is at /mcp (not /mcp/v1) + async with Client("http://localhost:8002/mcp") as client: + print("Connected to MCP server") - print("Available tools:") - tools = await client.list_tools() - for tool in tools: - print("**" * 20) - print(f" - {tool.name}: {tool.description}") + print("\nTesting tool calls via Server API...") - print("Available resources:") - resources = await client.list_resources() - for resource in resources: - print(f" - {resource.uri}: {resource.description}") + # Note: 'create_user' and 'get_user_info' are not supported by the Server API. + # We assume the user already exists or the Server API handles it implicitly. + # Using a demo user ID. + user_id = "fastmcp_demo_user" - print("Testing tool calls...") - - print(" Getting user info...") - result = await client.call_tool("get_user_info", {}) - print(f" Result: {result.content[0].text}") - - print(" Creating user...") - result = await client.call_tool( - "create_user", - {"user_id": "fastmcp_user", "role": "USER", "user_name": "FastMCP Test User"}, - ) - print(f"Result: {result.content[0].text}") - - print(" register cube...") + print("\n 1. Adding memory...") result = await client.call_tool( - "register_cube", - { - "cube_name_or_path": "cube_default_user", - "user_id": "fastmcp_user", - "cube_id": "fastmcp_user", + "add_memory", + arguments={ + "memory_content": "MemOS is a great tool for memory management.", + "user_id": user_id, }, ) print(f" Result: {result}") - print(" Adding memory...") + print("\n 2. Searching memories...") result = await client.call_tool( - "add_memory", - { - "memory_content": "This is a test memory from FastMCP client.", - "cube_id": "fastmcp_user", - "user_id": "fastmcp_user", - }, + "search_memories", + arguments={"query": "MemOS", "user_id": user_id}, ) - print(f" Result: {result.content[0].text}") + print(f" Result: {result}") - print(" Searching memories...") + print("\n 3. Chatting...") result = await client.call_tool( - "search_memories", {"query": "test memory", "user_id": "fastmcp_user"} + "chat", + arguments={"query": "What is MemOS?", "user_id": user_id}, ) - print(f" Result: {result.content[0].text[:200]}...") + print(f" Result: {result}") - print(" Testing chat...") - result = await client.call_tool( - "chat", {"query": "Hello! Tell me about yourself.", "user_id": "fastmcp_user"} - ) - print(f" Result: {result.content[0].text[:200]}...") + print("\n✓ All tests completed!") if __name__ == "__main__": diff --git a/examples/mem_mcp/simple_fastmcp_serve.py b/examples/mem_mcp/simple_fastmcp_serve.py index 78c05cd05..55ad4d84d 100644 --- a/examples/mem_mcp/simple_fastmcp_serve.py +++ b/examples/mem_mcp/simple_fastmcp_serve.py @@ -1,18 +1,67 @@ import argparse +import json import os -from memos.api.mcp_serve import MOSMCPStdioServer +import requests +from dotenv import load_dotenv +from fastmcp import FastMCP + + +load_dotenv() + +# Configuration +# This points to the Server API base URL (e.g., started via server_api.py) +API_BASE_URL = os.getenv("MEMOS_API_BASE_URL", "http://localhost:8001/product") + +# Create MCP Server +mcp = FastMCP("MemOS MCP via Server API") -if __name__ == "__main__": - import argparse - from dotenv import load_dotenv +@mcp.tool() +def add_memory(memory_content: str, user_id: str, cube_id: str | None = None): + """Add memory using the Server API.""" + payload = { + "user_id": user_id, + "messages": memory_content, + "writable_cube_ids": [cube_id] if cube_id else None, + } + try: + resp = requests.post(f"{API_BASE_URL}/add", json=payload) + resp.raise_for_status() + return resp.json()["message"] + except Exception as e: + return f"Error: {e}" - load_dotenv() +@mcp.tool() +def search_memories(query: str, user_id: str, cube_ids: str | None = None): + """Search memories using the Server API.""" + payload = {"query": query, "user_id": user_id, "readable_cube_ids": cube_ids} + try: + resp = requests.post(f"{API_BASE_URL}/search", json=payload) + resp.raise_for_status() + # The Server API search response structure matches product API mostly + return json.dumps(resp.json()["data"], ensure_ascii=False) + except Exception as e: + return f"Error: {e}" + + +@mcp.tool() +def chat(query: str, user_id: str): + """Chat using the Server API.""" + payload = {"query": query, "user_id": user_id} + try: + resp = requests.post(f"{API_BASE_URL}/chat/complete", json=payload) + resp.raise_for_status() + return resp.json()["data"]["response"] + except Exception as e: + return f"Error: {e}" + + +if __name__ == "__main__": # Parse command line arguments - parser = argparse.ArgumentParser(description="MOS MCP Server") + parser = argparse.ArgumentParser(description="MOS MCP Server via API") parser.add_argument( "--transport", choices=["stdio", "http", "sse"], @@ -24,14 +73,8 @@ args = parser.parse_args() - # Set environment variables - os.environ["OPENAI_API_BASE"] = os.getenv("OPENAI_API_BASE") - os.environ["OPENAI_API_KEY"] = os.getenv("OPENAI_API_KEY") - os.environ["MOS_TEXT_MEM_TYPE"] = "tree_text" # "tree_text" need set neo4j - os.environ["NEO4J_URI"] = os.getenv("NEO4J_URI") - os.environ["NEO4J_USER"] = os.getenv("NEO4J_USER") - os.environ["NEO4J_PASSWORD"] = os.getenv("NEO4J_PASSWORD") - - # Create and run MCP server - server = MOSMCPStdioServer() - server.run(transport=args.transport, host=args.host, port=args.port) + # For stdio transport, don't pass host and port + if args.transport == "stdio": + mcp.run(transport=args.transport) + else: + mcp.run(transport=args.transport, host=args.host, port=args.port) diff --git a/examples/mem_os/chat_w_generated_cube_explicit_memory.py b/examples/mem_os/chat_w_generated_cube_explicit_memory.py deleted file mode 100644 index d19a68486..000000000 --- a/examples/mem_os/chat_w_generated_cube_explicit_memory.py +++ /dev/null @@ -1,180 +0,0 @@ -from memos.configs.mem_cube import GeneralMemCubeConfig -from memos.configs.mem_os import MOSConfig -from memos.mem_cube.general import GeneralMemCube -from memos.mem_os.main import MOS - - -# 1. Create Mos Config -config = { - "user_id": "user03alice", - "chat_model": { - "backend": "huggingface", - "config": { - "model_name_or_path": "Qwen/Qwen3-1.7B", - "temperature": 0.1, - "remove_think_prefix": True, - "max_tokens": 4096, - }, - }, - "mem_reader": { - "backend": "simple_struct", - "config": { - "llm": { - "backend": "ollama", - "config": { - "model_name_or_path": "qwen3:0.6b", - "temperature": 0.0, - "remove_think_prefix": True, - "max_tokens": 8192, - }, - }, - "embedder": { - "backend": "ollama", - "config": { - "model_name_or_path": "nomic-embed-text:latest", - }, - }, - "chunker": { - "backend": "sentence", - "config": { - "tokenizer_or_token_counter": "gpt2", - "chunk_size": 512, - "chunk_overlap": 128, - "min_sentences_per_chunk": 1, - }, - }, - }, - }, - "max_turns_window": 20, - "top_k": 5, - "enable_textual_memory": True, - "enable_activation_memory": False, - "enable_parametric_memory": False, -} - -mos_config = MOSConfig(**config) -mos = MOS(mos_config) - -# 2. Initialize_memory_cube -config = GeneralMemCubeConfig.model_validate( - { - "user_id": "user03alice", - "cube_id": "user03alice/mem_cube_tree", - "text_mem": { - "backend": "tree_text", - "config": { - "extractor_llm": { - "backend": "ollama", - "config": { - "model_name_or_path": "qwen3:1.7b", - "temperature": 0.0, - "remove_think_prefix": True, - "max_tokens": 8192, - }, - }, - "dispatcher_llm": { - "backend": "ollama", - "config": { - "model_name_or_path": "qwen3:1.7b", - "temperature": 0.0, - "remove_think_prefix": True, - "max_tokens": 8192, - }, - }, - "graph_db": { - "backend": "neo4j", - "config": { - "uri": "bolt://localhost:7687", - "user": "neo4j", - "password": "12345678", - "db_name": "user03alice11", - "auto_create": True, - }, - }, - "embedder": { - "backend": "ollama", - "config": { - "model_name_or_path": "nomic-embed-text:latest", - }, - }, - }, - }, - "act_mem": { - "backend": "kv_cache", - "config": { - "memory_filename": "activation_memory.pickle", - "extractor_llm": { - "backend": "huggingface", - "config": { - "model_name_or_path": "Qwen/Qwen3-1.7B", - "temperature": 0.8, - "max_tokens": 1024, - "top_p": 0.9, - "top_k": 50, - "add_generation_prompt": True, - "remove_think_prefix": False, - }, - }, - }, - }, - "para_mem": { - "backend": "lora", - "config": { - "memory_filename": "parametric_memory.adapter", - "extractor_llm": { - "backend": "huggingface", - "config": { - "model_name_or_path": "Qwen/Qwen3-1.7B", - "temperature": 0.8, - "max_tokens": 1024, - "top_p": 0.9, - "top_k": 50, - "add_generation_prompt": True, - "remove_think_prefix": False, - }, - }, - }, - }, - }, -) - -# 3. Initialize the MemCube with the configuration -mem_cube = GeneralMemCube(config) -try: - mem_cube.dump("/tmp/user03alice/mem_cube_5") -except Exception as e: - print(e) - -# 4. Register the MemCube explicitly -mos.register_mem_cube("/tmp/user03alice/mem_cube_5", "user03alice") - -# 5. add, get, search memory -mos.add(memory_content="I like playing football.") - -get_all_results = mos.get_all() -print(f"Get all results after add memory: {get_all_results}") - -# 6. add mesaages -messages = [ - {"role": "user", "content": "I like playing football."}, - {"role": "assistant", "content": "yes football is my favorite game."}, -] -mos.add(messages) -get_all_results = mos.get_all() -print(f"Get all results after add mesaages: {get_all_results}") - -# 6. add doc -mos.add(doc_path="./examples/data") -get_all_results = mos.get_all() -print(f"Get all results after add doc: {get_all_results}") - -search_results = mos.search(query="my favorite football game") -print(f"Search results: {search_results}") - -# .chat -while True: - user_input = input("👤 [You] ").strip() - print() - response = mos.chat(user_input) - print(f"🤖 [Assistant] {response}\n") -print("📢 [System] MemChat has stopped.") diff --git a/examples/mem_os/chat_w_scheduler.py b/examples/mem_os/chat_w_scheduler.py deleted file mode 100644 index 28c4c31a9..000000000 --- a/examples/mem_os/chat_w_scheduler.py +++ /dev/null @@ -1,74 +0,0 @@ -import shutil -import sys - -from pathlib import Path - -from memos.configs.mem_cube import GeneralMemCubeConfig -from memos.configs.mem_os import MOSConfig -from memos.configs.mem_scheduler import AuthConfig -from memos.mem_cube.general import GeneralMemCube -from memos.mem_os.main import MOS - - -FILE_PATH = Path(__file__).absolute() -BASE_DIR = FILE_PATH.parent.parent.parent -sys.path.insert(0, str(BASE_DIR)) # Enable execution from any working directory - - -# set configs -mos_config = MOSConfig.from_yaml_file( - f"{BASE_DIR}/examples/data/config/mem_scheduler/memos_config_w_scheduler.yaml" -) - -mem_cube_config = GeneralMemCubeConfig.from_yaml_file( - f"{BASE_DIR}/examples/data/config/mem_scheduler/mem_cube_config_neo4j.yaml" -) - -# default local graphdb uri -if AuthConfig.default_config_exists(): - auth_config = AuthConfig.from_local_config() - - mos_config.mem_reader.config.llm.config.api_key = auth_config.openai.api_key - mos_config.mem_reader.config.llm.config.api_base = auth_config.openai.base_url - - mem_cube_config.text_mem.config.graph_db.config.uri = auth_config.graph_db.uri - mem_cube_config.text_mem.config.graph_db.config.user = auth_config.graph_db.user - mem_cube_config.text_mem.config.graph_db.config.password = auth_config.graph_db.password - mem_cube_config.text_mem.config.graph_db.config.db_name = auth_config.graph_db.db_name - mem_cube_config.text_mem.config.graph_db.config.auto_create = auth_config.graph_db.auto_create - -# Initialization -mos = MOS(mos_config) - -user_id = "user_1" -mos.create_user(user_id) - -mem_cube_id = "mem_cube_5" -mem_cube_name_or_path = f"{BASE_DIR}/outputs/mem_scheduler/{user_id}/{mem_cube_id}" - -if Path(mem_cube_name_or_path).exists(): - shutil.rmtree(mem_cube_name_or_path) - print(f"{mem_cube_name_or_path} is not empty, and has been removed.") - -mem_cube = GeneralMemCube(mem_cube_config) -mem_cube.dump(mem_cube_name_or_path) -mos.register_mem_cube( - mem_cube_name_or_path=mem_cube_name_or_path, mem_cube_id=mem_cube_id, user_id=user_id -) - -messages = [ - {"role": "user", "content": "I like playing football."}, - {"role": "assistant", "content": "I like playing football too."}, -] -mos.add(messages, user_id=user_id, mem_cube_id=mem_cube_id) - - -while True: - user_input = input("👤 [You] ").strip() - print() - response = mos.chat(user_input, user_id=user_id) - retrieved_memories = mos.get_all(mem_cube_id=mem_cube_id, user_id=user_id) - print(f"🤖 [Assistant] {response}\n") - for node in retrieved_memories["text_mem"][0]["memories"]["nodes"]: - if node["metadata"]["memory_type"] == "WorkingMemory": - print(f"🤖 [Assistant]working mem : {node['memory']}\n") diff --git a/examples/mem_os/cot_usage_example.py b/examples/mem_os/cot_usage_example.py deleted file mode 100644 index e3d10fa75..000000000 --- a/examples/mem_os/cot_usage_example.py +++ /dev/null @@ -1,342 +0,0 @@ -#!/usr/bin/env python3 -""" -MemOS CoT (Chain of Thought) Usage Example -This example demonstrates how to use CoT functionality with tree textual memory. -It shows how to: -1. Decompose complex questions into sub-questions -2. Get answers for sub-questions using tree_textual_memory -3. Use JSON configuration files with environment variable overrides -""" - -import json -import os - -# Load environment variables -from dotenv import load_dotenv - -from memos.configs.llm import LLMConfigFactory -from memos.configs.mem_reader import SimpleStructMemReaderConfig -from memos.configs.memory import TreeTextMemoryConfig -from memos.mem_os.main import MOS -from memos.mem_reader.simple_struct import SimpleStructMemReader -from memos.memories.textual.tree import TreeTextMemory - - -load_dotenv() - - -def load_and_modify_config(config_path: str) -> dict: - """Load JSON config and modify it with environment variables.""" - with open(config_path) as f: - config = json.load(f) - - # Get environment variables - openai_api_key = os.getenv("OPENAI_API_KEY") - openai_base_url = os.getenv("OPENAI_API_BASE", "https://api.openai.com/v1") - - # Modify config to use ollama for embedder and gpt-4o-mini for LLMs - if "embedder" in config: - config["embedder"] = { - "backend": "ollama", - "config": {"model_name_or_path": "nomic-embed-text:latest"}, - } - - # Modify LLM configs to use gpt-4o-mini - if "llm" in config: - config["llm"] = { - "backend": "openai", - "config": { - "model_name_or_path": "gpt-4o-mini", - "api_key": openai_api_key, - "api_base": openai_base_url, - "temperature": 0.5, - "remove_think_prefix": True, - "max_tokens": 8192, - }, - } - - if "extractor_llm" in config: - config["extractor_llm"] = { - "backend": "openai", - "config": { - "model_name_or_path": "gpt-4o-mini", - "api_key": openai_api_key, - "api_base": openai_base_url, - "temperature": 0.5, - "remove_think_prefix": True, - "max_tokens": 8192, - }, - } - - if "dispatcher_llm" in config: - config["dispatcher_llm"] = { - "backend": "openai", - "config": { - "model_name_or_path": "gpt-4o-mini", - "api_key": openai_api_key, - "api_base": openai_base_url, - "temperature": 0.5, - "remove_think_prefix": True, - "max_tokens": 8192, - }, - } - - # Modify graph_db config if present - if "graph_db" in config: - neo4j_uri = os.getenv("NEO4J_URI", "bolt://localhost:7687") - neo4j_user = os.getenv("NEO4J_USER", "neo4j") - neo4j_password = os.getenv("NEO4J_PASSWORD", "12345678") - - config["graph_db"] = { - "backend": "neo4j", - "config": { - "uri": neo4j_uri, - "user": neo4j_user, - "password": neo4j_password, - "db_name": "testlcy", - "auto_create": True, - "embedding_dimension": 768, - }, - } - - return config - - -def setup_llm_config(): - """Setup LLM configuration for CoT operations.""" - # Get environment variables - openai_api_key = os.getenv("OPENAI_API_KEY") - openai_base_url = os.getenv("OPENAI_API_BASE", "https://api.openai.com/v1") - # Use ollama with gpt-4o-mini model - return LLMConfigFactory( - backend="openai", - config={ - "model_name_or_path": "gpt-4o-mini", - "api_key": openai_api_key, - "api_base": openai_base_url, - "temperature": 0.5, - "remove_think_prefix": True, - "max_tokens": 8192, - }, - ) - - -def create_tree_memory(): - """Create a tree textual memory with sample data.""" - print("Creating tree textual memory...") - - # Load and modify configurations - tree_config_dict = load_and_modify_config("examples/data/config/tree_config.json") - reader_config_dict = load_and_modify_config( - "examples/data/config/simple_struct_reader_config.json" - ) - - # Create config objects - tree_config = TreeTextMemoryConfig.model_validate(tree_config_dict) - reader_config = SimpleStructMemReaderConfig.model_validate(reader_config_dict) - - # Create tree memory - tree_memory = TreeTextMemory(tree_config) - tree_memory.delete_all() # Clear existing data - - # Create memory reader - reader = SimpleStructMemReader(reader_config) - - # Sample conversation data - sample_conversations = [ - [ - {"role": "user", "content": "Tell me about China and its capital."}, - { - "role": "assistant", - "content": "China is a country in East Asia. Beijing is its capital city.", - }, - {"role": "user", "content": "Who is Lang Ping?"}, - { - "role": "assistant", - "content": "Lang Ping is a famous Chinese volleyball coach and former player.", - }, - {"role": "user", "content": "What about Madagascar?"}, - { - "role": "assistant", - "content": "Madagascar is an island country in the Indian Ocean. It's known for its unique wildlife.", - }, - {"role": "user", "content": "Tell me about trade between China and Madagascar."}, - { - "role": "assistant", - "content": "China and Madagascar have developed trade relations, particularly in agriculture and mining.", - }, - {"role": "user", "content": "What about the essential oil industry in Madagascar?"}, - { - "role": "assistant", - "content": "The essential oil industry is growing in Madagascar, especially on Nosy Be Island where vanilla and ylang-ylang are produced.", - }, - ] - ] - - # Acquire memories using the reader - memories = reader.get_memory( - sample_conversations, type="chat", info={"user_id": "cot_user", "session_id": "cot_session"} - ) - - # Add memories to tree structure - for memory_list in memories: - tree_memory.add(memory_list) - - print("✓ Added sample conversations to tree memory") - return tree_memory - - -def cot_decompose(): - """Test the cot_decompose functionality.""" - print("\n=== Testing CoT Decomposition ===") - - # Setup LLM config - llm_config = setup_llm_config() - - # Test questions - test_questions = [ - "Who is the current head coach of the gymnastics team in the capital of the country that Lang Ping represents?", - "What is the weather like today?", - "How did the trade relationship between Madagascar and China develop, and how does this relationship affect the market expansion of the essential oil industry on Nosy Be Island?", - ] - - for i, question in enumerate(test_questions, 1): - print(f"\nTest {i}: {question}") - result = MOS.cot_decompose(question, llm_config) - print(f"✓ Decomposition result: {result}") - - if result.get("is_complex", False): - sub_questions = result.get("sub_questions", []) - print(f"✓ Found {len(sub_questions)} sub-questions:") - for j, sub_q in enumerate(sub_questions, 1): - print(f" {j}. {sub_q}") - else: - print("✓ Question is not complex, no decomposition needed.") - - return llm_config - - -def get_sub_answers_with_tree_memory(): - """Test get_sub_answers with tree textual memory.""" - print("\n=== Testing get_sub_answers with Tree Textual Memory ===") - - # Setup - llm_config = setup_llm_config() - tree_memory = create_tree_memory() - - # Test sub-questions - sub_questions = [ - "Which country does Lang Ping represent in volleyball?", - "What is the capital of this country?", - "Who is the current head coach of the gymnastics team in this capital?", - ] - - print("Sub-questions to answer:") - for i, q in enumerate(sub_questions, 1): - print(f" {i}. {q}") - print("\nGenerating answers using tree memory and LLM...") - sub_questions, sub_answers = MOS.get_sub_answers( - sub_questions=sub_questions, search_engine=tree_memory, llm_config=llm_config, top_k=3 - ) - - print("✓ Generated answers:") - for i, (question, answer) in enumerate(zip(sub_questions, sub_answers, strict=False), 1): - print(f"\n Sub-question {i}: {question}") - print(f" Answer: {answer}") - - -def complete_cot_workflow(): - """Test the complete CoT workflow from decomposition to final synthesis.""" - print("\n=== Testing Complete CoT Workflow ===") - - # Setup - llm_config = setup_llm_config() - tree_memory = create_tree_memory() - - # Complex question - complex_question = "How did the trade relationship between Madagascar and China develop, and how does this relationship affect the market expansion of the essential oil industry on Nosy Be Island?" - - print(f"Original question: {complex_question}") - - try: - # Step 1: Decompose the question - print("\n1. Decomposing question...") - decomposition_result = MOS.cot_decompose(complex_question, llm_config) - print(f"✓ Decomposition result: {decomposition_result}") - - if not decomposition_result.get("is_complex", False): - print("Question is not complex, no decomposition needed.") - return - - sub_questions = decomposition_result.get("sub_questions", []) - print(f"✓ Found {len(sub_questions)} sub-questions:") - for i, q in enumerate(sub_questions, 1): - print(f" {i}. {q}") - - # Step 2: Get answers for sub-questions - print("\n2. Getting answers for sub-questions...") - sub_questions, sub_answers = MOS.get_sub_answers( - sub_questions=sub_questions, search_engine=tree_memory, llm_config=llm_config, top_k=3 - ) - - print("✓ Generated answers:") - for i, (question, answer) in enumerate(zip(sub_questions, sub_answers, strict=False), 1): - print(f"\n Sub-question {i}: {question}") - print(f" Answer: {answer}") - - # Step 3: Generate final synthesis - print("\n3. Generating final synthesis...") - # Build the sub-questions and answers text - qa_text = "" - for i, (question, answer) in enumerate(zip(sub_questions, sub_answers, strict=False), 1): - qa_text += f"Q{i}: {question}\nA{i}: {answer}\n\n" - - synthesis_prompt = f"""You are an expert at synthesizing information from multiple sources to provide comprehensive answers. - -Sub-questions and their answers: -{qa_text} -Please synthesize these answers into a comprehensive response that: -1. Addresses the original question completely -2. Integrates information from all sub-questions -3. Provides clear reasoning and connections -4. Is well-structured and easy to understand - -Original question: {complex_question} - -Your response:""" - - # Generate final answer - from memos.llms.factory import LLMFactory - - llm = LLMFactory.from_config(llm_config) - messages = [ - { - "role": "system", - "content": "You are a helpful assistant that synthesizes information from multiple sources.", - }, - {"role": "user", "content": synthesis_prompt}, - ] - - final_answer = llm.generate(messages) - print(f"\n✓ Final synthesized answer:\n{final_answer}") - - except Exception as e: - print(f"✗ Error in complete workflow: {e}") - - -def main(): - """Main function to run the CoT example.""" - print("MemOS CoT (Chain of Thought) Usage Example") - print("=" * 60) - - # Run the examples - cot_decompose() - get_sub_answers_with_tree_memory() - complete_cot_workflow() - - print("\n" + "=" * 60) - print("✓ All examples completed successfully!") - - -if __name__ == "__main__": - main() diff --git a/examples/mem_os/easy_memos.py b/examples/mem_os/easy_memos.py deleted file mode 100644 index a2e8014ed..000000000 --- a/examples/mem_os/easy_memos.py +++ /dev/null @@ -1,21 +0,0 @@ -#!/usr/bin/env python3 -""" -Simple test script for MOS.simple() functionality. -""" - -import os - -from memos.mem_os.main import MOS - - -# Set environment variables for testing -os.environ["OPENAI_API_BASE"] = "http://xxxxxxxxx" -os.environ["OPENAI_API_KEY"] = "sk-xxxxxxxxxx" -os.environ["MOS_TEXT_MEM_TYPE"] = "general_text" # "tree_text" need set neo4j - - -memory = MOS.simple() -print("MOS.simple() works!") -memory.add(memory_content="my favorite color is blue") -print(memory.chat("what is my favorite color?")) -# Your favorite color is blue! diff --git a/examples/mem_os/locomo_shared_database_memos.py b/examples/mem_os/locomo_shared_database_memos.py deleted file mode 100644 index 97efe1fe6..000000000 --- a/examples/mem_os/locomo_shared_database_memos.py +++ /dev/null @@ -1,203 +0,0 @@ -import json -import os - -from dotenv import load_dotenv - -from memos import log -from memos.configs.mem_cube import GeneralMemCubeConfig -from memos.configs.mem_os import MOSConfig -from memos.mem_cube.general import GeneralMemCube -from memos.mem_os.product import MOSProduct - - -load_dotenv() - - -logger = log.get_logger(__name__) - - -# === Load conversation === -with open("evaluation/data/locomo/locomo10.json", encoding="utf-8") as f: - conversation = json.load(f) - data = conversation[3] - speaker_a = data["conversation"]["speaker_a"] - speaker_b = data["conversation"]["speaker_b"] - conversation_i = data["conversation"] - -db_name = "shared-db-locomo-case" - -openapi_config = { - "model_name_or_path": "gpt-4o-mini", - "temperature": 0.8, - "max_tokens": 1024, - "api_key": "your-api-key-here", - "api_base": "https://api.openai.com/v1", -} - - -# === Create MOS Config === -def get_user_configs(user_name): - mos_config = MOSConfig( - user_id=user_name, - chat_model={"backend": "openai", "config": openapi_config}, - mem_reader={ - "backend": "simple_struct", - "config": { - "llm": {"backend": "openai", "config": openapi_config}, - "embedder": { - "backend": "universal_api", - "config": { - "provider": "openai", - "api_key": openapi_config["api_key"], - "model_name_or_path": "text-embedding-3-large", - "base_url": openapi_config["api_base"], - }, - }, - "chunker": { - "backend": "sentence", - "config": { - "tokenizer_or_token_counter": "gpt2", - "chunk_size": 512, - "chunk_overlap": 128, - "min_sentences_per_chunk": 1, - }, - }, - }, - }, - enable_textual_memory=True, - enable_activation_memory=False, - enable_parametric_memory=False, - top_k=5, - max_turns_window=20, - ) - - return mos_config - - -# === Get Memory Cube Config === -def get_mem_cube_config(user_name): - neo4j_uri = os.getenv("NEO4J_URI", "bolt://localhost:7687") - neo4j_config = { - "uri": neo4j_uri, - "user": "neo4j", - "password": "12345678", - "db_name": db_name, - "user_name": "will be updated", - "use_multi_db": False, - "embedding_dimension": 3072, - "auto_create": True, - } - cube_config = GeneralMemCubeConfig.model_validate( - { - "user_id": user_name, - "cube_id": f"{user_name}_cube", - "text_mem": { - "backend": "tree_text", - "config": { - "extractor_llm": {"backend": "openai", "config": openapi_config}, - "dispatcher_llm": {"backend": "openai", "config": openapi_config}, - "graph_db": {"backend": "neo4j", "config": neo4j_config}, - "embedder": { - "backend": "universal_api", - "config": { - "provider": "openai", - "api_key": openapi_config["api_key"], - "model_name_or_path": "text-embedding-3-large", - "base_url": openapi_config["api_base"], - }, - }, - "reorganize": True, - }, - }, - } - ) - - mem_cube = GeneralMemCube(cube_config) - return mem_cube - - -# === Initialize MOSProduct === -root_config = get_user_configs(user_name="system") -mos_product = MOSProduct(default_config=root_config) - - -# === Register both users === -users = {} -for speaker in [speaker_a, speaker_b]: - user_id = speaker.lower() + "_test" - config = get_user_configs(user_id) - mem_cube = get_mem_cube_config(user_id) - result = mos_product.user_register( - user_id=user_id, - user_name=speaker, - interests=f"I'm {speaker}", - default_mem_cube=mem_cube, - ) - users[speaker] = {"user_id": user_id, "default_cube_id": result["default_cube_id"]} - print(f"✅ Registered: {speaker} -> {result}") - -# === Process conversation, add to both roles === -i = 1 -MAX_CONVERSATION_FOR_TEST = 3 -while ( - f"session_{i}_date_time" in conversation_i and f"session_{i}" in conversation_i -) and i < MAX_CONVERSATION_FOR_TEST: - session_i = conversation_i[f"session_{i}"] - session_time = conversation_i[f"session_{i}_date_time"] - - print(f"\n=== Processing Session {i} | Time: {session_time} ===") - - role1_msgs, role2_msgs = [], [] - - for m in session_i: - if m["speaker"] == speaker_a: - role1_msgs.append( - { - "role": "user", - "content": f"{m['speaker']}:{m['text']}", - "chat_time": session_time, - } - ) - role2_msgs.append( - { - "role": "assistant", - "content": f"{m['speaker']}:{m['text']}", - "chat_time": session_time, - } - ) - elif m["speaker"] == speaker_b: - role1_msgs.append( - { - "role": "assistant", - "content": f"{m['speaker']}:{m['text']}", - "chat_time": session_time, - } - ) - role2_msgs.append( - { - "role": "user", - "content": f"{m['speaker']}:{m['text']}", - "chat_time": session_time, - } - ) - - print(f"\n[Session {i}] {speaker_a} will add {len(role1_msgs)} messages.") - print(f"[Session {i}] {speaker_b} will add {len(role2_msgs)} messages.") - - mos_product.add( - user_id=users[speaker_a]["user_id"], - messages=role1_msgs, - mem_cube_id=users[speaker_a]["default_cube_id"], - ) - mos_product.add( - user_id=users[speaker_b]["user_id"], - messages=role2_msgs, - mem_cube_id=users[speaker_b]["default_cube_id"], - ) - - print(f"[Session {i}] Added messages for both roles") - - i += 1 - -print("\n✅ All messages added for both roles.\n") -mos_product.mem_reorganizer_off() diff --git a/examples/mem_os/multi_user_memos_example.py b/examples/mem_os/multi_user_memos_example.py deleted file mode 100644 index ac7e68610..000000000 --- a/examples/mem_os/multi_user_memos_example.py +++ /dev/null @@ -1,162 +0,0 @@ -""" -Example demonstrating how to use MOSProduct for multi-user scenarios. -""" - -import os - -from memos.configs.mem_cube import GeneralMemCubeConfig -from memos.configs.mem_os import MOSConfig -from memos.mem_cube.general import GeneralMemCube -from memos.mem_os.product import MOSProduct - - -def get_config(user_name): - openapi_config = { - "model_name_or_path": "gpt-4o-mini", - "temperature": 0.8, - "max_tokens": 1024, - "top_p": 0.9, - "top_k": 50, - "remove_think_prefix": True, - "api_key": os.getenv("OPENAI_API_KEY"), - "api_base": os.getenv("OPENAI_API_BASE"), - } - # Create a default configuration - default_config = MOSConfig( - user_id="root", - chat_model={"backend": "openai", "config": openapi_config}, - mem_reader={ - "backend": "simple_struct", - "config": { - "llm": { - "backend": "openai", - "config": openapi_config, - }, - "embedder": { - "backend": "universal_api", - "config": { - "provider": os.getenv("MOS_EMBEDDER_PROVIDER", "openai"), - "api_key": os.getenv("MOS_EMBEDDER_API_KEY", "sk-xxxx"), - "model_name_or_path": os.getenv( - "MOS_EMBEDDER_MODEL", "text-embedding-3-large" - ), - "base_url": os.getenv("MOS_EMBEDDER_API_BASE", "http://openai.com"), - }, - }, - "chunker": { - "backend": "sentence", - "config": { - "tokenizer_or_token_counter": "gpt2", - "chunk_size": 512, - "chunk_overlap": 128, - "min_sentences_per_chunk": 1, - }, - }, - }, - }, - user_manager={ - "backend": "mysql", - "config": { - "host": os.getenv("MYSQL_HOST", "localhost"), - "port": int(os.getenv("MYSQL_PORT", "3306")), - "username": os.getenv("MYSQL_USERNAME", "root"), - "password": os.getenv("MYSQL_PASSWORD", "12345678"), - "database": os.getenv("MYSQL_DATABASE", "memos_users"), - "charset": os.getenv("MYSQL_CHARSET", "utf8mb4"), - }, - }, - enable_textual_memory=True, - enable_activation_memory=False, - top_k=5, - max_turns_window=20, - ) - default_cube_config = GeneralMemCubeConfig.model_validate( - { - "user_id": user_name, - "cube_id": f"{user_name}_default_cube", - "text_mem": { - "backend": "tree_text", - "config": { - "extractor_llm": {"backend": "openai", "config": openapi_config}, - "dispatcher_llm": {"backend": "openai", "config": openapi_config}, - "graph_db": { - "backend": "neo4j", - "config": { - "uri": os.getenv("NEO4J_URI", "bolt://localhost:7687"), - "user": os.getenv("NEO4J_USER", "neo4j"), - "password": os.getenv("NEO4J_PASSWORD", "12345678"), - "db_name": os.getenv( - "NEO4J_DB_NAME", "shared-tree-textual-memory-test" - ), - "user_name": f"memos{user_name.replace('-', '')}", - "embedding_dimension": int(os.getenv("EMBEDDING_DIMENSION", 768)), - "use_multi_db": False, - "auto_create": True, - }, - }, - "embedder": { - "backend": "universal_api", - "config": { - "provider": os.getenv("MOS_EMBEDDER_PROVIDER", "openai"), - "api_key": os.getenv("MOS_EMBEDDER_API_KEY", "sk-xxxx"), - "model_name_or_path": os.getenv( - "MOS_EMBEDDER_MODEL", "text-embedding-3-large" - ), - "base_url": os.getenv("MOS_EMBEDDER_API_BASE", "http://openai.com"), - }, - }, - }, - }, - "act_mem": {}, - "para_mem": {}, - } - ) - default_mem_cube = GeneralMemCube(default_cube_config) - return default_config, default_mem_cube - - -def main(): - default_config, default_mem_cube = get_config(user_name="alice") - # Initialize MOSProduct with default config - mos_product = MOSProduct(default_config=default_config) - - # Register first user with default config - result1 = mos_product.user_register( - user_id="alice", - user_name="alice", - interests="I'm interested in machine learning and AI research.", - default_mem_cube=default_mem_cube, - ) - print(f"User registration result: {result1}") - - # Chat with Alice - print("\n=== Chatting with Alice ===") - for response_chunk in mos_product.chat(query="What are my interests?", user_id="alice"): - print(response_chunk, end="") - - # Add memory for Alice - mos_product.add( - user_id="alice", - memory_content="I attended a machine learning conference last week.", - mem_cube_id=result1["default_cube_id"], - ) - - # Search memories for Alice - search_result = mos_product.search(query="conference", user_id="alice") - print(f"\nSearch result for Alice: {search_result}") - - # Search memories for Alice - search_result = mos_product.get_all(user_id="alice", memory_type="text_mem") - print(f"\nSearch result for Alice: {search_result}") - - # List all users - users = mos_product.list_users() - print(f"\nAll registered users: {users}") - - # Get user info - alice_info = mos_product.get_user_info("alice") - print(f"\nAlice's info: {alice_info}") - - -if __name__ == "__main__": - main() diff --git a/examples/mem_os/naive_memos.py b/examples/mem_os/naive_memos.py deleted file mode 100644 index 5d0039bb9..000000000 --- a/examples/mem_os/naive_memos.py +++ /dev/null @@ -1,30 +0,0 @@ -import uuid - -from memos.configs.mem_os import MOSConfig -from memos.mem_os.main import MOS - - -# init MOS -mos_config = MOSConfig.from_json_file("examples/data/config/simple_memos_config.json") -memory = MOS(mos_config) - -# create user -user_id = str(uuid.uuid4()) -memory.create_user(user_id=user_id) - -# register cube for user -memory.register_mem_cube("examples/data/mem_cube_2", user_id=user_id) - -# add memory for user -memory.add( - messages=[ - {"role": "user", "content": "I like playing football."}, - {"role": "assistant", "content": "I like playing football too."}, - ], - user_id=user_id, -) - -# Later, when you want to retrieve memory for user -retrieved_memories = memory.search(query="What do you like?", user_id=user_id) -# output text_memories: I like playing football, act_memories, para_memories -print(f"text_memories: {retrieved_memories['text_mem']}") diff --git a/examples/mem_os/persistent_memos_example.py b/examples/mem_os/persistent_memos_example.py deleted file mode 100644 index 16353be6a..000000000 --- a/examples/mem_os/persistent_memos_example.py +++ /dev/null @@ -1,192 +0,0 @@ -""" -Example demonstrating persistent user management in MemOS. - -This example shows how to use the PersistentUserManager to maintain -user configurations across service restarts. -""" - -import os -import tempfile - -from memos.configs.mem_os import MOSConfig -from memos.mem_os.product import MOSProduct -from memos.mem_user.persistent_user_manager import PersistentUserManager, UserRole - - -def create_sample_config(user_id: str) -> MOSConfig: - """Create a sample configuration for a user.""" - return MOSConfig( - user_id=user_id, - chat_model={ - "backend": "openai", - "config": { - "model_name_or_path": "gpt-3.5-turbo", - "api_key": "your-api-key-here", - "temperature": 0.7, - }, - }, - mem_reader={ - "backend": "naive", - "config": { - "llm": { - "backend": "openai", - "config": { - "model_name_or_path": "gpt-3.5-turbo", - "api_key": "your-api-key-here", - }, - }, - "embedder": { - "backend": "ollama", - "config": { - "model_name_or_path": "nomic-embed-text:latest", - }, - }, - }, - }, - enable_textual_memory=True, - enable_activation_memory=False, - top_k=5, - max_turns_window=20, - ) - - -def demonstrate_persistence(): - """Demonstrate the persistence functionality.""" - print("=== MemOS Persistent User Management Demo ===\n") - - # Create a temporary database for this demo - temp_dir = tempfile.mkdtemp() - db_path = os.path.join(temp_dir, "demo_memos.db") - - try: - # Step 1: Create a persistent user manager - print("1. Creating PersistentUserManager...") - user_manager = PersistentUserManager(db_path=db_path) - print(f" Database created at: {db_path}") - - # Step 2: Create some sample configurations - print("\n2. Creating sample user configurations...") - user_configs = {} - for i in range(3): - user_id = f"user_{i + 1}" - user_name = f"User {i + 1}" - config = create_sample_config(user_id) - user_configs[user_id] = config - - # Create user with configuration - created_id = user_manager.create_user_with_config( - user_name, config, UserRole.USER, user_id - ) - print(f" Created user: {user_name} (ID: {created_id})") - - # Step 3: Verify configurations are saved - print("\n3. Verifying configurations are saved...") - for user_id in user_configs: - config = user_manager.get_user_config(user_id) - if config: - print(f" ✓ Configuration found for {user_id}") - print(f" - Textual memory enabled: {config.enable_textual_memory}") - print(f" - Top-k: {config.top_k}") - else: - print(f" ✗ Configuration not found for {user_id}") - - # Step 4: Simulate service restart by creating a new manager instance - print("\n4. Simulating service restart...") - print(" Creating new PersistentUserManager instance...") - new_user_manager = PersistentUserManager(db_path=db_path) - - # Step 5: Verify configurations are restored - print("\n5. Verifying configurations are restored after restart...") - for user_id in user_configs: - config = new_user_manager.get_user_config(user_id) - if config: - print(f" ✓ Configuration restored for {user_id}") - else: - print(f" ✗ Configuration not restored for {user_id}") - - # Step 6: Create MOSProduct and demonstrate restoration - print("\n6. Creating MOSProduct with persistent user manager...") - default_config = create_sample_config("default_user") - mos_product = MOSProduct(default_config=default_config) - - # The MOSProduct should automatically restore user instances - print(f" Active user instances: {len(mos_product.user_instances)}") - for user_id in mos_product.user_instances: - print(f" - {user_id}") - - # Step 7: Demonstrate configuration update - print("\n7. Demonstrating configuration update...") - user_id = "user_1" - original_config = user_manager.get_user_config(user_id) - if original_config: - # Update configuration - updated_config = original_config.model_copy(deep=True) - updated_config.top_k = 10 - updated_config.enable_activation_memory = True - - success = user_manager.save_user_config(user_id, updated_config) - if success: - print(f" ✓ Updated configuration for {user_id}") - print(f" - New top-k: {updated_config.top_k}") - print(f" - Activation memory: {updated_config.enable_activation_memory}") - else: - print(f" ✗ Failed to update configuration for {user_id}") - - # Step 8: List all configurations - print("\n8. Listing all user configurations...") - all_configs = user_manager.list_user_configs() - print(f" Total configurations: {len(all_configs)}") - for user_id, config in all_configs.items(): - print( - f" - {user_id}: top_k={config.top_k}, textual_memory={config.enable_textual_memory}" - ) - - print("\n=== Demo completed successfully! ===") - print(f"Database file: {db_path}") - print("You can inspect this file to see the persistent data.") - - except Exception as e: - print(f"Error during demo: {e}") - raise - finally: - # Cleanup - if os.path.exists(db_path): - os.remove(db_path) - if os.path.exists(temp_dir): - os.rmdir(temp_dir) - - -def demonstrate_api_usage(): - """Demonstrate how the API would work with persistence.""" - print("\n=== API Usage Example ===") - print(""" - With the new persistent system, your API calls would work like this: - - 1. Register a user (configuration is automatically saved): - POST /product/users/register - { - "user_id": "john_doe", - "user_name": "John Doe", - "interests": "AI, machine learning, programming" - } - - 2. Get user configuration: - GET /product/users/john_doe/config - - 3. Update user configuration: - PUT /product/users/john_doe/config - { - "user_id": "john_doe", - "enable_activation_memory": true, - "top_k": 10, - ... - } - - 4. After service restart, all user instances are automatically restored - and the user can immediately use the system without re-registration. - """) - - -if __name__ == "__main__": - demonstrate_persistence() - demonstrate_api_usage() diff --git a/examples/mem_os/simple_memos.py b/examples/mem_os/simple_memos.py deleted file mode 100644 index be765e837..000000000 --- a/examples/mem_os/simple_memos.py +++ /dev/null @@ -1,42 +0,0 @@ -from memos.configs.mem_os import MOSConfig -from memos.mem_os.main import MOS - - -# init MOSConfig -mos_config = MOSConfig.from_json_file("examples/data/config/simple_memos_config.json") -mos = MOS(mos_config) - -# create user id -user_id = "lcy1" -mos.create_user(user_id=user_id) -users = mos.list_users() -print("\nAll users:") -for user in users: - print(f" - {user['user_name']} ({user['user_id']}) - Role: {user['role']}") - - -# load exist mem_cube from local -mos.register_mem_cube("examples/data/mem_cube_2", user_id=user_id) - -mos.add(memory_content="I like playing football.", user_id=user_id) - -get_all_results = mos.get_all(user_id=user_id) -print(f"Get all results for user : {get_all_results}") - -get_results = mos.get( - mem_cube_id="examples/data/mem_cube_2", - memory_id=get_all_results["text_mem"][0]["memories"][0].id, - user_id=user_id, -) -print(f"Get memories for user : {get_results}") - -search_results = mos.search(query="my favorite football game", user_id=user_id) -print(f"Search results for user : {search_results}") - - -while True: - user_input = input("👤 [You] ").strip() - print() - response = mos.chat(user_input, user_id=user_id) - print(f"🤖 [Assistant] {response}\n") -print("📢 [System] MemChat has stopped.") diff --git a/examples/mem_os/simple_openapi_memos.py b/examples/mem_os/simple_openapi_memos.py deleted file mode 100644 index a9faa3dc8..000000000 --- a/examples/mem_os/simple_openapi_memos.py +++ /dev/null @@ -1,300 +0,0 @@ -import os -import time -import uuid - -from datetime import datetime - -from dotenv import load_dotenv - -from memos.configs.mem_cube import GeneralMemCubeConfig -from memos.configs.mem_os import MOSConfig -from memos.mem_cube.general import GeneralMemCube -from memos.mem_os.main import MOS - - -load_dotenv() - -# 1. Create MOS Config and set openai config -print(f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to create MOS configuration...") -start_time = time.time() - -user_name = str(uuid.uuid4()) -print(user_name) - -# 1.1 Set openai config -openapi_config = { - "model_name_or_path": "gpt-4o-mini", - "temperature": 0.8, - "max_tokens": 1024, - "top_p": 0.9, - "top_k": 50, - "remove_think_prefix": True, - "api_key": os.getenv("OPENAI_API_KEY", "sk-xxxxx"), - "api_base": os.getenv("OPENAI_API_BASE", "https://api.openai.com/v1"), -} -# 1.2 Set neo4j config -neo4j_uri = os.getenv("NEO4J_URI", "bolt://localhost:7687") - -# 1.3 Create MOS Config -config = { - "user_id": user_name, - "chat_model": { - "backend": "openai", - "config": openapi_config, - }, - "mem_reader": { - "backend": "simple_struct", - "config": { - "llm": { - "backend": "openai", - "config": openapi_config, - }, - "embedder": { - "backend": "ollama", - "config": { - "model_name_or_path": "nomic-embed-text:latest", - }, - }, - "chunker": { - "backend": "sentence", - "config": { - "tokenizer_or_token_counter": "gpt2", - "chunk_size": 512, - "chunk_overlap": 128, - "min_sentences_per_chunk": 1, - }, - }, - }, - }, - "max_turns_window": 20, - "top_k": 5, - "enable_textual_memory": True, - "enable_activation_memory": False, - "enable_parametric_memory": False, -} - -mos_config = MOSConfig(**config) -# you can set PRO_MODE to True to enable CoT enhancement mos_config.PRO_MODE = True -mos = MOS(mos_config) - -print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] MOS configuration created successfully, time elapsed: {time.time() - start_time:.2f}s\n" -) - -# 2. Initialize memory cube -print(f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to initialize MemCube configuration...") -start_time = time.time() - -config = GeneralMemCubeConfig.model_validate( - { - "user_id": user_name, - "cube_id": f"{user_name}", - "text_mem": { - "backend": "tree_text", - "config": { - "extractor_llm": { - "backend": "openai", - "config": openapi_config, - }, - "dispatcher_llm": { - "backend": "openai", - "config": openapi_config, - }, - "graph_db": { - "backend": "neo4j", - "config": { - "uri": neo4j_uri, - "user": "neo4j", - "password": "12345678", - "db_name": f"db{user_name.replace('-', '')}", - "auto_create": True, - }, - }, - "embedder": { - "backend": "ollama", - "config": { - "model_name_or_path": "nomic-embed-text:latest", - }, - }, - }, - }, - "act_mem": {}, - "para_mem": {}, - }, -) - -print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] MemCube configuration initialization completed, time elapsed: {time.time() - start_time:.2f}s\n" -) - -# 3. Initialize the MemCube with the configuration -print(f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to create MemCube instance...") -start_time = time.time() - -mem_cube = GeneralMemCube(config) -try: - mem_cube.dump(f"/tmp/{user_name}/") - print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] MemCube created and saved successfully, time elapsed: {time.time() - start_time:.2f}s\n" - ) -except Exception as e: - print( - f"❌ [{datetime.now().strftime('%H:%M:%S')}] MemCube save failed: {e}, time elapsed: {time.time() - start_time:.2f}s\n" - ) - -# 4. Register the MemCube -print(f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to register MemCube...") -start_time = time.time() - -mos.register_mem_cube(f"/tmp/{user_name}", mem_cube_id=user_name) - -print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] MemCube registration completed, time elapsed: {time.time() - start_time:.2f}s\n" -) - -# 5. Add, get, search memory -print(f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to add single memory...") -start_time = time.time() - -mos.add(memory_content="I like playing football.") - -print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] Single memory added successfully, time elapsed: {time.time() - start_time:.2f}s" -) - -print(f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to get all memories...") -start_time = time.time() - -get_all_results = mos.get_all() - - -# Filter out embedding fields, keeping only necessary fields -def filter_memory_data(memories_data): - filtered_data = {} - for key, value in memories_data.items(): - if key == "text_mem": - filtered_data[key] = [] - for mem_group in value: - # Check if it's the new data structure (list of TextualMemoryItem objects) - if "memories" in mem_group and isinstance(mem_group["memories"], list): - # New data structure: directly a list of TextualMemoryItem objects - filtered_memories = [] - for memory_item in mem_group["memories"]: - # Create filtered dictionary - filtered_item = { - "id": memory_item.id, - "memory": memory_item.memory, - "metadata": {}, - } - # Filter metadata, excluding embedding - if hasattr(memory_item, "metadata") and memory_item.metadata: - for attr_name in dir(memory_item.metadata): - if not attr_name.startswith("_") and attr_name != "embedding": - attr_value = getattr(memory_item.metadata, attr_name) - if not callable(attr_value): - filtered_item["metadata"][attr_name] = attr_value - filtered_memories.append(filtered_item) - - filtered_group = { - "cube_id": mem_group.get("cube_id", ""), - "memories": filtered_memories, - } - filtered_data[key].append(filtered_group) - else: - # Old data structure: dictionary with nodes and edges - filtered_group = { - "memories": {"nodes": [], "edges": mem_group["memories"].get("edges", [])} - } - for node in mem_group["memories"].get("nodes", []): - filtered_node = { - "id": node.get("id"), - "memory": node.get("memory"), - "metadata": { - k: v - for k, v in node.get("metadata", {}).items() - if k != "embedding" - }, - } - filtered_group["memories"]["nodes"].append(filtered_node) - filtered_data[key].append(filtered_group) - else: - filtered_data[key] = value - return filtered_data - - -filtered_results = filter_memory_data(get_all_results) -print(f"Get all results after add memory: {filtered_results['text_mem'][0]['memories']}") - -print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] Get all memories completed, time elapsed: {time.time() - start_time:.2f}s\n" -) - -# 6. Add messages -print(f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to add conversation messages...") -start_time = time.time() - -messages = [ - {"role": "user", "content": "I like playing football."}, - {"role": "assistant", "content": "yes football is my favorite game."}, -] -mos.add(messages) - -print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] Conversation messages added successfully, time elapsed: {time.time() - start_time:.2f}s" -) - -print( - f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to get all memories (after adding messages)..." -) -start_time = time.time() - -get_all_results = mos.get_all() -filtered_results = filter_memory_data(get_all_results) -print(f"Get all results after add messages: {filtered_results}") - -print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] Get all memories completed, time elapsed: {time.time() - start_time:.2f}s\n" -) - -# 7. Add document -print(f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to add document...") -start_time = time.time() -## 7.1 add pdf for ./tmp/data if use doc mem mos.add(doc_path="./tmp/data/") -start_time = time.time() - -get_all_results = mos.get_all() -filtered_results = filter_memory_data(get_all_results) -print(f"Get all results after add doc: {filtered_results}") - -print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] Get all memories completed, time elapsed: {time.time() - start_time:.2f}s\n" -) - -# 8. Search -print(f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to search memories...") -start_time = time.time() - -search_results = mos.search(query="my favorite football game", user_id=user_name) -filtered_search_results = filter_memory_data(search_results) -print(f"Search results: {filtered_search_results}") - -print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] Memory search completed, time elapsed: {time.time() - start_time:.2f}s\n" -) - -# 9. Chat -print(f"🎯 [{datetime.now().strftime('%H:%M:%S')}] Starting chat mode...") -while True: - user_input = input("👤 [You] ").strip() - if user_input.lower() in ["quit", "exit"]: - break - - print() - chat_start_time = time.time() - response = mos.chat(user_input) - chat_duration = time.time() - chat_start_time - - print(f"🤖 [Assistant] {response}") - print(f"⏱️ [Response time: {chat_duration:.2f}s]\n") - -print("📢 [System] MemChat has stopped.") diff --git a/examples/mem_os/simple_openapi_memos_neo4j_community.py b/examples/mem_os/simple_openapi_memos_neo4j_community.py deleted file mode 100644 index aad1b8c77..000000000 --- a/examples/mem_os/simple_openapi_memos_neo4j_community.py +++ /dev/null @@ -1,315 +0,0 @@ -import os -import time -import uuid - -from datetime import datetime - -from dotenv import load_dotenv - -from memos.configs.mem_cube import GeneralMemCubeConfig -from memos.configs.mem_os import MOSConfig -from memos.mem_cube.general import GeneralMemCube -from memos.mem_os.main import MOS - - -load_dotenv() - -# 1. Create MOS Config and set openai config -print(f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to create MOS configuration...") -start_time = time.time() - -user_name = str(uuid.uuid4()) -print(user_name) - -# 1.1 Set openai config -openapi_config = { - "model_name_or_path": "gpt-4o-mini", - "temperature": 0.8, - "max_tokens": 1024, - "top_p": 0.9, - "top_k": 50, - "remove_think_prefix": True, - "api_key": os.getenv("OPENAI_API_KEY", "sk-xxxxx"), - "api_base": os.getenv("OPENAI_API_BASE", "https://api.openai.com/v1"), -} -embedder_config = { - "backend": "universal_api", - "config": { - "provider": "openai", - "api_key": os.getenv("OPENAI_API_KEY", "sk-xxxxx"), - "model_name_or_path": "text-embedding-3-large", - "base_url": os.getenv("OPENAI_API_BASE", "https://api.openai.com/v1"), - }, -} -EMBEDDING_DIMENSION = 3072 - -# 1.2 Set neo4j config -neo4j_uri = os.getenv("NEO4J_URI", "bolt://localhost:7687") - -# 1.3 Create MOS Config -config = { - "user_id": user_name, - "chat_model": { - "backend": "openai", - "config": openapi_config, - }, - "mem_reader": { - "backend": "simple_struct", - "config": { - "llm": { - "backend": "openai", - "config": openapi_config, - }, - "embedder": embedder_config, - "chunker": { - "backend": "sentence", - "config": { - "tokenizer_or_token_counter": "gpt2", - "chunk_size": 512, - "chunk_overlap": 128, - "min_sentences_per_chunk": 1, - }, - }, - }, - }, - "max_turns_window": 20, - "top_k": 5, - "enable_textual_memory": True, - "enable_activation_memory": False, - "enable_parametric_memory": False, -} - -mos_config = MOSConfig(**config) -# you can set PRO_MODE to True to enable CoT enhancement mos_config.PRO_MODE = True -mos = MOS(mos_config) - -print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] MOS configuration created successfully, time elapsed: {time.time() - start_time:.2f}s\n" -) - -# 2. Initialize memory cube -print(f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to initialize MemCube configuration...") -start_time = time.time() - -config = GeneralMemCubeConfig.model_validate( - { - "user_id": user_name, - "cube_id": f"{user_name}", - "text_mem": { - "backend": "tree_text", - "config": { - "extractor_llm": { - "backend": "openai", - "config": openapi_config, - }, - "dispatcher_llm": { - "backend": "openai", - "config": openapi_config, - }, - "embedder": embedder_config, - "graph_db": { - "backend": "neo4j-community", - "config": { - "uri": neo4j_uri, - "user": "neo4j", - "password": "12345678", - "db_name": "neo4j", - "user_name": "alice", - "use_multi_db": False, - "auto_create": False, - "embedding_dimension": EMBEDDING_DIMENSION, - "vec_config": { - "backend": "qdrant", - "config": { - "collection_name": "neo4j_vec_db", - "vector_dimension": EMBEDDING_DIMENSION, - "distance_metric": "cosine", - "host": "localhost", - "port": 6333, - }, - }, - }, - }, - "reorganize": True, - }, - }, - "act_mem": {}, - "para_mem": {}, - }, -) - -print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] MemCube configuration initialization completed, time elapsed: {time.time() - start_time:.2f}s\n" -) - -# 3. Initialize the MemCube with the configuration -print(f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to create MemCube instance...") -start_time = time.time() - -mem_cube = GeneralMemCube(config) -try: - mem_cube.dump(f"/tmp/{user_name}/") - print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] MemCube created and saved successfully, time elapsed: {time.time() - start_time:.2f}s\n" - ) -except Exception as e: - print( - f"❌ [{datetime.now().strftime('%H:%M:%S')}] MemCube save failed: {e}, time elapsed: {time.time() - start_time:.2f}s\n" - ) - -# 4. Register the MemCube -print(f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to register MemCube...") -start_time = time.time() - -mos.register_mem_cube(f"/tmp/{user_name}", mem_cube_id=user_name) - -print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] MemCube registration completed, time elapsed: {time.time() - start_time:.2f}s\n" -) - -# 5. Add, get, search memory -print(f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to add single memory...") -start_time = time.time() - -mos.add(memory_content="I like playing football.") - -print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] Single memory added successfully, time elapsed: {time.time() - start_time:.2f}s" -) - -print(f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to get all memories...") -start_time = time.time() - -get_all_results = mos.get_all() - - -# Filter out embedding fields, keeping only necessary fields -def filter_memory_data(memories_data): - filtered_data = {} - for key, value in memories_data.items(): - if key == "text_mem": - filtered_data[key] = [] - for mem_group in value: - # Check if it's the new data structure (list of TextualMemoryItem objects) - if "memories" in mem_group and isinstance(mem_group["memories"], list): - # New data structure: directly a list of TextualMemoryItem objects - filtered_memories = [] - for memory_item in mem_group["memories"]: - # Create filtered dictionary - filtered_item = { - "id": memory_item.id, - "memory": memory_item.memory, - "metadata": {}, - } - # Filter metadata, excluding embedding - if hasattr(memory_item, "metadata") and memory_item.metadata: - for attr_name in dir(memory_item.metadata): - if not attr_name.startswith("_") and attr_name != "embedding": - attr_value = getattr(memory_item.metadata, attr_name) - if not callable(attr_value): - filtered_item["metadata"][attr_name] = attr_value - filtered_memories.append(filtered_item) - - filtered_group = { - "cube_id": mem_group.get("cube_id", ""), - "memories": filtered_memories, - } - filtered_data[key].append(filtered_group) - else: - # Old data structure: dictionary with nodes and edges - filtered_group = { - "memories": {"nodes": [], "edges": mem_group["memories"].get("edges", [])} - } - for node in mem_group["memories"].get("nodes", []): - filtered_node = { - "id": node.get("id"), - "memory": node.get("memory"), - "metadata": { - k: v - for k, v in node.get("metadata", {}).items() - if k != "embedding" - }, - } - filtered_group["memories"]["nodes"].append(filtered_node) - filtered_data[key].append(filtered_group) - else: - filtered_data[key] = value - return filtered_data - - -filtered_results = filter_memory_data(get_all_results) -print(f"Get all results after add memory: {filtered_results['text_mem'][0]['memories']}") - -print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] Get all memories completed, time elapsed: {time.time() - start_time:.2f}s\n" -) - -# 6. Add messages -print(f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to add conversation messages...") -start_time = time.time() - -messages = [ - {"role": "user", "content": "I like playing football."}, - {"role": "assistant", "content": "yes football is my favorite game."}, -] -mos.add(messages) - -print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] Conversation messages added successfully, time elapsed: {time.time() - start_time:.2f}s" -) - -print( - f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to get all memories (after adding messages)..." -) -start_time = time.time() - -get_all_results = mos.get_all() -filtered_results = filter_memory_data(get_all_results) -print(f"Get all results after add messages: {filtered_results}") - -print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] Get all memories completed, time elapsed: {time.time() - start_time:.2f}s\n" -) - -# 7. Add document -print(f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to add document...") -start_time = time.time() -## 7.1 add pdf for ./tmp/data if use doc mem mos.add(doc_path="./tmp/data/") -start_time = time.time() - -get_all_results = mos.get_all() -filtered_results = filter_memory_data(get_all_results) -print(f"Get all results after add doc: {filtered_results}") - -print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] Get all memories completed, time elapsed: {time.time() - start_time:.2f}s\n" -) - -# 8. Search -print(f"🚀 [{datetime.now().strftime('%H:%M:%S')}] Starting to search memories...") -start_time = time.time() - -search_results = mos.search(query="my favorite football game", user_id=user_name) -filtered_search_results = filter_memory_data(search_results) -print(f"Search results: {filtered_search_results}") - -print( - f"✅ [{datetime.now().strftime('%H:%M:%S')}] Memory search completed, time elapsed: {time.time() - start_time:.2f}s\n" -) - -# 9. Chat -print(f"🎯 [{datetime.now().strftime('%H:%M:%S')}] Starting chat mode...") -while True: - user_input = input("👤 [You] ").strip() - if user_input.lower() in ["quit", "exit"]: - break - - print() - chat_start_time = time.time() - response = mos.chat(user_input) - chat_duration = time.time() - chat_start_time - - print(f"🤖 [Assistant] {response}") - print(f"⏱️ [Response time: {chat_duration:.2f}s]\n") - -print("📢 [System] MemChat has stopped.") diff --git a/examples/mem_os/simple_prefs_memos_product.py b/examples/mem_os/simple_prefs_memos_product.py deleted file mode 100644 index 40ec920f5..000000000 --- a/examples/mem_os/simple_prefs_memos_product.py +++ /dev/null @@ -1,399 +0,0 @@ -from memos.configs.mem_cube import GeneralMemCubeConfig -from memos.configs.mem_os import MOSConfig -from memos.mem_cube.general import GeneralMemCube -from memos.mem_os.product import MOSProduct - - -def get_config(user_id: str): - llm_config = { - "backend": "openai", - "config": { - "model_name_or_path": "gpt-4o-mini", - "api_key": "sk-xxxxx", - "api_base": "http://xxxx/v1", - "temperature": 0.1, - "remove_think_prefix": True, - "max_tokens": 4096, - }, - } - - embedder_config = { - "backend": "ollama", - "config": {"model_name_or_path": "nomic-embed-text:latest"}, - } - - # init MOS - mos_config = { - "user_id": user_id, - "chat_model": llm_config, - "mem_reader": { - "backend": "simple_struct", - "config": { - "llm": llm_config, - "embedder": embedder_config, - "chunker": { - "backend": "sentence", - "config": { - "tokenizer_or_token_counter": "gpt2", - "chunk_size": 512, - "chunk_overlap": 128, - "min_sentences_per_chunk": 1, - }, - }, - }, - }, - "max_turns_window": 20, - "top_k": 5, - "enable_textual_memory": True, - "enable_activation_memory": False, - "enable_parametric_memory": False, - "enable_preference_memory": True, - } - - cube_config = { - "model_schema": "memos.configs.mem_cube.GeneralMemCubeConfig", - "user_id": user_id, - "cube_id": f"{user_id}/mem_cube", - "text_mem": { - "backend": "tree_text", - "config": { - "cube_id": f"{user_id}/mem_cube", - "extractor_llm": llm_config, - "dispatcher_llm": llm_config, - "graph_db": { - "backend": "neo4j", - "config": { - "uri": "bolt://localhost:7687", - "user": "neo4j", - "password": "12345678", - "db_name": "neo4j", - "user_name": "memosneo4j", - "embedding_dimension": 768, - "use_multi_db": False, - "auto_create": False, - }, - }, - "embedder": embedder_config, - }, - }, - "act_mem": {"backend": "uninitialized", "config": {}}, - "para_mem": {"backend": "uninitialized", "config": {}}, - "pref_mem": { - "backend": "pref_text", - "config": { - "cube_id": f"{user_id}/mem_cube", - "extractor_llm": llm_config, - "vector_db": { - "backend": "milvus", - "config": { - "collection_name": [ - "explicit_preference", - "implicit_preference", - ], - "vector_dimension": 768, - "distance_metric": "cosine", - "uri": "./milvus_demo.db", - }, - }, - "embedder": embedder_config, - "extractor": {"backend": "naive", "config": {}}, - "adder": {"backend": "naive", "config": {}}, - "retriever": {"backend": "naive", "config": {}}, - }, - }, - } - mos_config = MOSConfig(**mos_config) - cube_config = GeneralMemCubeConfig.model_validate(cube_config) - - return mos_config, cube_config - - -user_id = "superman" -mos_config, mem_cube_config = get_config(user_id) - -memory = MOSProduct(default_config=mos_config) - -# register user and mem cube -mem_cube = GeneralMemCube(mem_cube_config) - -register_result = memory.user_register( - user_id=user_id, user_name=user_id, default_mem_cube=mem_cube -) -print(f"register_result: {register_result}") - -# add memory for user -memory.add( - messages=[ - # Sports and hobbies - {"role": "user", "content": "I like playing football."}, - { - "role": "assistant", - "content": "I like playing football too. It's a great way to stay active and work as a team.", - }, - {"role": "user", "content": "Do you enjoy watching basketball games?"}, - { - "role": "assistant", - "content": "Yes, I find basketball exciting to watch, especially the fast-paced action and strategic plays.", - }, - {"role": "user", "content": "I'm learning to play tennis this summer."}, - { - "role": "assistant", - "content": "That's wonderful! Tennis is a fantastic sport that improves both physical fitness and mental focus.", - }, - # Food and cooking - {"role": "user", "content": "What's your favorite type of cuisine?"}, - { - "role": "assistant", - "content": "I really enjoy Italian cuisine, especially pasta dishes and authentic pizza.", - }, - {"role": "user", "content": "I love cooking Asian food at home."}, - { - "role": "assistant", - "content": "Asian cooking is amazing! The balance of flavors and techniques is so interesting to learn.", - }, - {"role": "user", "content": "Have you ever tried making homemade bread?"}, - { - "role": "assistant", - "content": "Yes, there's something magical about the process of making bread from scratch. The smell is incredible!", - }, - # Travel and places - {"role": "user", "content": "I'm planning a trip to Japan next year."}, - { - "role": "assistant", - "content": "Japan is an incredible destination! The culture, food, and technology blend is fascinating.", - }, - {"role": "user", "content": "What's the most beautiful place you've ever seen?"}, - { - "role": "assistant", - "content": "I find mountain landscapes particularly breathtaking, especially during sunrise or sunset.", - }, - {"role": "user", "content": "I prefer beach vacations over city trips."}, - { - "role": "assistant", - "content": "Beach vacations are so relaxing! The sound of waves and warm sand is perfect for unwinding.", - }, - # Technology and innovation - {"role": "user", "content": "What do you think about artificial intelligence?"}, - { - "role": "assistant", - "content": "AI is transforming many aspects of our lives. It's exciting to see how it can help solve complex problems.", - }, - {"role": "user", "content": "I'm learning to code in Python."}, - { - "role": "assistant", - "content": "Python is a great language to start with! It's versatile and has a wonderful community.", - }, - {"role": "user", "content": "Do you use any productivity apps?"}, - { - "role": "assistant", - "content": "I find task management apps really helpful for organizing daily activities and staying focused.", - }, - # Music and entertainment - {"role": "user", "content": "What kind of music do you enjoy?"}, - { - "role": "assistant", - "content": "I appreciate many genres, but I particularly enjoy jazz and classical music for their complexity.", - }, - {"role": "user", "content": "I play the guitar in my free time."}, - { - "role": "assistant", - "content": "That's fantastic! Playing an instrument is such a rewarding creative outlet.", - }, - {"role": "user", "content": "Have you seen any good movies lately?"}, - { - "role": "assistant", - "content": "I enjoy films that make you think deeply about human nature and society.", - }, - # Books and learning - {"role": "user", "content": "Do you read books regularly?"}, - { - "role": "assistant", - "content": "Yes, reading is one of my favorite ways to learn and explore new ideas.", - }, - {"role": "user", "content": "I'm currently reading a science fiction novel."}, - { - "role": "assistant", - "content": "Science fiction is great for exploring possibilities and imagining different futures.", - }, - {"role": "user", "content": "What's your favorite subject to learn about?"}, - { - "role": "assistant", - "content": "I'm fascinated by psychology and how the human mind works.", - }, - # Nature and environment - {"role": "user", "content": "I love hiking in the mountains."}, - { - "role": "assistant", - "content": "Hiking is wonderful! It's a great way to connect with nature and get exercise.", - }, - {"role": "user", "content": "Are you concerned about climate change?"}, - { - "role": "assistant", - "content": "Yes, it's one of the most important challenges of our time that requires collective action.", - }, - {"role": "user", "content": "I try to recycle and reduce waste at home."}, - { - "role": "assistant", - "content": "That's commendable! Every small action contributes to environmental protection.", - }, - # Work and career - {"role": "user", "content": "What do you do for work?"}, - { - "role": "assistant", - "content": "I work in technology, helping people solve problems and improve their daily lives.", - }, - {"role": "user", "content": "I'm considering a career change."}, - { - "role": "assistant", - "content": "Career changes can be challenging but also very rewarding when you find the right fit.", - }, - {"role": "user", "content": "Do you work from home or in an office?"}, - { - "role": "assistant", - "content": "I work remotely, which gives me flexibility in how I organize my day.", - }, - # Health and wellness - {"role": "user", "content": "I've started doing yoga in the morning."}, - { - "role": "assistant", - "content": "Yoga is excellent for both physical and mental well-being. Great way to start the day!", - }, - {"role": "user", "content": "How do you manage stress?"}, - { - "role": "assistant", - "content": "I find meditation and regular exercise help me stay balanced and focused.", - }, - {"role": "user", "content": "I'm trying to eat more vegetables."}, - { - "role": "assistant", - "content": "That's a great health goal! Vegetables provide so many essential nutrients.", - }, - # Art and creativity - {"role": "user", "content": "Do you enjoy painting or drawing?"}, - { - "role": "assistant", - "content": "I appreciate visual arts, though I'm more of an observer than a creator.", - }, - {"role": "user", "content": "I'm learning photography."}, - { - "role": "assistant", - "content": "Photography is a beautiful way to capture moments and express your perspective.", - }, - {"role": "user", "content": "What's your favorite type of art?"}, - { - "role": "assistant", - "content": "I'm drawn to abstract art because it allows for personal interpretation.", - }, - # Family and relationships - {"role": "user", "content": "I have two siblings."}, - { - "role": "assistant", - "content": "Having siblings can be wonderful - they're often our first friends and lifelong companions.", - }, - {"role": "user", "content": "Do you have any pets?"}, - { - "role": "assistant", - "content": "I don't have pets, but I think they bring so much joy and companionship to people's lives.", - }, - {"role": "user", "content": "I'm planning to visit my grandparents this weekend."}, - { - "role": "assistant", - "content": "That's lovely! Grandparents often have such interesting stories and wisdom to share.", - }, - # Education and growth - {"role": "user", "content": "I'm taking an online course in data science."}, - { - "role": "assistant", - "content": "Data science is a fascinating field with many applications across different industries.", - }, - {"role": "user", "content": "What's the best way to learn a new language?"}, - { - "role": "assistant", - "content": "Immersion and consistent practice are key, along with finding methods that work for your learning style.", - }, - {"role": "user", "content": "I enjoy attending workshops and conferences."}, - { - "role": "assistant", - "content": "Workshops are great for hands-on learning and networking with like-minded people.", - }, - # Hobbies and interests - {"role": "user", "content": "I collect vintage postcards."}, - { - "role": "assistant", - "content": "What a unique hobby! Postcards offer glimpses into different times and places.", - }, - {"role": "user", "content": "Do you garden?"}, - { - "role": "assistant", - "content": "I appreciate gardening, though I'm more of a plant observer than a gardener myself.", - }, - {"role": "user", "content": "I love solving puzzles and brain teasers."}, - { - "role": "assistant", - "content": "Puzzles are excellent for keeping the mind sharp and developing problem-solving skills.", - }, - # Future and goals - {"role": "user", "content": "What are your goals for next year?"}, - { - "role": "assistant", - "content": "I aim to continue learning and growing, helping more people with their challenges.", - }, - {"role": "user", "content": "I want to start my own business someday."}, - { - "role": "assistant", - "content": "Entrepreneurship is exciting! It requires courage, planning, and persistence.", - }, - {"role": "user", "content": "Do you believe in setting New Year's resolutions?"}, - { - "role": "assistant", - "content": "I think any time is good for setting positive goals and making positive changes.", - }, - # Philosophy and life - {"role": "user", "content": "What's the meaning of life to you?"}, - { - "role": "assistant", - "content": "I believe it's about learning, growing, and making positive connections with others.", - }, - {"role": "user", "content": "I try to practice gratitude daily."}, - { - "role": "assistant", - "content": "Gratitude is such a powerful practice that can transform your perspective on life.", - }, - {"role": "user", "content": "What's your biggest fear?"}, - { - "role": "assistant", - "content": "I think fear of not being able to help when someone really needs it would be challenging.", - }, - # Technology and daily life - {"role": "user", "content": "How has technology changed your daily routine?"}, - { - "role": "assistant", - "content": "Technology has made many tasks more efficient and connected me with people worldwide.", - }, - {"role": "user", "content": "I'm trying to reduce my screen time."}, - { - "role": "assistant", - "content": "That's a healthy goal! Finding balance with technology is important for well-being.", - }, - {"role": "user", "content": "Do you use social media?"}, - { - "role": "assistant", - "content": "I think social media can be a great tool for connection when used mindfully.", - }, - # Final thoughts - {"role": "user", "content": "This has been a great conversation!"}, - { - "role": "assistant", - "content": "I've really enjoyed our chat too! It's wonderful to explore so many different topics together.", - }, - ], - user_id=user_id, - mem_cube_id=register_result["default_cube_id"], -) - -retrieved_memories = memory.search(query="What do you like?", user_id=user_id) -print( - f"len_pref_memories: {len(retrieved_memories['pref_mem'][0]['memories'])}" - if retrieved_memories["pref_mem"] - else 0 -) diff --git a/examples/mem_os/simple_treekvcache_memos.py b/examples/mem_os/simple_treekvcache_memos.py deleted file mode 100644 index 2f3408afd..000000000 --- a/examples/mem_os/simple_treekvcache_memos.py +++ /dev/null @@ -1,46 +0,0 @@ -import os - -from memos.configs.mem_cube import GeneralMemCubeConfig -from memos.configs.mem_os import MOSConfig -from memos.mem_cube.general import GeneralMemCube -from memos.mem_os.main import MOS - - -# init MOSConfig by deafult user -# note kvcache must at chatllm backend by huggingface -# gpu need set -os.environ["CUDA_VISIBLE_DEVICES"] = "1" -user_id = "root" -cube_id = "root/mem_cube_kv_cache" -tmp_cube_path = "/tmp/default/mem_cube_5" - -mos_config = MOSConfig.from_json_file("examples/data/config/simple_treekvcache_memos_config.json") -mos = MOS(mos_config) - - -# 2. Initialize_memory_cube -cube_config = GeneralMemCubeConfig.from_json_file( - "examples/data/config/simple_treekvcache_cube_config.json" -) - -# 3. Initialize the MemCube with the configuration and dump cube -mem_cube = GeneralMemCube(cube_config) -try: - mem_cube.dump(tmp_cube_path) -except Exception as e: - print(e) - -# 4. Register the MemCube explicitly -mos.register_mem_cube(tmp_cube_path, mem_cube_id=cube_id, user_id=user_id) - -# 5. Extract kv memory and add kv cache_mem -extract_kvmem = mos.mem_cubes[cube_id].act_mem.extract("I like football") -mos.mem_cubes[cube_id].act_mem.add([extract_kvmem]) - -# .chat -while True: - user_input = input("👤 [You] ").strip() - print() - response = mos.chat(user_input) - print(f"🤖 [Assistant] {response}\n") -print("📢 [System] MemChat has stopped.") diff --git a/examples/mem_os/simple_vllm_memos.py b/examples/mem_os/simple_vllm_memos.py deleted file mode 100644 index ffa7a3a24..000000000 --- a/examples/mem_os/simple_vllm_memos.py +++ /dev/null @@ -1,61 +0,0 @@ -#!/usr/bin/env python3 -""" -Simple example demonstrating how to use VLLMLLM with an existing vLLM server. -Requires a vLLM server to be running. -""" - -from typing import TYPE_CHECKING - -from memos.configs.llm import VLLMLLMConfig -from memos.llms.vllm import VLLMLLM - - -if TYPE_CHECKING: - from memos.types import MessageDict - - -def main(): - """Main function demonstrating VLLMLLM usage.""" - - # Configuration for connecting to existing vLLM server - config = VLLMLLMConfig( - model_name_or_path="/mnt/afs/models/hf_models/Qwen2.5-7B", # MUST MATCH the --model arg of vLLM server - api_key="", # Not needed for local server - api_base="http://localhost:8088/v1", # vLLM server address with /v1 - temperature=0.7, - max_tokens=512, - top_p=0.9, - model_schema="memos.configs.llm.VLLMLLMConfig", - ) - - # Initialize VLLM LLM - print("Initializing VLLM LLM...") - llm = VLLMLLM(config) - - # Test messages for KV cache building - print("\nBuilding KV cache for system messages...") - system_messages: list[MessageDict] = [ - {"role": "system", "content": "You are a helpful AI assistant."}, - {"role": "user", "content": "Hello! Can you tell me about vLLM?"}, - ] - try: - prompt = llm.build_vllm_kv_cache(system_messages) - print(f"✓ KV cache built successfully for prompt: '{prompt[:100]}...'") - except Exception as e: - print(f"✗ Failed to build KV cache: {e}") - - # Test with different messages for generation - print("\nGenerating response...") - user_messages: list[MessageDict] = [ - {"role": "system", "content": "You are a helpful AI assistant. Please Introduce yourself "}, - {"role": "user", "content": "What are the benefits of using vLLM?"}, - ] - try: - response = llm.generate(user_messages) - print(f"Response: {response}") - except Exception as e: - print(f"Error generating response: {e}") - - -if __name__ == "__main__": - main() diff --git a/examples/mem_reader/README.md b/examples/mem_reader/README.md new file mode 100644 index 000000000..3677d0502 --- /dev/null +++ b/examples/mem_reader/README.md @@ -0,0 +1,90 @@ +# MemReader Examples + +This directory contains examples and sample code demonstrating how to use the `MemReader` module in MemOS. `MemReader` is responsible for parsing various types of input data (text, chat history, files, images) into structured memory formats. + +## 📂 Directory Structure + +```text +examples/mem_reader/ +├── builders.py # Factory functions to initialize Reader components +├── parser_demos/ # Demos for individual parser components +│ ├── demo_image.py # Example: Parsing image content +│ ├── demo_string.py # Example: Parsing string content +│ └── ... # Other specific parser demos +├── runners/ # Main execution scripts for running examples +│ ├── run_simple.py # Runner for SimpleStructMemReader +│ └── run_multimodal.py# Runner for MultiModalStructMemReader +├── samples.py # Sample data (chat logs, test cases) +├── settings.py # Configuration management (loads from .env) +└── utils.py # Utility functions (printing, formatting) +``` + +## 🚀 Getting Started + +### 1. Configuration + +Before running the examples, ensure you have configured your environment variables. Copy the `.env.example` file in the project root to `.env` and fill in the necessary API keys. + +The `settings.py` file loads these configurations. Key variables include: +- `OPENAI_API_KEY`: For LLM and Embeddings. +- `MOS_CHAT_MODEL`: Default model for chat (e.g., `gpt-4o`). +- `MOS_EMBEDDER_MODEL`: Model for embeddings. + +### 2. Running Examples + +We provide two main runners to demonstrate different capabilities: + +#### A. Simple Reader (`run_simple.py`) + +Demonstrates the `SimpleStructMemReader`, which is optimized for text-based chat history and documents. + +**Features:** +- **Fast Mode**: Quick parsing without LLM (regex/rule-based). +- **Fine Mode**: Detailed parsing using LLM. +- **Transfer**: Converting Fast memories to Fine memories. +- **Document Parsing**: Reading text files. + +**Usage:** +```bash +python -m examples.mem_reader.runners.run_simple +``` + +#### B. Multimodal Reader (`run_multimodal.py`) + +Demonstrates the `MultiModalStructMemReader`, which handles complex inputs like images, files, and mixed content types. + +**Features:** +- Supports **String**, **Multimodal**, and **Raw** input types. +- Configurable output format (Text/JSON). +- Selectable test cases. + +**Usage:** +```bash +# Run all examples in 'fine' mode +python -m examples.mem_reader.runners.run_multimodal --example all --mode fine + +# Run specific example (e.g., multimodal inputs) +python -m examples.mem_reader.runners.run_multimodal --example multimodal + +# View help for more options +python -m examples.mem_reader.runners.run_multimodal --help +``` + +### 3. Parser Demos + +If you want to understand how specific parsers work internally (e.g., how the system parses a User message vs. an Assistant message), check the `parser_demos/` directory. + +**Usage:** +```bash +python -m examples.mem_reader.parser_demos.demo_user +python -m examples.mem_reader.parser_demos.demo_image +``` + +## 🧩 Key Components + +- **`SimpleStructMemReader`**: Best for standard text-based chat applications. It's lightweight and efficient. +- **`MultiModalStructMemReader`**: Designed for advanced agents that handle images, file attachments, and complex tool interactions. + +## 🛠️ Customization + +You can modify `settings.py` or `builders.py` to change the underlying LLM backend (e.g., switching from OpenAI to Ollama) or adjust chunking strategies. diff --git a/examples/mem_reader/builders.py b/examples/mem_reader/builders.py new file mode 100644 index 000000000..89c39a5de --- /dev/null +++ b/examples/mem_reader/builders.py @@ -0,0 +1,78 @@ +"""Builder functions for initializing MemReader components. + +This module provides factory functions to create configured instances of +LLMs, Embedders, and MemReaders, simplifying the setup process in examples. +""" + +from typing import Any + +from memos.configs.embedder import EmbedderConfigFactory +from memos.configs.llm import LLMConfigFactory +from memos.configs.mem_reader import ( + MultiModalStructMemReaderConfig, + SimpleStructMemReaderConfig, +) +from memos.configs.parser import ParserConfigFactory +from memos.embedders.factory import EmbedderFactory +from memos.llms.factory import LLMFactory +from memos.mem_reader.multi_modal_struct import MultiModalStructMemReader +from memos.mem_reader.simple_struct import SimpleStructMemReader +from memos.parsers.factory import ParserFactory + +from .settings import get_embedder_config, get_llm_config, get_reader_config + + +def build_llm_and_embedder() -> tuple[Any, Any]: + """Initialize and return configured LLM and Embedder instances.""" + llm_config_dict = get_llm_config() + embedder_config_dict = get_embedder_config() + + llm_config = LLMConfigFactory.model_validate(llm_config_dict) + embedder_config = EmbedderConfigFactory.model_validate(embedder_config_dict) + + llm = LLMFactory.from_config(llm_config) + embedder = EmbedderFactory.from_config(embedder_config) + + return embedder, llm + + +def build_file_parser() -> Any: + """Initialize and return a configured file parser (MarkItDown). + + Returns: + Configured parser instance or None if initialization fails. + """ + try: + parser_config = ParserConfigFactory.model_validate( + { + "backend": "markitdown", + "config": {}, + } + ) + return ParserFactory.from_config(parser_config) + except Exception as e: + print(f"⚠️ Warning: Could not initialize file parser: {e}") + return None + + +def build_simple_reader() -> SimpleStructMemReader: + """Initialize and return a configured SimpleStructMemReader. + + Returns: + Configured SimpleStructMemReader instance. + """ + config_dict = get_reader_config() + # Simple reader doesn't need file parser + config = SimpleStructMemReaderConfig(**config_dict) + return SimpleStructMemReader(config) + + +def build_multimodal_reader() -> MultiModalStructMemReader: + """Initialize and return a configured MultiModalStructMemReader. + + Returns: + Configured MultiModalStructMemReader instance. + """ + config_dict = get_reader_config() + config = MultiModalStructMemReaderConfig(**config_dict) + return MultiModalStructMemReader(config) diff --git a/examples/mem_reader/compare_simple_vs_multimodal.py b/examples/mem_reader/compare_simple_vs_multimodal.py deleted file mode 100644 index fa12ac211..000000000 --- a/examples/mem_reader/compare_simple_vs_multimodal.py +++ /dev/null @@ -1,461 +0,0 @@ -"""Compare SimpleStructMemReader and MultiModalStructMemReader outputs. - -This example demonstrates the differences between simple_struct and multi_modal_struct -in both fast and fine modes. -""" - -import os -import sys - -from pathlib import Path - -from dotenv import load_dotenv - -from memos.configs.mem_reader import ( - MultiModalStructMemReaderConfig, - SimpleStructMemReaderConfig, -) -from memos.memories.textual.item import TextualMemoryItem - - -# Add src directory to path -project_root = Path(__file__).parent.parent.parent -src_path = project_root / "src" -if str(src_path) not in sys.path: - sys.path.insert(0, str(src_path)) - -# Load environment variables -load_dotenv() - - -def get_reader_config() -> dict: - """Get reader configuration from environment variables.""" - openai_api_key = os.getenv("OPENAI_API_KEY") - openai_base_url = os.getenv("OPENAI_API_BASE", "https://api.openai.com/v1") - ollama_api_base = os.getenv("OLLAMA_API_BASE", "http://localhost:11434") - - # LLM config - llm_backend = os.getenv("MEM_READER_LLM_BACKEND", "openai") - if llm_backend == "ollama": - llm_config = { - "backend": "ollama", - "config": { - "model_name_or_path": os.getenv("MEM_READER_LLM_MODEL", "qwen3:0.6b"), - "api_base": ollama_api_base, - "temperature": float(os.getenv("MEM_READER_LLM_TEMPERATURE", "0.0")), - "remove_think_prefix": os.getenv( - "MEM_READER_LLM_REMOVE_THINK_PREFIX", "true" - ).lower() - == "true", - "max_tokens": int(os.getenv("MEM_READER_LLM_MAX_TOKENS", "8192")), - }, - } - else: # openai - llm_config = { - "backend": "openai", - "config": { - "model_name_or_path": os.getenv("MEM_READER_LLM_MODEL", "gpt-4o-mini"), - "api_key": openai_api_key or os.getenv("MEMRADER_API_KEY", "EMPTY"), - "api_base": openai_base_url, - "temperature": float(os.getenv("MEM_READER_LLM_TEMPERATURE", "0.5")), - "remove_think_prefix": os.getenv( - "MEM_READER_LLM_REMOVE_THINK_PREFIX", "true" - ).lower() - == "true", - "max_tokens": int(os.getenv("MEM_READER_LLM_MAX_TOKENS", "8192")), - }, - } - - # Embedder config - embedder_backend = os.getenv( - "MEM_READER_EMBEDDER_BACKEND", os.getenv("MOS_EMBEDDER_BACKEND", "ollama") - ) - if embedder_backend == "universal_api": - embedder_config = { - "backend": "universal_api", - "config": { - "provider": os.getenv( - "MEM_READER_EMBEDDER_PROVIDER", - os.getenv("MOS_EMBEDDER_PROVIDER", "openai"), - ), - "api_key": os.getenv( - "MEM_READER_EMBEDDER_API_KEY", - os.getenv("MOS_EMBEDDER_API_KEY", openai_api_key or "sk-xxxx"), - ), - "model_name_or_path": os.getenv( - "MEM_READER_EMBEDDER_MODEL", - os.getenv("MOS_EMBEDDER_MODEL", "text-embedding-3-large"), - ), - "base_url": os.getenv( - "MEM_READER_EMBEDDER_API_BASE", - os.getenv("MOS_EMBEDDER_API_BASE", openai_base_url), - ), - }, - } - else: # ollama - embedder_config = { - "backend": "ollama", - "config": { - "model_name_or_path": os.getenv( - "MEM_READER_EMBEDDER_MODEL", - os.getenv("MOS_EMBEDDER_MODEL", "nomic-embed-text:latest"), - ), - "api_base": ollama_api_base, - }, - } - - return { - "llm": llm_config, - "embedder": embedder_config, - "chunker": { - "backend": "sentence", - "config": { - "tokenizer_or_token_counter": "gpt2", - "chunk_size": 512, - "chunk_overlap": 128, - "min_sentences_per_chunk": 1, - }, - }, - } - - -def print_memory_item(item: TextualMemoryItem, prefix: str = "", max_length: int = 500): - """Print a memory item in a readable format.""" - print(f"{prefix}Memory ID: {item.id}") - print(f"{prefix}Memory Type: {item.metadata.memory_type}") - print(f"{prefix}Tags: {item.metadata.tags}") - memory_preview = ( - item.memory[:max_length] + "..." if len(item.memory) > max_length else item.memory - ) - print(f"{prefix}Memory: {memory_preview}") - print(f"{prefix}Key: {item.metadata.key}") - if item.metadata.background: - bg_preview = ( - item.metadata.background[:max_length] + "..." - if len(item.metadata.background) > max_length - else item.metadata.background - ) - print(f"{prefix}Background: {bg_preview}") - print(f"{prefix}Sources count: {len(item.metadata.sources) if item.metadata.sources else 0}") - print() - - -def compare_readers(): - """Compare SimpleStructMemReader and MultiModalStructMemReader.""" - print("=" * 80) - print("Comparing SimpleStructMemReader vs MultiModalStructMemReader") - print("=" * 80) - print() - - # Test data - simple chat messages - scene_data = [ - [ - {"role": "user", "chat_time": "3 May 2025", "content": "I'm feeling a bit down today."}, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "I'm sorry to hear that. Do you want to talk about what's been going on?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "It's just been a tough couple of days, you know? Everything feels a bit overwhelming, and I just can't seem to shake it off.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "It sounds like you're going through a lot right now. Sometimes it helps to talk things out. Is there something specific that's been weighing on you, or is it more of a general feeling?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "It's a mix, I guess. Work's been really stressful, and on top of that, I've been feeling kinda disconnected from the people around me.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "That can be really draining, especially when you're feeling isolated on top of the stress. Do you think there's something from your past that's contributing to how you're feeling now? Sometimes our emotions are tied to older experiences.", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "Yeah, now that you mention it… I've been thinking a lot about my childhood lately. I didn't have the easiest upbringing. My parents were always busy, and I often felt like I had to figure things out on my own. I remember one winter, I got really sick with a high fever. My parents were both working, and I didn't want to bother them. I just lay in bed, crying by myself, not knowing what to do.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "That must have been very lonely. Growing up without that support can really make you feel like you have to handle everything on your own. How do you think that experience has affected you now?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "Yeah, it was really lonely. I remember spending a lot of time doing things on my own, like drawing, because it was my only way to escape. I used to draw animals, and I created my own little world on paper. It was peaceful in a way, just me and my imagination, with no outside distractions.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "Drawing sounds like a beautiful way to express yourself and find some peace in that solitude. It must have been a very quiet and private world for you. Do you think that sense of loneliness from back then has shaped how you connect with others today?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "I think it has. I've always found it hard to let others in. I feel like I have to handle everything on my own. If I rely on someone else, I feel vulnerable, and I start to feel that loneliness again.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "It sounds like this was a protective mechanism that helped you cope with your feelings of loneliness back then. But as you've grown, it seems like this pattern might not be serving you as well anymore. Have you thought about trying to break this cycle, or finding new ways to connect with others?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "I've thought about it, though it's hard. But I also know I can't keep going like this forever. Maybe it's time to seek some help.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "It's a big step, but a positive one. Exploring your emotions and past experiences could really help you understand how they've shaped your present. Therapy could be a great place to start, as it would help you connect with both yourself and others. Would you be open to trying it?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "I'm open to trying. Thank you for encouraging me. It feels like it might be a good starting point for me.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "I'm really glad to hear that. Taking that first step is always difficult, but you've made an important decision. If you need help finding the right therapist, or just want to talk, I'm here for you.", - }, - ], - [ - { - "role": "user", - "chat_time": "3 May 2025", - "content": "This week I've been feeling a bit off, especially when playing tennis. My body just doesn't feel right.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "It sounds like you've been having some physical discomfort lately. Where specifically are you feeling unwell? Is there something particular about tennis that's been difficult?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "I just feel a bit fatigued, and every time I play tennis, my movements feel off. I used to play pretty well, but lately, I've been feeling weak and my movements aren't as coordinated as they used to be.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "Physical discomfort can definitely affect performance, especially in a sport like tennis that requires quick reactions and coordination. Have you noticed anything specific that might be causing these changes? Could it be overtraining, or is there another physical issue making you feel off?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "I think it might be due to stress and lack of sleep recently. I just feel drained all the time. Plus, I've been frustrated with my tennis performance, which makes me feel even worse, like I'm stuck in a vicious cycle.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "Stress and poor sleep quality can definitely affect your physical performance, especially in sports. When you're also feeling down about your performance, it's easy to fall into a negative cycle. Have you tried anything to relieve some of the stress, or to improve your sleep?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "I've tried some relaxation techniques, like deep breathing and simple meditation, but it doesn't seem to help much. I still can't focus during tennis, and I feel like my mind is distracted by other things.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "Relaxation techniques like deep breathing and meditation can help reduce stress, but sometimes we need a more structured approach to improve both our physical and mental state. For example, you might want to focus on specific aspects of your tennis training, or adjust your rest and recovery time. Have you thought about setting smaller goals for yourself? Like focusing on a specific tennis move each day, rather than expecting perfection right away? That might help you gradually regain confidence.", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "That's a good idea. I think I've been expecting too much from myself, which just makes me more anxious. Maybe I can start with small practice sessions and gradually find my rhythm again.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "Exactly! Setting small and clear goals will help reduce pressure and give you a sense of accomplishment. And don't forget that rest is just as important—taking time to recover will help you feel better physically and mentally. Both physical and mental recovery are key to improving your performance and health.", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "Thanks, I feel like I need to rethink a few things. Maybe I really need to be kinder to myself and take things step by step.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "You're on the right track. Rethinking things and giving yourself space and patience is so important. You've already taken the first step, now just keep moving forward, one step at a time. If you need anything, I'm always here to help.", - }, - ], - [ - { - "role": "user", - "chat_time": "3 May 2025", - "content": "Yesterday at 3 PM I had a meeting with my team about the new project. We discussed the timeline and I'm concerned we won't meet the deadline of December 15th.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "I understand your concern about the project timeline. What specific challenges do you think might prevent your team from meeting the December 15th deadline?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "We have too many requirements to implement in just two weeks. The backend team needs at least until December 10th to finish their part, which leaves almost no time for testing before the deadline.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "That's definitely a tight timeline. Have you considered prioritizing the requirements to focus on the most critical features first? You might also want to discuss with stakeholders about possibly extending the deadline or reducing the scope for this initial release.", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "That's a good point. We have a follow-up meeting scheduled for tomorrow morning at 9:30 AM. I'll suggest we prioritize features and maybe propose pushing the deadline to January 5th next year.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "That sounds like a solid approach. Coming to the meeting with both a prioritization strategy and a specific alternative deadline shows you've thought through the problem carefully. It might also help to bring some data about similar past projects and how long they took to complete, if you have that information available.", - }, - ], - ] - - info = {"user_id": "test_user", "session_id": "test_session"} - - # Get config - config_dict = get_reader_config() - - # Initialize readers - print("Initializing readers...") - simple_config = SimpleStructMemReaderConfig(**config_dict) - multimodal_config = MultiModalStructMemReaderConfig(**config_dict) - - from memos.mem_reader.multi_modal_struct import MultiModalStructMemReader - from memos.mem_reader.simple_struct import SimpleStructMemReader - - simple_reader = SimpleStructMemReader(simple_config) - multimodal_reader = MultiModalStructMemReader(multimodal_config) - print("Readers initialized.\n") - print(f"Using LLM: {config_dict['llm']['backend']}") - print(f"Using Embedder: {config_dict['embedder']['backend']}") - print() - - # Test FAST mode - print("=" * 80) - print("FAST MODE COMPARISON") - print("=" * 80) - print() - - print("-" * 80) - print("SimpleStructMemReader (FAST):") - print("-" * 80) - try: - simple_fast = simple_reader.get_memory(scene_data, "chat", info, mode="fast") - if simple_fast and len(simple_fast) > 0: - for scene_idx, scene_memories in enumerate(simple_fast): - print(f"\nScene {scene_idx + 1}:") - for item_idx, item in enumerate(scene_memories): - print_memory_item(item, prefix=f" [{item_idx + 1}] ") - else: - print(" No memories generated.") - except Exception as e: - print(f" Error: {e}") - import traceback - - traceback.print_exc() - - print("\n" + "-" * 80) - print("MultiModalStructMemReader (FAST):") - print("-" * 80) - try: - multimodal_fast = multimodal_reader.get_memory(scene_data, "chat", info, mode="fast") - if multimodal_fast and len(multimodal_fast) > 0: - for scene_idx, scene_memories in enumerate(multimodal_fast): - print(f"\nScene {scene_idx + 1}:") - for item_idx, item in enumerate(scene_memories): - print_memory_item(item, prefix=f" [{item_idx + 1}] ") - else: - print(" No memories generated.") - except Exception as e: - print(f" Error: {e}") - import traceback - - traceback.print_exc() - - # Test FINE mode - print("\n" + "=" * 80) - print("FINE MODE COMPARISON") - print("=" * 80) - print() - - print("-" * 80) - print("SimpleStructMemReader (FINE):") - print("-" * 80) - try: - simple_fine = simple_reader.get_memory(scene_data, "chat", info, mode="fine") - if simple_fine and len(simple_fine) > 0: - for scene_idx, scene_memories in enumerate(simple_fine): - print(f"\nScene {scene_idx + 1}:") - for item_idx, item in enumerate(scene_memories): - print_memory_item(item, prefix=f" [{item_idx + 1}] ") - else: - print(" No memories generated.") - except Exception as e: - print(f" Error: {e}") - import traceback - - traceback.print_exc() - - print("\n" + "-" * 80) - print("MultiModalStructMemReader (FINE):") - print("-" * 80) - try: - multimodal_fine = multimodal_reader.get_memory(scene_data, "chat", info, mode="fine") - if multimodal_fine and len(multimodal_fine) > 0: - for scene_idx, scene_memories in enumerate(multimodal_fine): - print(f"\nScene {scene_idx + 1}:") - for item_idx, item in enumerate(scene_memories): - print_memory_item(item, prefix=f" [{item_idx + 1}] ") - else: - print(" No memories generated.") - except Exception as e: - print(f" Error: {e}") - import traceback - - traceback.print_exc() - - # Summary comparison - print("\n" + "=" * 80) - print("SUMMARY") - print("=" * 80) - print() - - def count_memories(memories_list): - """Count total memories across all scenes.""" - if not memories_list: - return 0 - return sum(len(scene) for scene in memories_list if scene) - - simple_fast_count = count_memories(simple_fast) if "simple_fast" in locals() else 0 - multimodal_fast_count = count_memories(multimodal_fast) if "multimodal_fast" in locals() else 0 - simple_fine_count = count_memories(simple_fine) if "simple_fine" in locals() else 0 - multimodal_fine_count = count_memories(multimodal_fine) if "multimodal_fine" in locals() else 0 - - print(f"SimpleStructMemReader FAST: {simple_fast_count} memories") - print(f"MultiModalStructMemReader FAST: {multimodal_fast_count} memories") - print(f"SimpleStructMemReader FINE: {simple_fine_count} memories") - print(f"MultiModalStructMemReader FINE: {multimodal_fine_count} memories") - print() - - print("Key Differences:") - print("1. Both readers should produce similar results for simple text messages") - print("2. MultiModalStructMemReader can handle multimodal content (images, files, etc.)") - print("3. FINE mode uses LLM to extract structured memories from aggregated windows") - print("4. FAST mode directly aggregates messages into windows without LLM processing") - - -if __name__ == "__main__": - compare_readers() diff --git a/examples/mem_reader/multimodal_struct_reader.py b/examples/mem_reader/multimodal_struct_reader.py deleted file mode 100644 index 790b13f85..000000000 --- a/examples/mem_reader/multimodal_struct_reader.py +++ /dev/null @@ -1,1042 +0,0 @@ -#!/usr/bin/env python3 -""" -MultiModalStructMemReader Example Script - -This script demonstrates various use cases for MultiModalStructMemReader, -including different message types, modes (fast/fine), and output formats. - -Usage: - python multimodal_struct_reader.py --example all - python multimodal_struct_reader.py --example string_message --mode fast - python multimodal_struct_reader.py --example multimodal --format json -""" - -import argparse -import json -import os -import sys -import time - -from pathlib import Path -from typing import Any - -from dotenv import load_dotenv - -from memos.configs.mem_reader import MultiModalStructMemReaderConfig -from memos.mem_reader.multi_modal_struct import MultiModalStructMemReader -from memos.memories.textual.item import TextualMemoryItem - - -# Add src directory to path -src_path = Path(__file__).parent.parent.parent / "src" -sys.path.insert(0, str(src_path)) - -# Load environment variables -load_dotenv() - - -# ============================================================================ -# Test Case Definitions -# ============================================================================ - - -class TestCase: - """Base class for test cases.""" - - def __init__( - self, - name: str, - description: str, - scene_data: Any, - expected_count: dict[str, int] | None = None, - ): - """ - Initialize a test case. - - Args: - name: Test case name - description: Test case description - scene_data: Scene data to test - expected_count: Expected memory count for each mode (optional) - """ - self.name = name - self.description = description - self.scene_data = scene_data - self.expected_count = expected_count or {} - - def get_info(self) -> dict[str, Any]: - """Get info dict for this test case.""" - return { - "user_id": "test_user", - "session_id": f"session_{self.name}", - "test_case": self.name, - } - - -# String message test cases -STRING_MESSAGE_CASES = [ - TestCase( - name="string_simple", - description="Simple string message", - scene_data=["今天心情不错，喝了咖啡。"], - expected_count={"fast": 1, "fine": 1}, # StringParser returns [] in - # fast mode - ), - TestCase( - name="string_multiple", - description="Multiple string messages", - scene_data=[ - "这是第一条消息。", - "这是第二条消息。", - "这是第三条消息。", - ], - ), -] - -# Standard chat message test cases -CHAT_MESSAGE_CASES = [ - TestCase( - name="chat_simple", - description="Simple chat conversation", - scene_data=[ - [ - { - "role": "user", - "content": "Hello, how are you?", - "chat_time": "2025-01-01T10:00:00Z", - }, - { - "role": "assistant", - "content": "I'm doing well, thank you!", - "chat_time": "2025-01-01T10:00:01Z", - }, - ] - ], - ), - TestCase( - name="chat_with_system", - description="Chat with system message", - scene_data=[ - [ - { - "role": "system", - "content": [{"type": "text", "text": "You are a helpful assistant."}], - "chat_time": "2025-01-01T10:00:00Z", - }, - { - "role": "user", - "content": "What's the weather?", - "chat_time": "2025-01-01T10:00:01Z", - }, - { - "role": "assistant", - "content": "I don't have access to weather data.", - "chat_time": "2025-01-01T10:00:02Z", - }, - ] - ], - ), - TestCase( - name="chat_long_conversation", - description="Long conversation with multiple turns", - scene_data=[ - [ - { - "role": "user", - "chat_time": "3 May 2025", - "content": "I'm feeling a bit down today.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "I'm sorry to hear that. Do you want to talk about what's been going on?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "It's just been a tough couple of days.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "It sounds like you're going through a lot right now.", - }, - ] - ], - ), - TestCase( - name="chat_with_list_content", - description="", - scene_data=[ - [ - { - "role": "user", - "content": [ - { - "type": "text", - "text": "我是测试base64", - }, - { - "type": "file", - "file": { - "file_data": "Hello World", - "filename": "2102b64c-25a2-481c-a940-4325496baf39.txt", - "file_id": "90ee1bcf-5295-4b75-91a4-23fe1f7ab30a", - }, - }, - { - "type": "image_url", - "image_url": { - "url": "https://play-groud-test-1.oss-cn-shanghai.aliyuncs.com/algorithmImages/2025/12/01/ce545319ba6d4d21a0aebcb75337acc3.jpeg" - }, - }, - ], - "message_id": "1995458892790317057", - } - ] - ], - ), -] - -# Tool-related test cases -TOOL_MESSAGE_CASES = [ - TestCase( - name="tool_assistant_with_calls", - description="Assistant message with tool_calls", - scene_data=[ - [ - { - "role": "assistant", - "content": None, - "tool_calls": [ - { - "id": "tool-call-weather-1", - "type": "function", - "function": { - "name": "get_weather", - "arguments": '{"location": "北京"}', - }, - } - ], - "chat_time": "2025-11-24T10:12:00Z", - "message_id": "assistant-with-call-1", - } - ] - ], - ), - TestCase( - name="tool_with_result", - description="Tool call with result message", - scene_data=[ - [ - { - "role": "assistant", - "content": None, - "tool_calls": [ - { - "id": "tool-call-weather-1", - "type": "function", - "function": { - "name": "get_weather", - "arguments": '{"location": "北京"}', - }, - } - ], - "chat_time": "2025-11-24T10:12:00Z", - }, - { - "role": "tool", - "content": "北京今天天气晴朗，温度25°C，湿度60%。", - "tool_call_id": "tool-call-weather-1", - "chat_time": "2025-11-24T10:12:05Z", - }, - ] - ], - ), - TestCase( - name="tool_custom_format", - description="Custom tool format (tool_description, tool_input, tool_output)", - scene_data=[ - [ - { - "type": "tool_description", - "name": "get_weather", - "description": "获取指定地点的当前天气信息", - "parameters": { - "type": "object", - "properties": {"location": {"type": "string", "description": "城市名称"}}, - "required": ["location"], - }, - }, - { - "type": "tool_input", - "call_id": "call_123", - "name": "get_weather", - "argument": {"location": "北京"}, - }, - { - "type": "tool_output", - "call_id": "call_123", - "name": "get_weather", - "output": {"weather": "晴朗", "temperature": 25, "humidity": 60}, - }, - ] - ], - ), -] - -# Multimodal message test cases -MULTIMODAL_MESSAGE_CASES = [ - TestCase( - name="multimodal_text_image", - description="User message with text and image", - scene_data=[ - [ - { - "role": "user", - "content": [ - {"type": "text", "text": "帮我看看这张图片大概是什么内容？"}, - { - "type": "image_url", - "image_url": { - "url": "https://example.com/mountain_lake.jpg", - "detail": "high", - }, - }, - ], - "chat_time": "2025-11-24T10:20:00Z", - "message_id": "mm-img-1", - } - ] - ], - ), - TestCase( - name="multimodal_text_file", - description="User message with text and file", - scene_data=[ - [ - { - "role": "user", - "content": [ - {"type": "text", "text": "请阅读这个PDF，总结里面的要点。"}, - {"type": "file", "file": {"file_id": "file_123", "filename": "report.pdf"}}, - ], - "chat_time": "2025-11-24T10:21:00Z", - "message_id": "mm-file-1", - } - ] - ], - ), - TestCase( - name="oss_text_file", - description="User message with text and file", - scene_data=[ - [ - { - "role": "user", - "content": [ - {"type": "text", "text": "请阅读这个PDF，总结里面的要点。"}, - { - "type": "file", - "file": { - "file_id": "file_123", - "filename": "report.pdf", - "file_data": "@http://139.196.232.20:9090/graph-test/algorithm/2025_11_13/1763043889_1763043782_PM1%E8%BD%A6%E9%97%B4PMT%E9%9D%B4%E5%8E%8B%E8%BE%B9%E5%8E%8B%E5%8E%8B%E5%8A%9B%E6%97%A0%E6%B3%95%E5%BB%BA%E7%AB%8B%E6%95%85%E9%9A%9C%E6%8A%A5%E5%91%8A20240720.md", - }, - }, - ], - "chat_time": "2025-11-24T10:21:00Z", - "message_id": "mm-file-1", - } - ] - ], - ), - TestCase( - name="pure_data_file", - description="User message with text and file", - scene_data=[ - [ - { - "role": "user", - "content": [ - {"type": "text", "text": "请阅读这个PDF，总结里面的要点。"}, - { - "type": "file", - "file": { - "file_id": "file_123", - "filename": "report.pdf", - "file_data": "明文记忆是系统与用户对话、操作等交互中动态习得，以及外部提供的、可显式管理的结构化知识形态，通常以文档、提示模板、图结构或用户规则等形式存在。它具备编辑性、可共享性与治理友好性，适合存储需要频繁修改、可审计或多方协同使用的信息。 在 MemOS 中，明文记忆可用于动态生成推理上下文、个性化偏好注入、多代理协作共享等场景，成为连接人类输入与模型认知的关键桥梁。激活记忆是指模型在推理过程中产生的瞬时性认知状态，包括 KV cache、隐藏层激活、注意力权重等中间张量结构。它通常用于维持上下文连续性、对话一致性与行为风格控制。 MemOS 将激活记忆抽象为可调度资源，支持按需唤醒、延迟卸载与结构变换。例如，某些上下文状态可以被压缩为“半结构化记忆片段”用于未来复用，也可以在任务级别转化为参数化模块，支持短期记忆的长期化演进。这一机制为模型行为一致性、风格保持与状态持续性提供了基础。", - }, - }, - ], - "chat_time": "2025-11-24T10:21:00Z", - "message_id": "mm-file-1", - } - ] - ], - ), - TestCase( - name="local_data_file", - description="User message with text and file", - scene_data=[ - [ - { - "role": "user", - "content": [ - {"type": "text", "text": "请阅读这个PDF，总结里面的要点。"}, - { - "type": "file", - "file": { - "file_id": "file_123", - "filename": "report.pdf", - "file_data": "./my_local_file/report.pdf", - }, - }, - ], - "chat_time": "2025-11-24T10:21:00Z", - "message_id": "mm-file-1", - } - ] - ], - ), - TestCase( - name="internet_file", - description="User message with text and file", - scene_data=[ - [ - { - "role": "user", - "content": [ - {"type": "text", "text": "请阅读这个PDF，总结里面的要点。"}, - { - "type": "file", - "file": { - "file_id": "file_123", - "filename": "report.pdf", - "file_data": "https://upload.wikimedia.org/wikipedia/commons/c/cb/NLC416-16jh004830-88775_%E7%B4%85%E6%A8%93%E5%A4%A2.pdf", - }, - }, - ], - "chat_time": "2025-11-24T10:21:00Z", - "message_id": "mm-file-1", - } - ] - ], - ), - TestCase( - name="multimodal_mixed", - description="Mixed multimodal message (text + file + image)", - scene_data=[ - [ - { - "role": "user", - "content": [ - {"type": "text", "text": "请同时分析这个报告和图表。"}, - { - "type": "file", - "file": {"file_id": "file_789", "filename": "analysis_report.pdf"}, - }, - { - "type": "image_url", - "image_url": {"url": "https://example.com/chart.png", "detail": "auto"}, - }, - ], - "chat_time": "2025-11-24T10:23:00Z", - "message_id": "mixed-1", - } - ] - ], - ), - TestCase( - name="multimodal_audio", - description="Audio-only message", - scene_data=[ - [ - { - "role": "user", - "content": [ - { - "type": "input_audio", - "input_audio": {"data": "base64_encoded_audio_here", "format": "mp3"}, - } - ], - "chat_time": "2025-11-24T10:22:00Z", - "message_id": "audio-1", - } - ] - ], - ), -] - -# Raw input item test cases -RAW_INPUT_CASES = [ - TestCase( - name="raw_text_items", - description="Pure text input items without dialog context", - scene_data=[ - [ - {"type": "text", "text": "这是一段独立的文本输入，没有明确的对话上下文。"}, - {"type": "text", "text": "它依然会被抽取和写入明文记忆。"}, - ] - ], - ), - TestCase( - name="raw_file_item", - description="Pure file input by file_id", - scene_data=[ - [{"type": "file", "file": {"file_id": "file_uploaded_123", "filename": "document.pdf"}}] - ], - ), - # File parameter test cases - covering all combinations - TestCase( - name="file_only_file_id", - description="File with only file_id parameter", - scene_data=[[{"type": "file", "file": {"file_id": "file_only_id_123"}}]], - ), - TestCase( - name="file_only_filename", - description="File with only filename parameter", - scene_data=[[{"type": "file", "file": {"filename": "document_only.pdf"}}]], - ), - TestCase( - name="file_only_file_data_base64", - description="File with only file_data (base64 encoded)", - scene_data=[ - [ - { - "type": "file", - "file": { - "file_data": "data:application/pdf;base64,JVBERi0xLjQKJdPr6eEKMSAwIG9iago8PAovVHlwZSAvQ2F0YWxvZwovUGFnZXMgMiAwIFIKPj4KZW5kb2JqCjIgMCBvYmoKPDwKL1R5cGUgL1BhZ2VzCi9LaWRzIFszIDAgUl0KL0NvdW50IDEKPD4KZW5kb2JqCjMgMCBvYmoKPDwKL1R5cGUgL1BhZ2UKL1BhcmVudCAyIDAgUgovTWVkaWFCb3ggWzAgMCA2MTIgNzkyXQovUmVzb3VyY2VzIDw8Ci9Gb250IDw8Ci9GMSA0IDAgUgo+Pgo+PgovQ29udGVudHMgNSAwIFIKPj4KZW5kb2JqCjQgMCBvYmoKPDwKL1R5cGUgL0ZvbnQKL1N1YnR5cGUgL1R5cGUxCi9CYXNlRm9udCAvSGVsdmV0aWNhCj4+CmVuZG9iag==" - }, - } - ] - ], - ), - TestCase( - name="file_only_file_data_url", - description="File with only file_data (URL)", - scene_data=[ - [ - { - "type": "file", - "file": {"file_data": "https://example.com/documents/report.pdf"}, - } - ] - ], - ), - TestCase( - name="file_only_file_data_text", - description="File with only file_data (plain text content)", - scene_data=[ - [ - { - "type": "file", - "file": { - "file_data": "This is a plain text file content. It contains multiple lines.\nLine 2 of the file.\nLine 3 of the file." - }, - } - ] - ], - ), - TestCase( - name="file_file_data_and_file_id", - description="File with file_data and file_id", - scene_data=[ - [ - { - "type": "file", - "file": { - "file_data": "https://example.com/documents/data.pdf", - "file_id": "file_with_data_123", - }, - } - ] - ], - ), - TestCase( - name="file_file_data_and_filename", - description="File with file_data and filename", - scene_data=[ - [ - { - "type": "file", - "file": { - "file_data": "This is file content with filename.", - "filename": "content_with_name.txt", - }, - } - ] - ], - ), - TestCase( - name="file_file_id_and_filename", - description="File with file_id and filename (existing case)", - scene_data=[ - [{"type": "file", "file": {"file_id": "file_uploaded_123", "filename": "document.pdf"}}] - ], - ), - TestCase( - name="file_all_parameters", - description="File with all parameters (file_data, file_id, filename)", - scene_data=[ - [ - { - "type": "file", - "file": { - "file_data": "https://example.com/documents/complete.pdf", - "file_id": "file_complete_123", - "filename": "complete_document.pdf", - }, - } - ] - ], - ), - TestCase( - name="file_no_parameters", - description="File with no parameters (should return [File: unknown])", - scene_data=[[{"type": "file", "file": {}}]], - ), -] - -# Assistant message test cases -ASSISTANT_MESSAGE_CASES = [ - TestCase( - name="assistant_with_refusal", - description="Assistant message with refusal", - scene_data=[ - [ - { - "role": "assistant", - "content": [{"type": "text", "text": "I can help you with that."}], - "refusal": "I cannot provide information about that topic.", - "chat_time": "2025-11-24T10:30:00Z", - } - ] - ], - ), - TestCase( - name="assistant_with_audio", - description="Assistant message with audio", - scene_data=[ - [ - { - "role": "assistant", - "content": "Here's the audio response.", - "audio": {"id": "audio_response_123"}, - "chat_time": "2025-11-24T10:31:00Z", - } - ] - ], - ), -] - -# All test cases organized by category -TEST_CASES = { - "string": STRING_MESSAGE_CASES, - "chat": CHAT_MESSAGE_CASES, - "tool": TOOL_MESSAGE_CASES, - "multimodal": MULTIMODAL_MESSAGE_CASES, - "raw": RAW_INPUT_CASES, - "assistant": ASSISTANT_MESSAGE_CASES, -} - -# Flattened list of all test cases -ALL_TEST_CASES = {case.name: case for cases in TEST_CASES.values() for case in cases} - - -# ============================================================================ -# Utility Functions -# ============================================================================ - - -def print_textual_memory_item(item: TextualMemoryItem, prefix: str = "", max_length: int = 500): - """Print a memory item in a readable format.""" - print(f"{prefix}Memory ID: {item.id}") - print(f"{prefix}Memory Type: {item.metadata.memory_type}") - if item.metadata.tags: - print(f"{prefix}Tags: {item.metadata.tags}") - memory_preview = ( - item.memory[:max_length] + "..." if len(item.memory) > max_length else item.memory - ) - print(f"{prefix}Memory: {memory_preview}") - if item.metadata.key: - print(f"{prefix}Key: {item.metadata.key}") - if item.metadata.sources: - sources_count = len(item.metadata.sources) if isinstance(item.metadata.sources, list) else 1 - print(f"{prefix}Sources count: {sources_count}") - print() - - -def print_textual_memory_item_json(item: TextualMemoryItem, indent: int = 2): - """Print a memory item as formatted JSON.""" - data = item.to_dict() - if "metadata" in data and "embedding" in data["metadata"]: - embedding = data["metadata"]["embedding"] - if embedding: - data["metadata"]["embedding"] = f"[vector of {len(embedding)} dimensions]" - print(json.dumps(data, indent=indent, ensure_ascii=False)) - - -def get_reader_config() -> dict[str, Any]: - """ - Get reader configuration from environment variables. - - Returns: - Configuration dictionary for MultiModalStructMemReaderConfig - """ - openai_api_key = os.getenv("OPENAI_API_KEY") - openai_base_url = os.getenv("OPENAI_API_BASE", "https://api.openai.com/v1") - ollama_api_base = os.getenv("OLLAMA_API_BASE", "http://localhost:11434") - - # Get LLM backend and config - llm_backend = os.getenv("MEM_READER_LLM_BACKEND", "openai") - if llm_backend == "ollama": - llm_config = { - "backend": "ollama", - "config": { - "model_name_or_path": os.getenv("MEM_READER_LLM_MODEL", "qwen3:0.6b"), - "api_base": ollama_api_base, - "temperature": float(os.getenv("MEM_READER_LLM_TEMPERATURE", "0.0")), - "remove_think_prefix": os.getenv( - "MEM_READER_LLM_REMOVE_THINK_PREFIX", "true" - ).lower() - == "true", - "max_tokens": int(os.getenv("MEM_READER_LLM_MAX_TOKENS", "8192")), - }, - } - else: # openai - llm_config = { - "backend": "openai", - "config": { - "model_name_or_path": os.getenv("MEM_READER_LLM_MODEL", "gpt-4o-mini"), - "api_key": openai_api_key or os.getenv("MEMRADER_API_KEY", "EMPTY"), - "api_base": openai_base_url, - "temperature": float(os.getenv("MEM_READER_LLM_TEMPERATURE", "0.5")), - "remove_think_prefix": os.getenv( - "MEM_READER_LLM_REMOVE_THINK_PREFIX", "true" - ).lower() - == "true", - "max_tokens": int(os.getenv("MEM_READER_LLM_MAX_TOKENS", "8192")), - }, - } - - # Get embedder backend and config - embedder_backend = os.getenv( - "MEM_READER_EMBEDDER_BACKEND", os.getenv("MOS_EMBEDDER_BACKEND", "ollama") - ) - if embedder_backend == "universal_api": - embedder_config = { - "backend": "universal_api", - "config": { - "provider": os.getenv( - "MEM_READER_EMBEDDER_PROVIDER", os.getenv("MOS_EMBEDDER_PROVIDER", "openai") - ), - "api_key": os.getenv( - "MEM_READER_EMBEDDER_API_KEY", - os.getenv("MOS_EMBEDDER_API_KEY", openai_api_key or "sk-xxxx"), - ), - "model_name_or_path": os.getenv( - "MEM_READER_EMBEDDER_MODEL", - os.getenv("MOS_EMBEDDER_MODEL", "text-embedding-3-large"), - ), - "base_url": os.getenv( - "MEM_READER_EMBEDDER_API_BASE", - os.getenv("MOS_EMBEDDER_API_BASE", openai_base_url), - ), - }, - } - else: # ollama - embedder_config = { - "backend": "ollama", - "config": { - "model_name_or_path": os.getenv( - "MEM_READER_EMBEDDER_MODEL", - os.getenv("MOS_EMBEDDER_MODEL", "nomic-embed-text:latest"), - ), - "api_base": ollama_api_base, - }, - } - - # Get direct markdown hostnames from environment variable - direct_markdown_hostnames = None - env_hostnames = os.getenv("FILE_PARSER_DIRECT_MARKDOWN_HOSTNAMES", "139.196.232.20") - if env_hostnames: - direct_markdown_hostnames = [h.strip() for h in env_hostnames.split(",") if h.strip()] - - return { - "llm": llm_config, - "embedder": embedder_config, - "chunker": { - "backend": "sentence", - "config": { - "tokenizer_or_token_counter": "gpt2", - "chunk_size": 512, - "chunk_overlap": 128, - "min_sentences_per_chunk": 1, - }, - }, - "direct_markdown_hostnames": direct_markdown_hostnames, - } - - -def count_memories(memory_results: list[list[TextualMemoryItem]]) -> int: - """Count total number of memory items across all scenes.""" - return sum(len(mem_list) for mem_list in memory_results) - - -# ============================================================================ -# Main Functions -# ============================================================================ - - -def run_test_case( - test_case: TestCase, reader: MultiModalStructMemReader, mode: str = "fast", format: str = "text" -): - """ - Run a single test case. - - Args: - test_case: Test case to run - reader: MultiModalStructMemReader instance - mode: Processing mode ("fast" or "fine") - format: Output format ("text" or "json") - """ - print(f"\n{'=' * 80}") - print(f"Test Case: {test_case.name}") - print(f"Description: {test_case.description}") - print(f"Mode: {mode.upper()}") - print(f"{'=' * 80}\n") - - info = test_case.get_info() - start_time = time.time() - - try: - memory_results = reader.get_memory(test_case.scene_data, type="chat", info=info, mode=mode) - elapsed_time = time.time() - start_time - - total_count = count_memories(memory_results) - print(f"✅ Completed in {elapsed_time:.2f}s") - print(f"📊 Generated {total_count} memory items across {len(memory_results)} scenes\n") - - # Check expected count if provided - if test_case.expected_count and mode in test_case.expected_count: - expected = test_case.expected_count[mode] - if total_count == expected: - print(f"✅ Expected count matches: {expected}") - else: - print(f"⚠️ Expected {expected}, got {total_count}") - - # Print sample results - print("\nSample Results:") - print("-" * 80) - for scene_idx, mem_list in enumerate(memory_results[:3]): # Show first 3 scenes - if not mem_list: - continue - print(f"\nScene {scene_idx + 1}:") - for item_idx, item in enumerate(mem_list[:2]): # Show first 2 items per scene - print(f"\n [Item {item_idx + 1}]") - if format == "json": - print_textual_memory_item_json(item, indent=4) - else: - print_textual_memory_item(item, prefix=" ", max_length=300) - - except Exception as e: - print(f"❌ Error: {e}") - import traceback - - traceback.print_exc() - - -def run_all_test_cases(reader: MultiModalStructMemReader, mode: str = "fast", format: str = "text"): - """Run all test cases.""" - print(f"\n{'=' * 80}") - print(f"Running All Test Cases (Mode: {mode.upper()})") - print(f"{'=' * 80}\n") - - total_cases = len(ALL_TEST_CASES) - for idx, (name, test_case) in enumerate(ALL_TEST_CASES.items(), 1): - print(f"\n[{idx}/{total_cases}] Running: {name}") - run_test_case(test_case, reader, mode=mode, format=format) - - -def run_category( - category: str, reader: MultiModalStructMemReader, mode: str = "fast", format: str = "text" -): - """Run all test cases in a category.""" - if category not in TEST_CASES: - print(f"❌ Unknown category: {category}") - print(f"Available categories: {', '.join(TEST_CASES.keys())}") - return - - cases = TEST_CASES[category] - print(f"\n{'=' * 80}") - print(f"Running Category: {category.upper()} ({len(cases)} test cases)") - print(f"Mode: {mode.upper()}") - print(f"{'=' * 80}\n") - - for idx, test_case in enumerate(cases, 1): - print(f"\n[{idx}/{len(cases)}] {test_case.name}") - run_test_case(test_case, reader, mode=mode, format=format) - - -def compare_modes(test_case: TestCase, reader: MultiModalStructMemReader, format: str = "text"): - """Compare fast and fine modes for a test case.""" - print(f"\n{'=' * 80}") - print(f"Comparing Fast vs Fine Mode: {test_case.name}") - print(f"{'=' * 80}\n") - - info = test_case.get_info() - - # Fast mode - print("⚡ FAST Mode:") - print("-" * 80) - start_time = time.time() - fast_results = reader.get_memory(test_case.scene_data, type="chat", info=info, mode="fast") - fast_time = time.time() - start_time - fast_count = count_memories(fast_results) - print(f"Time: {fast_time:.2f}s, Items: {fast_count}") - - # Fine mode - print("\n🔄 FINE Mode:") - print("-" * 80) - start_time = time.time() - fine_results = reader.get_memory(test_case.scene_data, type="chat", info=info, mode="fine") - fine_time = time.time() - start_time - fine_count = count_memories(fine_results) - print(f"Time: {fine_time:.2f}s, Items: {fine_count}") - - # Comparison - print("\n📈 Comparison:") - print(f" Fast: {fast_time:.2f}s, {fast_count} items") - print(f" Fine: {fine_time:.2f}s, {fine_count} items") - if fast_time > 0: - print(f" Speed: {fine_time / fast_time:.1f}x difference") - - # Show samples - if format == "text": - print("\n--- Fast Mode Sample (first item) ---") - if fast_results and fast_results[0]: - print_textual_memory_item(fast_results[0][0], prefix=" ", max_length=300) - - print("\n--- Fine Mode Sample (first item) ---") - if fine_results and fine_results[0]: - print_textual_memory_item(fine_results[0][0], prefix=" ", max_length=300) - - -def list_test_cases(): - """List all available test cases.""" - print("\n" + "=" * 80) - print("Available Test Cases") - print("=" * 80 + "\n") - - for category, cases in TEST_CASES.items(): - print(f"📁 {category.upper()} ({len(cases)} cases):") - for case in cases: - print(f" • {case.name}: {case.description}") - print() - - -def main(): - """Main entry point.""" - parser = argparse.ArgumentParser( - description="Test MultiModalStructMemReader with various use cases", - formatter_class=argparse.RawDescriptionHelpFormatter, - epilog=""" -Examples: - # Run all test cases in fast mode - python multimodal_struct_reader.py --example all --mode fast - - # Run a specific test case - python multimodal_struct_reader.py --example chat_simple --mode fine - - # Run a category of test cases - python multimodal_struct_reader.py --example multimodal --mode fast - - # Compare fast vs fine mode - python multimodal_struct_reader.py --example chat_simple --compare - - # List all available test cases - python multimodal_struct_reader.py --list - - # Output in JSON format - python multimodal_struct_reader.py --example chat_simple --format json - """, - ) - - parser.add_argument( - "--example", - type=str, - default="oss_text_file", - help="Test case name, category name, or 'all' to run all cases (default: all)", - ) - parser.add_argument( - "--mode", - choices=["fast", "fine"], - default="fine", - help="Processing mode: fast (quick) or fine (with LLM) (default: fast)", - ) - parser.add_argument( - "--format", - choices=["text", "json"], - default="text", - help="Output format: text (readable) or json (structured) (default: text)", - ) - parser.add_argument( - "--compare", - action="store_true", - help="Compare fast and fine modes (only works with specific test case)", - ) - parser.add_argument( - "--list", - action="store_true", - help="List all available test cases and exit", - ) - parser.add_argument( - "--max-memory-length", - type=int, - default=500, - help="Maximum length of memory content to display (default: 500)", - ) - - args = parser.parse_args() - - # List test cases and exit - if args.list: - list_test_cases() - return - - # Initialize reader - print("Initializing MultiModalStructMemReader...") - try: - config_dict = get_reader_config() - reader_config = MultiModalStructMemReaderConfig.model_validate(config_dict) - reader = MultiModalStructMemReader(reader_config) - print("✅ Reader initialized\n") - except Exception as e: - print(f"❌ Failed to initialize reader: {e}") - import traceback - - traceback.print_exc() - return - - # Run test cases - if args.example == "all": - run_all_test_cases(reader, mode=args.mode, format=args.format) - elif args.example in ALL_TEST_CASES: - test_case = ALL_TEST_CASES[args.example] - if args.compare: - compare_modes(test_case, reader, format=args.format) - else: - run_test_case(test_case, reader, mode=args.mode, format=args.format) - elif args.example in TEST_CASES: - run_category(args.example, reader, mode=args.mode, format=args.format) - else: - print(f"❌ Unknown test case or category: {args.example}") - print("\nAvailable options:") - print(" Categories:", ", ".join(TEST_CASES.keys())) - print(" Test cases:", ", ".join(ALL_TEST_CASES.keys())) - print("\nUse --list to see all available test cases") - - -if __name__ == "__main__": - main() diff --git a/examples/mem_reader/parser/__init__.py b/examples/mem_reader/parser/__init__.py deleted file mode 100644 index 3a947ae89..000000000 --- a/examples/mem_reader/parser/__init__.py +++ /dev/null @@ -1 +0,0 @@ -"""Parser examples for different message types.""" diff --git a/examples/mem_reader/parser/config_utils.py b/examples/mem_reader/parser/config_utils.py deleted file mode 100644 index 225b8b5b4..000000000 --- a/examples/mem_reader/parser/config_utils.py +++ /dev/null @@ -1,132 +0,0 @@ -"""Shared configuration utilities for parser examples. - -This module provides configuration functions that match the configuration -logic in examples/mem_reader/multimodal_struct_reader.py. -""" - -import os - -from typing import Any - -from memos.configs.embedder import EmbedderConfigFactory -from memos.configs.llm import LLMConfigFactory -from memos.embedders.factory import EmbedderFactory -from memos.llms.factory import LLMFactory - - -def get_reader_config() -> dict[str, Any]: - """ - Get reader configuration from environment variables. - - Returns a dictionary that can be used to create MultiModalStructMemReaderConfig. - Matches the configuration logic in examples/mem_reader/multimodal_struct_reader.py. - - Returns: - Configuration dictionary with llm, embedder, and chunker configs - """ - openai_api_key = os.getenv("OPENAI_API_KEY") - openai_base_url = os.getenv("OPENAI_API_BASE", "https://api.openai.com/v1") - ollama_api_base = os.getenv("OLLAMA_API_BASE", "http://localhost:11434") - - # Get LLM backend and config - llm_backend = os.getenv("MEM_READER_LLM_BACKEND", "openai") - if llm_backend == "ollama": - llm_config = { - "backend": "ollama", - "config": { - "model_name_or_path": os.getenv("MEM_READER_LLM_MODEL", "qwen3:0.6b"), - "api_base": ollama_api_base, - "temperature": float(os.getenv("MEM_READER_LLM_TEMPERATURE", "0.0")), - "remove_think_prefix": os.getenv( - "MEM_READER_LLM_REMOVE_THINK_PREFIX", "true" - ).lower() - == "true", - "max_tokens": int(os.getenv("MEM_READER_LLM_MAX_TOKENS", "8192")), - }, - } - else: # openai - llm_config = { - "backend": "openai", - "config": { - "model_name_or_path": os.getenv("MEM_READER_LLM_MODEL", "gpt-4o-mini"), - "api_key": openai_api_key or os.getenv("MEMRADER_API_KEY", "EMPTY"), - "api_base": openai_base_url, - "temperature": float(os.getenv("MEM_READER_LLM_TEMPERATURE", "0.5")), - "remove_think_prefix": os.getenv( - "MEM_READER_LLM_REMOVE_THINK_PREFIX", "true" - ).lower() - == "true", - "max_tokens": int(os.getenv("MEM_READER_LLM_MAX_TOKENS", "8192")), - }, - } - - # Get embedder backend and config - embedder_backend = os.getenv( - "MEM_READER_EMBEDDER_BACKEND", os.getenv("MOS_EMBEDDER_BACKEND", "ollama") - ) - if embedder_backend == "universal_api": - embedder_config = { - "backend": "universal_api", - "config": { - "provider": os.getenv( - "MEM_READER_EMBEDDER_PROVIDER", os.getenv("MOS_EMBEDDER_PROVIDER", "openai") - ), - "api_key": os.getenv( - "MEM_READER_EMBEDDER_API_KEY", - os.getenv("MOS_EMBEDDER_API_KEY", openai_api_key or "sk-xxxx"), - ), - "model_name_or_path": os.getenv( - "MEM_READER_EMBEDDER_MODEL", - os.getenv("MOS_EMBEDDER_MODEL", "text-embedding-3-large"), - ), - "base_url": os.getenv( - "MEM_READER_EMBEDDER_API_BASE", - os.getenv("MOS_EMBEDDER_API_BASE", openai_base_url), - ), - }, - } - else: # ollama - embedder_config = { - "backend": "ollama", - "config": { - "model_name_or_path": os.getenv( - "MEM_READER_EMBEDDER_MODEL", - os.getenv("MOS_EMBEDDER_MODEL", "nomic-embed-text:latest"), - ), - "api_base": ollama_api_base, - }, - } - - return { - "llm": llm_config, - "embedder": embedder_config, - "chunker": { - "backend": "sentence", - "config": { - "tokenizer_or_token_counter": "gpt2", - "chunk_size": 512, - "chunk_overlap": 128, - "min_sentences_per_chunk": 1, - }, - }, - } - - -def init_embedder_and_llm(): - """ - Initialize embedder and LLM from environment variables. - - Returns: - Tuple of (embedder, llm) instances - """ - config_dict = get_reader_config() - - # Initialize embedder - embedder_config = EmbedderConfigFactory.model_validate(config_dict["embedder"]) - embedder = EmbedderFactory.from_config(embedder_config) - - # Initialize LLM - llm_config = LLMConfigFactory.model_validate(config_dict["llm"]) - llm = LLMFactory.from_config(llm_config) - - return embedder, llm diff --git a/examples/mem_reader/parser/example_assistant_parser.py b/examples/mem_reader/parser/example_assistant_parser.py deleted file mode 100644 index a77f04a68..000000000 --- a/examples/mem_reader/parser/example_assistant_parser.py +++ /dev/null @@ -1,94 +0,0 @@ -"""Example demonstrating AssistantParser usage. - -AssistantParser handles assistant messages in chat conversations. -""" - -import sys - -from pathlib import Path - -from dotenv import load_dotenv - -from memos.mem_reader.read_multi_modal.assistant_parser import AssistantParser - - -# Handle imports for both script and module usage -try: - from .config_utils import init_embedder_and_llm -except ImportError: - # When running as script, add parent directory to path - sys.path.insert(0, str(Path(__file__).parent)) - from config_utils import init_embedder_and_llm - -# Load environment variables -load_dotenv() - - -def main(): - """Demonstrate AssistantParser usage.""" - print("=== AssistantParser Example ===\n") - - # 1. Initialize embedder and LLM (using shared config) - embedder, llm = init_embedder_and_llm() - - # 3. Create AssistantParser - parser = AssistantParser(embedder=embedder, llm=llm) - - # 4. Example assistant messages - assistant_messages = [ - { - "role": "assistant", - "content": "I'm sorry to hear that you're feeling down. Would you like to talk about what's been going on?", - "chat_time": "2025-01-15T10:00:30", - "message_id": "msg_001", - }, - { - "role": "assistant", - "content": "Based on the document you provided, I can see several key points: 1) The project timeline, 2) Budget considerations, and 3) Resource allocation.", - "chat_time": "2025-01-15T10:05:30", - "message_id": "msg_002", - }, - { - "role": "assistant", - "content": "Here's a Python solution for your problem:\n```python\ndef solve_problem():\n return 'solution'\n```", - "chat_time": "2025-01-15T10:10:30", - "message_id": "msg_003", - }, - ] - - print("📝 Processing assistant messages:\n") - for i, message in enumerate(assistant_messages, 1): - print(f"Assistant Message {i}:") - print(f" Content: {message['content'][:60]}...") - - # Create source from assistant message - info = {"user_id": "user1", "session_id": "session1"} - source = parser.create_source(message, info) - - print(" ✅ Created SourceMessage:") - print(f" - Type: {source.type}") - print(f" - Role: {source.role}") - print(f" - Content: {source.content[:60]}...") - print(f" - Chat Time: {source.chat_time}") - print(f" - Message ID: {source.message_id}") - print() - - # Parse in fast mode - memory_items = parser.parse_fast(message, info) - print(f" 📊 Fast mode generated {len(memory_items)} memory item(s)") - if memory_items: - print(f" - Memory: {memory_items[0].memory[:60]}...") - print(f" - Memory Type: {memory_items[0].metadata.memory_type}") - print(f" - Tags: {memory_items[0].metadata.tags}") - print() - - # Rebuild assistant message from source - rebuilt = parser.rebuild_from_source(source) - print(f" 🔄 Rebuilt message: role={rebuilt['role']}, content={rebuilt['content'][:40]}...") - print() - - print("✅ AssistantParser example completed!") - - -if __name__ == "__main__": - main() diff --git a/examples/mem_reader/parser/example_file_content_parser.py b/examples/mem_reader/parser/example_file_content_parser.py deleted file mode 100644 index 06071a70c..000000000 --- a/examples/mem_reader/parser/example_file_content_parser.py +++ /dev/null @@ -1,132 +0,0 @@ -"""Example demonstrating FileContentParser usage. - -FileContentParser handles file content parts in multimodal messages (RawMessageList). -""" - -import sys - -from pathlib import Path - -from dotenv import load_dotenv - -from memos.configs.parser import ParserConfigFactory -from memos.mem_reader.read_multi_modal.file_content_parser import FileContentParser -from memos.parsers.factory import ParserFactory - - -# Handle imports for both script and module usage -try: - from .config_utils import init_embedder_and_llm -except ImportError: - # When running as script, add parent directory to path - sys.path.insert(0, str(Path(__file__).parent)) - from config_utils import init_embedder_and_llm - -# Load environment variables -load_dotenv() - - -def main(): - """Demonstrate FileContentParser usage.""" - print("=== FileContentParser Example ===\n") - - # 1. Initialize embedder and LLM (using shared config) - embedder, llm = init_embedder_and_llm() - - # 3. Initialize parser for file content parsing (optional) - try: - parser_config = ParserConfigFactory.model_validate( - { - "backend": "markitdown", - "config": {}, - } - ) - file_parser = ParserFactory.from_config(parser_config) - except Exception as e: - print(f"⚠️ Warning: Could not initialize file parser: {e}") - print(" FileContentParser will work without a parser, but file parsing will be limited.") - file_parser = None - - # 4. Create FileContentParser - parser = FileContentParser(embedder=embedder, llm=llm, parser=file_parser) - - # 5. Example file content parts - file_content_parts = [ - { - "type": "file", - "file": { - "filename": "document.pdf", - "file_id": "file_123", - "file_data": "This is the content extracted from the PDF file...", - }, - }, - { - "type": "file", - "file": { - "filename": "report.docx", - "file_id": "file_456", - "file_data": "Report content: Analysis of Q4 performance...", - }, - }, - { - "type": "file", - "file": { - "filename": "data.csv", - "file_id": "file_789", - "path": "/path/to/data.csv", # Alternative: using path instead of file_data - }, - }, - ] - - print("📝 Processing file content parts:\n") - for i, part in enumerate(file_content_parts, 1): - print(f"File Content Part {i}:") - file_info = part.get("file", {}) - print(f" Filename: {file_info.get('filename', 'unknown')}") - print(f" File ID: {file_info.get('file_id', 'N/A')}") - - # Create source from file content part - info = {"user_id": "user1", "session_id": "session1"} - source = parser.create_source(part, info) - - print(" ✅ Created SourceMessage:") - print(f" - Type: {source.type}") - print(f" - Doc Path: {source.doc_path}") - if source.content: - print(f" - Content: {source.content[:60]}...") - if hasattr(source, "original_part") and source.original_part: - print(" - Has original_part: Yes") - print() - - # Rebuild file content part from source - rebuilt = parser.rebuild_from_source(source) - print(" 🔄 Rebuilt part:") - print(f" - Type: {rebuilt['type']}") - print(f" - Filename: {rebuilt['file'].get('filename', 'N/A')}") - print() - - # 6. Example with actual file path (if parser is available) - if file_parser: - print("📄 Testing file parsing with actual file path:\n") - # Note: This is just an example - actual file parsing would require a real file - example_file_part = { - "type": "file", - "file": { - "filename": "example.txt", - "path": "examples/mem_reader/text1.txt", # Using existing test file - }, - } - - try: - source = parser.create_source(example_file_part, info) - print(f" ✅ Created SourceMessage for file: {source.doc_path}") - # The parser would parse the file content if the file exists - except Exception as e: - print(f" ⚠️ File parsing note: {e}") - print() - - print("✅ FileContentParser example completed!") - - -if __name__ == "__main__": - main() diff --git a/examples/mem_reader/parser/example_multi_modal_parser.py b/examples/mem_reader/parser/example_multi_modal_parser.py deleted file mode 100644 index 3638d8d5e..000000000 --- a/examples/mem_reader/parser/example_multi_modal_parser.py +++ /dev/null @@ -1,400 +0,0 @@ -"""Example demonstrating MultiModalParser parser selection. - -This example verifies that different input types correctly return -the corresponding parser instances. - -MessagesType Definition (from src/memos/types/general_types.py): - MessagesType = str | MessageList | RawMessageList - - Where: - - str: Simple string messages - - MessageList: list[ChatCompletionMessageParam] - ChatCompletionMessageParam = ( - ChatCompletionSystemMessageParam | - ChatCompletionUserMessageParam | - ChatCompletionAssistantMessageParam | - ChatCompletionToolMessageParam - ) - - RawMessageList: list[RawMessageDict] - RawMessageDict = ChatCompletionContentPartTextParam | File - - Note: User/Assistant messages can have multimodal content (list of parts): - - {"type": "text", "text": "..."} - - {"type": "file", "file": {...}} - - {"type": "image_url", "image_url": {...}} - - {"type": "input_audio", "input_audio": {...}} -""" - -import sys - -from pathlib import Path - -from dotenv import load_dotenv - -from memos.mem_reader.read_multi_modal.multi_modal_parser import MultiModalParser - - -# Add src directory to path for imports -project_root = Path(__file__).parent.parent.parent.parent -src_path = project_root / "src" -if str(src_path) not in sys.path: - sys.path.insert(0, str(src_path)) - - -# Handle imports for both script and module usage -try: - from .config_utils import init_embedder_and_llm -except ImportError: - # When running as script, add parent directory to path - sys.path.insert(0, str(Path(__file__).parent)) - from config_utils import init_embedder_and_llm - -# Load environment variables -load_dotenv() - - -def parser_selection(): - """Test that different input types return the correct parser.""" - print("=== MultiModalParser Parser Selection Test ===\n") - - # 1. Initialize embedder and LLM - embedder, llm = init_embedder_and_llm() - - # 2. Create MultiModalParser - parser = MultiModalParser(embedder=embedder, llm=llm) - - # 3. Test cases: different input types - test_cases = [ - # String input -> StringParser - { - "name": "String input", - "message": "This is a simple string message", - "expected_parser_type": "StringParser", - }, - # RawMessageList: text type -> TextContentParser - { - "name": "Text content part (RawMessageList)", - "message": {"type": "text", "text": "This is a text content part"}, - "expected_parser_type": "TextContentParser", - }, - # RawMessageList: file type -> FileContentParser - { - "name": "File content part (RawMessageList)", - "message": { - "type": "file", - "file": { - "filename": "example.pdf", - "file_data": "File content here", - }, - }, - "expected_parser_type": "FileContentParser", - }, - # RawMessageList: image_url type -> None (type_parsers uses "image" key, not "image_url") - { - "name": "Image content part (RawMessageList - image_url type)", - "message": { - "type": "image_url", - "image_url": { - "url": "https://example.com/image.jpg", - "detail": "auto", - }, - }, - "expected_parser_type": None, # type_parsers has "image" key, but message has "image_url" type - "should_return_none": True, - }, - # RawMessageList: input_audio type -> None (type_parsers uses "audio" key, not "input_audio") - { - "name": "Audio content part (RawMessageList - input_audio type)", - "message": { - "type": "input_audio", - "input_audio": { - "data": "base64_encoded_audio_data", - "format": "mp3", - }, - }, - "expected_parser_type": None, # type_parsers has "audio" key, but message has "input_audio" type - "should_return_none": True, - }, - # MessageList: system role -> SystemParser - { - "name": "System message", - "message": { - "role": "system", - "content": "You are a helpful assistant.", - }, - "expected_parser_type": "SystemParser", - }, - # MessageList: user role -> UserParser - { - "name": "User message (simple)", - "message": { - "role": "user", - "content": "Hello, how are you?", - }, - "expected_parser_type": "UserParser", - }, - # MessageList: user role with multimodal content -> UserParser - { - "name": "User message (multimodal with text and file)", - "message": { - "role": "user", - "content": [ - {"type": "text", "text": "What's in this image?"}, - {"type": "file", "file": {"filename": "image.jpg", "file_data": ""}}, - ], - }, - "expected_parser_type": "UserParser", - }, - # MessageList: user role with image_url content -> UserParser - { - "name": "User message (with image_url)", - "message": { - "role": "user", - "content": [ - {"type": "text", "text": "What's in this image?"}, - { - "type": "image_url", - "image_url": {"url": "https://example.com/image.jpg"}, - }, - ], - }, - "expected_parser_type": "UserParser", - }, - # MessageList: user role with input_audio content -> UserParser - { - "name": "User message (with input_audio)", - "message": { - "role": "user", - "content": [ - {"type": "text", "text": "Listen to this audio"}, - { - "type": "input_audio", - "input_audio": {"data": "base64_data", "format": "wav"}, - }, - ], - }, - "expected_parser_type": "UserParser", - }, - # MessageList: assistant role -> AssistantParser - { - "name": "Assistant message (simple)", - "message": { - "role": "assistant", - "content": "I'm doing well, thank you!", - }, - "expected_parser_type": "AssistantParser", - }, - # MessageList: assistant role with tool_calls -> AssistantParser - { - "name": "Assistant message (with tool_calls)", - "message": { - "role": "assistant", - "content": None, - "tool_calls": [ - { - "id": "call_123", - "type": "function", - "function": { - "name": "get_weather", - "arguments": '{"location": "Beijing"}', - }, - } - ], - }, - "expected_parser_type": "AssistantParser", - }, - # MessageList: tool role -> ToolParser - { - "name": "Tool message", - "message": { - "role": "tool", - "content": "Tool execution result", - "tool_call_id": "call_123", - }, - "expected_parser_type": "ToolParser", - }, - ] - - print("Testing parser selection for different input types:\n") - all_passed = True - - for i, test_case in enumerate(test_cases, 1): - message = test_case["message"] - expected_type = test_case.get("expected_parser_type") - test_name = test_case["name"] - should_return_none = test_case.get("should_return_none", False) - - # Get parser using internal method - selected_parser = parser._get_parser(message) - - # Handle cases where None is expected - if should_return_none or expected_type is None: - if selected_parser is None: - print(f"✅ Test {i}: {test_name}") - print(" Expected: None (parser not implemented yet or not found)") - print(" Got: None") - if expected_type: - print(f" Note: {expected_type} is not yet implemented") - else: - print(f"⚠️ Test {i}: {test_name}") - print(" Expected: None") - print(f" Got: {type(selected_parser).__name__}") - print(" Note: Parser found but may not be fully implemented") - print() - continue - - # Check if parser was found - if selected_parser is None: - print(f"❌ Test {i}: {test_name}") - print(f" Expected: {expected_type}") - print(" Got: None (parser not found)") - print(f" Message: {message}\n") - all_passed = False - continue - - # Get actual parser type name - actual_type = type(selected_parser).__name__ - - # Verify parser type - if actual_type == expected_type: - print(f"✅ Test {i}: {test_name}") - print(f" Expected: {expected_type}") - print(f" Got: {actual_type}") - print(f" Parser instance: {selected_parser}") - else: - print(f"❌ Test {i}: {test_name}") - print(f" Expected: {expected_type}") - print(f" Got: {actual_type}") - print(f" Message: {message}") - all_passed = False - print() - - # Test edge cases - print("\n=== Testing Edge Cases ===\n") - - edge_cases = [ - { - "name": "Unknown message type (not dict, not str)", - "message": 12345, - "should_return_none": True, - }, - { - "name": "Dict without type or role", - "message": {"content": "Some content"}, - "should_return_none": True, - }, - { - "name": "Unknown type in RawMessageList", - "message": {"type": "unknown_type", "data": "some data"}, - "should_return_none": True, - }, - { - "name": "Unknown role in MessageList", - "message": {"role": "unknown_role", "content": "some content"}, - "should_return_none": True, - }, - { - "name": "List of messages (MessageList - not handled by _get_parser)", - "message": [ - {"role": "user", "content": "Message 1"}, - {"role": "assistant", "content": "Message 2"}, - ], - "should_return_none": True, # Lists are handled in parse(), not _get_parser() - }, - { - "name": "List of RawMessageList items (not handled by _get_parser)", - "message": [ - {"type": "text", "text": "Text content 1"}, - {"type": "file", "file": {"filename": "doc.pdf", "file_data": ""}}, - ], - "should_return_none": True, # Lists are handled in parse(), not _get_parser() - }, - ] - - for i, test_case in enumerate(edge_cases, 1): - message = test_case["message"] - should_return_none = test_case["should_return_none"] - test_name = test_case["name"] - - selected_parser = parser._get_parser(message) - - if should_return_none: - if selected_parser is None: - print(f"✅ Edge Case {i}: {test_name}") - print(" Correctly returned None") - else: - print(f"❌ Edge Case {i}: {test_name}") - print(" Expected: None") - print(f" Got: {type(selected_parser).__name__}") - all_passed = False - else: - if selected_parser is not None: - print(f"✅ Edge Case {i}: {test_name}") - print(f" Got parser: {type(selected_parser).__name__}") - else: - print(f"❌ Edge Case {i}: {test_name}") - print(" Expected: Parser") - print(" Got: None") - all_passed = False - print() - - # Summary - print("=" * 60) - if all_passed: - print("✅ All tests passed! Parser selection is working correctly.") - else: - print("❌ Some tests failed. Please check the output above.") - print("=" * 60) - - -def parser_instances(): - """Test that parser instances are correctly initialized.""" - print("\n=== Parser Instance Verification ===\n") - - embedder, llm = init_embedder_and_llm() - parser = MultiModalParser(embedder=embedder, llm=llm) - - # Verify all parser instances are initialized - parsers_to_check = { - "string_parser": "StringParser", - "system_parser": "SystemParser", - "user_parser": "UserParser", - "assistant_parser": "AssistantParser", - "tool_parser": "ToolParser", - "text_content_parser": "TextContentParser", - "file_content_parser": "FileContentParser", - } - - print("Checking parser instance initialization:\n") - all_initialized = True - - for attr_name, expected_type in parsers_to_check.items(): - parser_instance = getattr(parser, attr_name, None) - if parser_instance is None: - print(f"❌ {attr_name}: Not initialized") - all_initialized = False - else: - actual_type = type(parser_instance).__name__ - if actual_type == expected_type: - print(f"✅ {attr_name}: {actual_type}") - else: - print(f"❌ {attr_name}: Expected {expected_type}, got {actual_type}") - all_initialized = False - - print() - if all_initialized: - print("✅ All parser instances are correctly initialized!") - else: - print("❌ Some parser instances are missing or incorrect.") - print() - - -def main(): - """Run all tests.""" - parser_selection() - parser_instances() - print("\n✅ MultiModalParser example completed!") - - -if __name__ == "__main__": - main() diff --git a/examples/mem_reader/parser/example_string_parser.py b/examples/mem_reader/parser/example_string_parser.py deleted file mode 100644 index 3ec658a0e..000000000 --- a/examples/mem_reader/parser/example_string_parser.py +++ /dev/null @@ -1,66 +0,0 @@ -"""Example demonstrating StringParser usage. - -StringParser handles simple string messages that need to be converted to memory items. -""" - -import sys - -from pathlib import Path - -from dotenv import load_dotenv - -from memos.mem_reader.read_multi_modal.string_parser import StringParser - - -# Handle imports for both script and module usage -try: - from .config_utils import init_embedder_and_llm -except ImportError: - # When running as script, add parent directory to path - sys.path.insert(0, str(Path(__file__).parent)) - from config_utils import init_embedder_and_llm - -# Load environment variables -load_dotenv() - - -def main(): - """Demonstrate StringParser usage.""" - print("=== StringParser Example ===\n") - - # 1. Initialize embedder and LLM (using shared config) - embedder, llm = init_embedder_and_llm() - - # 3. Create StringParser - parser = StringParser(embedder=embedder, llm=llm) - - # 4. Example string messages - string_messages = [ - "This is a simple text message that needs to be parsed.", - "Another string message for processing.", - "StringParser handles plain text strings and converts them to SourceMessage objects.", - ] - - print("📝 Processing string messages:\n") - for i, message in enumerate(string_messages, 1): - print(f"Message {i}: {message[:50]}...") - - # Create source from string - info = {"user_id": "user1", "session_id": "session1"} - source = parser.create_source(message, info) - - print(" ✅ Created SourceMessage:") - print(f" - Type: {source.type}") - print(f" - Content: {source.content[:50]}...") - print() - - # Rebuild string from source - rebuilt = parser.rebuild_from_source(source) - print(f" 🔄 Rebuilt string: {rebuilt[:50]}...") - print() - - print("✅ StringParser example completed!") - - -if __name__ == "__main__": - main() diff --git a/examples/mem_reader/parser/example_system_parser.py b/examples/mem_reader/parser/example_system_parser.py deleted file mode 100644 index bc684a32b..000000000 --- a/examples/mem_reader/parser/example_system_parser.py +++ /dev/null @@ -1,158 +0,0 @@ -"""Example demonstrating SystemParser usage. - -SystemParser handles system messages in chat conversations. -Note: System messages support multimodal content, but only text parts are allowed -(not file, image_url, or input_audio like user messages). -""" - -import sys - -from pathlib import Path - -from dotenv import load_dotenv - - -try: - from .print_utils import pretty_print_dict -except ImportError: - # Fallback if print_utils is not available - def pretty_print_dict(d): - import json - - print(json.dumps(d, indent=2, ensure_ascii=False)) - - -from memos.mem_reader.read_multi_modal.system_parser import SystemParser - - -# Handle imports for both script and module usage -try: - from .config_utils import init_embedder_and_llm -except ImportError: - # When running as script, add parent directory to path - sys.path.insert(0, str(Path(__file__).parent)) - from config_utils import init_embedder_and_llm - -# Load environment variables -load_dotenv() - - -def main(): - """Demonstrate SystemParser usage.""" - print("=== SystemParser Example ===\n") - - # 1. Initialize embedder and LLM (using shared config) - embedder, llm = init_embedder_and_llm() - - # 3. Create SystemParser - parser = SystemParser(embedder=embedder, llm=llm) - - # 4. Example system messages (simple text) - simple_system_message = { - "role": "system", - "content": "You are a helpful assistant that provides clear and concise answers.", - "chat_time": "2025-01-15T10:00:00", - "message_id": "msg_001", - } - - print("📝 Example 1: Simple text system message\n") - pretty_print_dict(simple_system_message) - - info = {"user_id": "user1", "session_id": "session1"} - source = parser.create_source(simple_system_message, info) - - print(" ✅ Created SourceMessage:") - print(f" - Type: {source.type}") - print(f" - Role: {source.role}") - print(f" - Content: {source.content[:60]}...") - print(f" - Chat Time: {source.chat_time}") - print(f" - Message ID: {source.message_id}") - print() - - # Parse in fast mode - memory_items = parser.parse_fast(simple_system_message, info) - print(f" 📊 Fast mode generated {len(memory_items)} memory item(s)") - if memory_items: - print(f" - Memory: {memory_items[0].memory[:60]}...") - print(f" - Memory Type: {memory_items[0].metadata.memory_type}") - print(f" - Tags: {memory_items[0].metadata.tags}") - print() - - # 5. Example multimodal system message (multiple text parts) - # Note: System messages only support text parts, not file/image/audio - multimodal_system_message = { - "role": "system", - "content": [ - {"type": "text", "text": "You are a helpful assistant."}, - {"type": "text", "text": "Always provide clear and concise answers."}, - {"type": "text", "text": "If you don't know something, say so."}, - ], - "chat_time": "2025-01-15T10:05:00", - "message_id": "msg_002", - } - - print("📝 Example 2: Multimodal system message (multiple text parts)\n") - pretty_print_dict(multimodal_system_message) - print(f"Message contains {len(multimodal_system_message['content'])} text parts") - - sources = parser.create_source(multimodal_system_message, info) - if isinstance(sources, list): - print(f" ✅ Created {len(sources)} SourceMessage(s):") - for i, src in enumerate(sources, 1): - print(f" [{i}] Type: {src.type}, Role: {src.role}") - print(f" Content: {src.content[:50]}...") - else: - print(f" ✅ Created SourceMessage: Type={sources.type}") - print() - - # Parse in fast mode - memory_items = parser.parse_fast(multimodal_system_message, info) - print(f" 📊 Fast mode generated {len(memory_items)} memory item(s)") - if memory_items: - print(f" - Memory: {memory_items[0].memory[:60]}...") - print(f" - Memory Type: {memory_items[0].metadata.memory_type}") - print(f" - Tags: {memory_items[0].metadata.tags}") - # Show sources from memory item - if memory_items[0].metadata.sources: - print(f" - Sources: {len(memory_items[0].metadata.sources)} SourceMessage(s)") - print() - - # 6. Example with structured system instructions - structured_system_message = { - "role": "system", - "content": [ - { - "type": "text", - "text": "You are a coding assistant specialized in Python programming.", - }, - {"type": "text", "text": "Always write clean, well-documented code."}, - {"type": "text", "text": "Explain your reasoning when providing solutions."}, - ], - "chat_time": "2025-01-15T10:10:00", - "message_id": "msg_003", - } - - print("📝 Example 3: Structured system instructions (multiple text parts)\n") - pretty_print_dict(structured_system_message) - - sources = parser.create_source(structured_system_message, info) - if isinstance(sources, list): - print(f" ✅ Created {len(sources)} SourceMessage(s):") - for i, src in enumerate(sources, 1): - print(f" [{i}] Type: {src.type}, Role: {src.role}") - print(f" Content: {src.content[:50]}...") - print() - - # Rebuild examples - print("🔄 Rebuilding messages from sources:\n") - if isinstance(sources, list) and sources: - rebuilt = parser.rebuild_from_source(sources[0]) - else: - rebuilt = parser.rebuild_from_source(source) - if rebuilt: - pretty_print_dict(rebuilt) - print("✅ SystemParser example completed!") - - -if __name__ == "__main__": - main() diff --git a/examples/mem_reader/parser/example_text_content_parser.py b/examples/mem_reader/parser/example_text_content_parser.py deleted file mode 100644 index 1eb64d033..000000000 --- a/examples/mem_reader/parser/example_text_content_parser.py +++ /dev/null @@ -1,72 +0,0 @@ -"""Example demonstrating TextContentParser usage. - -TextContentParser handles text content parts in multimodal messages (RawMessageList). -""" - -import sys - -from pathlib import Path - -from dotenv import load_dotenv - -from memos.mem_reader.read_multi_modal.text_content_parser import TextContentParser - - -# Handle imports for both script and module usage -try: - from .config_utils import init_embedder_and_llm -except ImportError: - # When running as script, add parent directory to path - sys.path.insert(0, str(Path(__file__).parent)) - from config_utils import init_embedder_and_llm - -# Load environment variables -load_dotenv() - - -def main(): - """Demonstrate TextContentParser usage.""" - print("=== TextContentParser Example ===\n") - - # 1. Initialize embedder and LLM (using shared config) - embedder, llm = init_embedder_and_llm() - - # 3. Create TextContentParser - parser = TextContentParser(embedder=embedder, llm=llm) - - # 4. Example text content parts - text_content_parts = [ - {"type": "text", "text": "This is a simple text content part."}, - {"type": "text", "text": "TextContentParser handles text parts in multimodal messages."}, - { - "type": "text", - "text": "This parser is used when processing RawMessageList items that contain text content.", - }, - ] - - print("📝 Processing text content parts:\n") - for i, part in enumerate(text_content_parts, 1): - print(f"Text Content Part {i}:") - print(f" Text: {part['text'][:60]}...") - - # Create source from text content part - info = {"user_id": "user1", "session_id": "session1"} - source = parser.create_source(part, info) - - print(" ✅ Created SourceMessage:") - print(f" - Type: {source.type}") - print(f" - Content: {source.content[:60]}...") - if hasattr(source, "original_part") and source.original_part: - print(" - Has original_part: Yes") - print() - - # Rebuild text content part from source - rebuilt = parser.rebuild_from_source(source) - print(f" 🔄 Rebuilt part: type={rebuilt['type']}, text={rebuilt['text'][:40]}...") - print() - - print("✅ TextContentParser example completed!") - - -if __name__ == "__main__": - main() diff --git a/examples/mem_reader/parser/example_tool_parser.py b/examples/mem_reader/parser/example_tool_parser.py deleted file mode 100644 index bf3f4e333..000000000 --- a/examples/mem_reader/parser/example_tool_parser.py +++ /dev/null @@ -1,101 +0,0 @@ -"""Example demonstrating ToolParser usage. - -ToolParser handles tool/function call messages in chat conversations. -""" - -import sys - -from pathlib import Path - -from dotenv import load_dotenv - -from memos.mem_reader.read_multi_modal.tool_parser import ToolParser - - -# Handle imports for both script and module usage -try: - from .config_utils import init_embedder_and_llm -except ImportError: - # When running as script, add parent directory to path - sys.path.insert(0, str(Path(__file__).parent)) - from config_utils import init_embedder_and_llm - -# Load environment variables -load_dotenv() - - -def main(): - """Demonstrate ToolParser usage.""" - print("=== ToolParser Example ===\n") - - # 1. Initialize embedder and LLM (using shared config) - embedder, llm = init_embedder_and_llm() - - # 3. Create ToolParser - parser = ToolParser(embedder=embedder, llm=llm) - - # 4. Example tool messages - tool_messages = [ - { - "role": "tool", - "content": '{"result": "Weather in New York: 72°F, sunny"}', - "tool_call_id": "call_abc123", - "chat_time": "2025-01-15T10:00:30", - "message_id": "msg_001", - }, - { - "role": "tool", - "content": '{"status": "success", "data": {"items": [1, 2, 3]}}', - "tool_call_id": "call_def456", - "chat_time": "2025-01-15T10:05:30", - "message_id": "msg_002", - }, - { - "role": "tool", - "content": "Database query executed successfully. Retrieved 5 records.", - "tool_call_id": "call_ghi789", - "chat_time": "2025-01-15T10:10:30", - "message_id": "msg_003", - }, - ] - - print("📝 Processing tool messages:\n") - for i, message in enumerate(tool_messages, 1): - print(f"Tool Message {i}:") - print(f" Content: {message['content'][:60]}...") - print(f" Tool Call ID: {message['tool_call_id']}") - - # Create source from tool message - info = {"user_id": "user1", "session_id": "session1"} - source = parser.create_source(message, info) - - print(" ✅ Created SourceMessage:") - print(f" - Type: {source.type}") - print(f" - Role: {source.role}") - print(f" - Content: {source.content[:60]}...") - print(f" - Chat Time: {source.chat_time}") - print(f" - Message ID: {source.message_id}") - print() - - # Parse in fast mode - memory_items = parser.parse_fast(message, info) - print(f" 📊 Fast mode generated {len(memory_items)} memory item(s)") - if memory_items: - print(f" - Memory: {memory_items[0].memory[:60]}...") - print(f" - Memory Type: {memory_items[0].metadata.memory_type}") - print(f" - Tags: {memory_items[0].metadata.tags}") - print() - - # Rebuild tool message from source - rebuilt = parser.rebuild_from_source(source) - print(" 🔄 Rebuilt message:") - print(f" - Role: {rebuilt['role']}") - print(f" - Tool Call ID: {rebuilt.get('tool_call_id', 'N/A')}") - print(f" - Content: {rebuilt['content'][:40]}...") - print() - - print("✅ ToolParser example completed!") - - -if __name__ == "__main__": - main() diff --git a/examples/mem_reader/parser/example_user_parser.py b/examples/mem_reader/parser/example_user_parser.py deleted file mode 100644 index 78a75b94f..000000000 --- a/examples/mem_reader/parser/example_user_parser.py +++ /dev/null @@ -1,135 +0,0 @@ -"""Example demonstrating UserParser usage. - -UserParser handles user messages, including multimodal messages with text, files, images, etc. -""" - -import sys - -from pathlib import Path - -from dotenv import load_dotenv -from print_utils import pretty_print_dict - -from memos.mem_reader.read_multi_modal.user_parser import UserParser - - -# Handle imports for both script and module usage -try: - from .config_utils import init_embedder_and_llm -except ImportError: - # When running as script, add parent directory to path - sys.path.insert(0, str(Path(__file__).parent)) - from config_utils import init_embedder_and_llm - -# Load environment variables -load_dotenv() - - -def main(): - """Demonstrate UserParser usage.""" - print("=== UserParser Example ===\n") - - # 1. Initialize embedder and LLM (using shared config) - embedder, llm = init_embedder_and_llm() - - # 3. Create UserParser - parser = UserParser(embedder=embedder, llm=llm) - - # 4. Example user messages (simple text) - simple_user_message = { - "role": "user", - "content": "I'm feeling a bit down today. Can you help me?", - "chat_time": "2025-01-15T10:00:00", - "message_id": "msg_001", - } - - print("📝 Example 1: Simple text user message\n") - pretty_print_dict(simple_user_message) - - info = {"user_id": "user1", "session_id": "session1"} - # Parse in fast mode - memory_items = parser.parse_fast(simple_user_message, info) - print(f" 📊 Fast mode generated {len(memory_items)} memory item(s)") - if memory_items: - print(f" - Memory: {memory_items[0].memory[:60]}...") - print(f" - Memory Type: {memory_items[0].metadata.memory_type}") - print() - - # 5. Example multimodal user message (text + file) - multimodal_user_message = { - "role": "user", - "content": [ - {"type": "text", "text": "Please analyze this document:"}, - { - "type": "file", - "file": { - "filename": "report.pdf", - "file_id": "file_123", - "file_data": "This is the content of the PDF file...", - }, - }, - ], - "chat_time": "2025-01-15T10:05:00", - "message_id": "msg_002", - } - - print("📝 Example 2: Multimodal user message (text + file)\n") - pretty_print_dict(multimodal_user_message) - print(f"Message contains {len(multimodal_user_message['content'])} parts") - - # Parse in fast mode - memory_items = parser.parse_fast(multimodal_user_message, info) - print(f" 📊 Fast mode generated {len(memory_items)} memory item(s)") - for memory_item in memory_items: - sources = memory_item.metadata.sources - print(f" ✅ Created {len(sources)} SourceMessage(s):") - for i, src in enumerate(sources, 1): - print(f" [{i}] Type: {src.type}, Role: {src.role}") - if src.type == "text": - print(f" Content: {src.content[:50]}...") - elif src.type == "file": - print(f" Doc Path: {src.doc_path}") - print() - - # 6. Example with image_url (future support) - image_user_message = { - "role": "user", - "content": [ - {"type": "text", "text": "What's in this image?"}, - { - "type": "image_url", - "image_url": {"url": "https://example.com/image.jpg"}, - }, - ], - "chat_time": "2025-01-15T10:10:00", - "message_id": "msg_003", - } - print("📝 Example 3: User message with image\n") - print(f"Message contains {len(image_user_message['content'])} parts") - pretty_print_dict(image_user_message) - - # Parse in fast mode - memory_items = parser.parse_fast(image_user_message, info) - print(f" 📊 Fast mode generated {len(memory_items)} memory item(s)") - for memory_item in memory_items: - sources = memory_item.metadata.sources - print(f" ✅ Created {len(sources)} SourceMessage(s):") - for i, src in enumerate(sources, 1): - print(f" [{i}] Type: {src.type}, Role: {src.role}") - if src.type == "text": - print(f" Content: {src.content[:50]}...") - elif src.type == "file": - print(f" Doc Path: {src.doc_path}") - elif src.type == "image": - print(f" Image Path: {src.image_path}") - - # Rebuild examples - print("🔄 Rebuilding messages from sources:\n") - rebuilt_simple = parser.rebuild_from_source(sources[1]) - if rebuilt_simple: - pretty_print_dict(rebuilt_simple) - print("✅ UserParser example completed!") - - -if __name__ == "__main__": - main() diff --git a/examples/mem_reader/parser/print_utils.py b/examples/mem_reader/parser/print_utils.py deleted file mode 100644 index 5eba1fa76..000000000 --- a/examples/mem_reader/parser/print_utils.py +++ /dev/null @@ -1,11 +0,0 @@ -import pprint - - -def pretty_print_dict(d: dict): - text = pprint.pformat(d, indent=2, width=120) - border = "═" * (max(len(line) for line in text.split("\n")) + 4) - - print(f"╔{border}╗") - for line in text.split("\n"): - print(f"║ {line.ljust(len(border) - 2)} ║") - print(f"╚{border}╝") diff --git a/examples/mem_reader/parser_demos/__init__.py b/examples/mem_reader/parser_demos/__init__.py new file mode 100644 index 000000000..e69de29bb diff --git a/examples/mem_reader/parser_demos/_base.py b/examples/mem_reader/parser_demos/_base.py new file mode 100644 index 000000000..883578a21 --- /dev/null +++ b/examples/mem_reader/parser_demos/_base.py @@ -0,0 +1,76 @@ +"""Base class and utilities for parser demos.""" + +from typing import Any + +from examples.mem_reader.builders import build_llm_and_embedder +from examples.mem_reader.utils import pretty_print_dict +from memos.memories.textual.item import SourceMessage + + +class BaseParserDemo: + """Base class for all parser demos.""" + + def __init__(self): + print(f"\n🚀 Initializing {self.__class__.__name__}...") + self.embedder, self.llm = build_llm_and_embedder() + self.parser = self.create_parser() + print("✅ Initialization complete.\n") + + def create_parser(self): + """Create and return the specific parser instance.""" + raise NotImplementedError + + def run(self): + """Run the main demo logic.""" + raise NotImplementedError + + def demo_source_creation( + self, message: Any, info: dict, **kwargs + ) -> SourceMessage | list[SourceMessage]: + """Demonstrate creating a SourceMessage from raw input.""" + print(f"📝 Creating SourceMessage from: {str(message)[:100]}...") + source = self.parser.create_source(message, info, **kwargs) + + if isinstance(source, list): + print(f" ✅ Created {len(source)} SourceMessage(s)") + for i, s in enumerate(source): + print(f" [{i}] Type: {s.type}, Role: {getattr(s, 'role', 'N/A')}") + else: + print(" ✅ Created SourceMessage:") + print(f" - Type: {source.type}") + if hasattr(source, "role"): + print(f" - Role: {source.role}") + if source.content: + print(f" - Content: {str(source.content)[:60]}...") + + return source + + def demo_rebuild(self, source: SourceMessage | list[SourceMessage]): + """Demonstrate rebuilding raw message from SourceMessage.""" + print("\n🔄 Rebuilding message from source...") + + # Handle list of sources (take first one for demo if it's a list) + src_to_rebuild = source[0] if isinstance(source, list) else source + + rebuilt = self.parser.rebuild_from_source(src_to_rebuild) + print(" ✅ Rebuilt result:") + if isinstance(rebuilt, dict): + pretty_print_dict(rebuilt) + else: + print(f" {rebuilt}") + + def demo_parse_fast(self, message: Any, info: dict): + """Demonstrate fast parsing (if supported).""" + if not hasattr(self.parser, "parse_fast"): + return + + print("\n⚡️ Running parse_fast...") + try: + memory_items = self.parser.parse_fast(message, info) + print(f" 📊 Generated {len(memory_items)} memory item(s)") + if memory_items: + item = memory_items[0] + print(f" - Memory: {item.memory[:60]}...") + print(f" - Type: {item.metadata.memory_type}") + except Exception as e: + print(f" ⚠️ parse_fast not applicable or failed: {e}") diff --git a/examples/mem_reader/parser_demos/demo_assistant.py b/examples/mem_reader/parser_demos/demo_assistant.py new file mode 100644 index 000000000..7c6a0a1e8 --- /dev/null +++ b/examples/mem_reader/parser_demos/demo_assistant.py @@ -0,0 +1,44 @@ +"""Demo for AssistantParser.""" + +from examples.mem_reader.samples import ASSISTANT_MESSAGE_CASES +from memos.mem_reader.read_multi_modal.assistant_parser import AssistantParser + +from ._base import BaseParserDemo + + +class AssistantParserDemo(BaseParserDemo): + def create_parser(self): + parser = AssistantParser(embedder=self.embedder, llm=self.llm) + + # Workaround: AssistantParser.rebuild_from_source is empty in src. + # Patch it to return content for demo visualization, aligning with legacy behavior. + original_rebuild = parser.rebuild_from_source + + def patched_rebuild(source): + if source.role == "assistant": + # Only handling simple text content as per legacy example scope + return { + "role": "assistant", + "content": source.content, + } + return original_rebuild(source) + + parser.rebuild_from_source = patched_rebuild + return parser + + def run(self): + print("=== AssistantParser Demo ===") + + info = {"user_id": "user1", "session_id": "session1"} + + for case in ASSISTANT_MESSAGE_CASES: + print(f"\n--- Case: {case.description} ---") + for msg in case.scene_data: + source = self.demo_source_creation(msg, info) + self.demo_rebuild(source) + self.demo_parse_fast(msg, info) + + +if __name__ == "__main__": + demo = AssistantParserDemo() + demo.run() diff --git a/examples/mem_reader/parser_demos/demo_file_content.py b/examples/mem_reader/parser_demos/demo_file_content.py new file mode 100644 index 000000000..f0d11b15f --- /dev/null +++ b/examples/mem_reader/parser_demos/demo_file_content.py @@ -0,0 +1,68 @@ +"""Demo for FileContentParser.""" + +from examples.mem_reader.builders import build_file_parser +from examples.mem_reader.samples import FILE_CONTENT_PARTS, FILE_CONTENT_REAL_FILE_PART +from memos.mem_reader.read_multi_modal.file_content_parser import FileContentParser + +from ._base import BaseParserDemo + + +class FileContentParserDemo(BaseParserDemo): + def create_parser(self): + # Initialize the underlying file parser (MarkItDown) + file_parser_impl = build_file_parser() + + return FileContentParser( + embedder=self.embedder, + llm=self.llm, + parser=file_parser_impl, + ) + + def run(self): + print("=== FileContentParser Demo ===") + + info = {"user_id": "user1", "session_id": "session1"} + + print("📝 Processing file content parts:\n") + for i, part in enumerate(FILE_CONTENT_PARTS, 1): + print(f"File Content Part {i}:") + file_info = part.get("file", {}) + print(f" Filename: {file_info.get('filename', 'unknown')}") + print(f" File ID: {file_info.get('file_id', 'N/A')}") + + # Create source from file content part + source = self.parser.create_source(part, info) + + print(" ✅ Created SourceMessage:") + print(f" - Type: {source.type}") + print(f" - Doc Path: {source.doc_path}") + if source.content: + print(f" - Content: {source.content[:60]}...") + if hasattr(source, "original_part") and source.original_part: + print(" - Has original_part: Yes") + print() + + # Rebuild file content part from source + rebuilt = self.parser.rebuild_from_source(source) + print(" 🔄 Rebuilt part:") + print(f" - Type: {rebuilt.get('type')}") + print(f" - Filename: {rebuilt.get('file', {}).get('filename', 'N/A')}") + + print() + + # 6. Example with actual file path (if parser is available) + if getattr(self.parser, "parser", None): + print("📄 Testing file parsing with actual file path:\n") + + try: + source = self.parser.create_source(FILE_CONTENT_REAL_FILE_PART, info) + print(f" ✅ Created SourceMessage for file: {source.doc_path}") + # The parser would parse the file content if the file exists + except Exception as e: + print(f" ⚠️ File parsing note: {e}") + print() + + +if __name__ == "__main__": + demo = FileContentParserDemo() + demo.run() diff --git a/examples/mem_reader/parser_demos/demo_image.py b/examples/mem_reader/parser_demos/demo_image.py new file mode 100644 index 000000000..a3ff3b334 --- /dev/null +++ b/examples/mem_reader/parser_demos/demo_image.py @@ -0,0 +1,87 @@ +"""Demo for ImageParser.""" + +import base64 +import copy + +from pathlib import Path + +from examples.mem_reader.samples import IMAGE_MESSAGE_CASES +from memos.mem_reader.read_multi_modal.image_parser import ImageParser + +from ._base import BaseParserDemo + + +class ImageParserDemo(BaseParserDemo): + def create_parser(self): + return ImageParser(embedder=self.embedder, llm=self.llm) + + def run(self): + print("🚀 Initializing ImageParserDemo...") + print("✅ Initialization complete.") + print("=== ImageParser Demo ===\n") + + info = {"user_id": "user1", "session_id": "session1"} + + test_cases = copy.deepcopy(IMAGE_MESSAGE_CASES) + + # Add Local Image (Base64) if exists + local_img_path = Path(__file__).parent.parent / "test_image.png" + if local_img_path.exists(): + with open(local_img_path, "rb") as f: + b64_data = base64.b64encode(f.read()).decode("utf-8") + test_cases.append( + { + "type": "image_url", + "image_url": { + "url": f"data:image/png;base64,{b64_data}", + "detail": "auto", + }, + "_note": "Local Image (Base64)", + } + ) + + for i, msg in enumerate(test_cases, 1): + print(f"--- Case {i}: Image URL message ---") + + # 1. Create SourceMessage + print(f"📝 Creating SourceMessage from: {msg}") + source = self.parser.create_source(msg, info) + print(" ✅ Created SourceMessage:") + print(f" - Type: {source.type}") + print(f" - URL: {getattr(source, 'url', 'N/A')}") + + # 2. Rebuild from Source + print("🔄 Rebuilding message from source...") + rebuilt = self.parser.rebuild_from_source(source) + print(f" ✅ Rebuilt result: {rebuilt}") + + # 3. Fast Parse (Expected Empty) + print("⚡️ Running parse_fast (expecting empty)...") + fast_results = self.parser.parse_fast(msg, info) + if not fast_results: + print(" ✅ Got empty list as expected (images require fine mode).") + else: + print(f" ⚠️ Unexpected fast results: {len(fast_results)} items") + + # 4. Fine Parse (Vision Model) + print("🧠 Running parse_fine (Vision Model)...") + # Note: This might fail if the configured LLM doesn't support vision or if the URL is unreachable + try: + fine_results = self.parser.parse_fine(msg, info) + if not fine_results: + print( + " ⚠️ No memories generated (LLM might not support vision or image inaccessible)." + ) + else: + print(f" 📊 Generated {len(fine_results)} memory item(s):") + for item in fine_results: + print(f" - Memory: {item.memory[:100]}...") + except Exception as e: + print(f" ❌ Error during fine parsing: {e}") + + print() + + +if __name__ == "__main__": + demo = ImageParserDemo() + demo.run() diff --git a/examples/mem_reader/parser_demos/demo_multi_modal.py b/examples/mem_reader/parser_demos/demo_multi_modal.py new file mode 100644 index 000000000..3302ef073 --- /dev/null +++ b/examples/mem_reader/parser_demos/demo_multi_modal.py @@ -0,0 +1,346 @@ +"""Demo for MultiModalParser.""" + +from examples.mem_reader.builders import build_file_parser +from memos.mem_reader.read_multi_modal.multi_modal_parser import MultiModalParser + +from ._base import BaseParserDemo + + +class MultiModalParserDemo(BaseParserDemo): + def create_parser(self): + file_parser = build_file_parser() + return MultiModalParser(embedder=self.embedder, llm=self.llm, parser=file_parser) + + def run(self): + self.parser_selection() + self.parser_instances() + print("\n✅ MultiModalParser example completed!") + + def parser_selection(self): + """Test that different input types return the correct parser.""" + print("=== MultiModalParser Parser Selection Test ===\n") + + # Test cases: different input types + test_cases = [ + # String input -> StringParser + { + "name": "String input", + "message": "This is a simple string message", + "expected_parser_type": "StringParser", + }, + # RawMessageList: text type -> TextContentParser + { + "name": "Text content part (RawMessageList)", + "message": {"type": "text", "text": "This is a text content part"}, + "expected_parser_type": "TextContentParser", + }, + # RawMessageList: file type -> FileContentParser + { + "name": "File content part (RawMessageList)", + "message": { + "type": "file", + "file": { + "filename": "example.pdf", + "file_data": "File content here", + }, + }, + "expected_parser_type": "FileContentParser", + }, + # RawMessageList: image_url type -> ImageParser + { + "name": "Image content part (RawMessageList - image_url type)", + "message": { + "type": "image_url", + "image_url": { + "url": "https://example.com/image.jpg", + "detail": "auto", + }, + }, + "expected_parser_type": "ImageParser", + }, + # RawMessageList: input_audio type -> None (type_parsers uses "audio" key, not "input_audio") + { + "name": "Audio content part (RawMessageList - input_audio type)", + "message": { + "type": "input_audio", + "input_audio": { + "data": "base64_encoded_audio_data", + "format": "mp3", + }, + }, + "expected_parser_type": None, # type_parsers has "audio" key, but message has "input_audio" type + "should_return_none": True, + }, + # MessageList: system role -> SystemParser + { + "name": "System message", + "message": { + "role": "system", + "content": "You are a helpful assistant.", + }, + "expected_parser_type": "SystemParser", + }, + # MessageList: user role -> UserParser + { + "name": "User message (simple)", + "message": { + "role": "user", + "content": "Hello, how are you?", + }, + "expected_parser_type": "UserParser", + }, + # MessageList: user role with multimodal content -> UserParser + { + "name": "User message (multimodal with text and file)", + "message": { + "role": "user", + "content": [ + {"type": "text", "text": "What's in this image?"}, + {"type": "file", "file": {"filename": "image.jpg", "file_data": ""}}, + ], + }, + "expected_parser_type": "UserParser", + }, + # MessageList: user role with image_url content -> UserParser + { + "name": "User message (with image_url)", + "message": { + "role": "user", + "content": [ + {"type": "text", "text": "What's in this image?"}, + { + "type": "image_url", + "image_url": {"url": "https://example.com/image.jpg"}, + }, + ], + }, + "expected_parser_type": "UserParser", + }, + # MessageList: user role with input_audio content -> UserParser + { + "name": "User message (with input_audio)", + "message": { + "role": "user", + "content": [ + {"type": "text", "text": "Listen to this audio"}, + { + "type": "input_audio", + "input_audio": {"data": "base64_data", "format": "wav"}, + }, + ], + }, + "expected_parser_type": "UserParser", + }, + # MessageList: assistant role -> AssistantParser + { + "name": "Assistant message (simple)", + "message": { + "role": "assistant", + "content": "I'm doing well, thank you!", + }, + "expected_parser_type": "AssistantParser", + }, + # MessageList: assistant role with tool_calls -> AssistantParser + { + "name": "Assistant message (with tool_calls)", + "message": { + "role": "assistant", + "content": None, + "tool_calls": [ + { + "id": "call_123", + "type": "function", + "function": { + "name": "get_weather", + "arguments": '{"location": "Beijing"}', + }, + } + ], + }, + "expected_parser_type": "AssistantParser", + }, + # MessageList: tool role -> ToolParser + { + "name": "Tool message", + "message": { + "role": "tool", + "content": "Tool execution result", + "tool_call_id": "call_123", + }, + "expected_parser_type": "ToolParser", + }, + ] + + print("Testing parser selection for different input types:\n") + all_passed = True + + for i, test_case in enumerate(test_cases, 1): + message = test_case["message"] + expected_type = test_case.get("expected_parser_type") + test_name = test_case["name"] + should_return_none = test_case.get("should_return_none", False) + + # Get parser using internal method + selected_parser = self.parser._get_parser(message) + + # Handle cases where None is expected + if should_return_none or expected_type is None: + if selected_parser is None: + print(f"✅ Test {i}: {test_name}") + print(" Expected: None (parser not implemented yet or not found)") + print(" Got: None") + if expected_type: + print(f" Note: {expected_type} is not yet implemented") + else: + print(f"⚠️ Test {i}: {test_name}") + print(" Expected: None") + print(f" Got: {type(selected_parser).__name__}") + print(" Note: Parser found but may not be fully implemented") + print() + continue + + # Check if parser was found + if selected_parser is None: + print(f"❌ Test {i}: {test_name}") + print(f" Expected: {expected_type}") + print(" Got: None (parser not found)") + print(f" Message: {message}\n") + all_passed = False + continue + + # Get actual parser type name + actual_type = type(selected_parser).__name__ + + # Verify parser type + if actual_type == expected_type: + print(f"✅ Test {i}: {test_name}") + print(f" Expected: {expected_type}") + print(f" Got: {actual_type}") + print(f" Parser instance: {selected_parser}") + else: + print(f"❌ Test {i}: {test_name}") + print(f" Expected: {expected_type}") + print(f" Got: {actual_type}") + print(f" Message: {message}") + all_passed = False + print() + + # Test edge cases + print("\n=== Testing Edge Cases ===\n") + + edge_cases = [ + { + "name": "Unknown message type (not dict, not str)", + "message": 12345, + "should_return_none": True, + }, + { + "name": "Dict without type or role", + "message": {"content": "Some content"}, + "should_return_none": True, + }, + { + "name": "Unknown type in RawMessageList", + "message": {"type": "unknown_type", "data": "some data"}, + "should_return_none": True, + }, + { + "name": "Unknown role in MessageList", + "message": {"role": "unknown_role", "content": "some content"}, + "should_return_none": True, + }, + { + "name": "List of messages (MessageList - not handled by _get_parser)", + "message": [ + {"role": "user", "content": "Message 1"}, + {"role": "assistant", "content": "Message 2"}, + ], + "should_return_none": True, # Lists are handled in parse(), not _get_parser() + }, + { + "name": "List of RawMessageList items (not handled by _get_parser)", + "message": [ + {"type": "text", "text": "Text content 1"}, + {"type": "file", "file": {"filename": "doc.pdf", "file_data": ""}}, + ], + "should_return_none": True, # Lists are handled in parse(), not _get_parser() + }, + ] + + for i, test_case in enumerate(edge_cases, 1): + message = test_case["message"] + should_return_none = test_case["should_return_none"] + test_name = test_case["name"] + + selected_parser = self.parser._get_parser(message) + + if should_return_none: + if selected_parser is None: + print(f"✅ Edge Case {i}: {test_name}") + print(" Correctly returned None") + else: + print(f"❌ Edge Case {i}: {test_name}") + print(" Expected: None") + print(f" Got: {type(selected_parser).__name__}") + all_passed = False + else: + if selected_parser is not None: + print(f"✅ Edge Case {i}: {test_name}") + print(f" Got parser: {type(selected_parser).__name__}") + else: + print(f"❌ Edge Case {i}: {test_name}") + print(" Expected: Parser") + print(" Got: None") + all_passed = False + print() + + # Summary + print("=" * 60) + if all_passed: + print("✅ All tests passed! Parser selection is working correctly.") + else: + print("❌ Some tests failed. Please check the output above.") + print("=" * 60) + + def parser_instances(self): + """Test that parser instances are correctly initialized.""" + print("\n=== Parser Instance Verification ===\n") + + # Verify all parser instances are initialized + parsers_to_check = { + "string_parser": "StringParser", + "system_parser": "SystemParser", + "user_parser": "UserParser", + "assistant_parser": "AssistantParser", + "tool_parser": "ToolParser", + "text_content_parser": "TextContentParser", + "file_content_parser": "FileContentParser", + } + + print("Checking parser instance initialization:\n") + all_initialized = True + + for attr_name, expected_type in parsers_to_check.items(): + parser_instance = getattr(self.parser, attr_name, None) + if parser_instance is None: + print(f"❌ {attr_name}: Not initialized") + all_initialized = False + else: + actual_type = type(parser_instance).__name__ + if actual_type == expected_type: + print(f"✅ {attr_name}: {actual_type}") + else: + print(f"❌ {attr_name}: Expected {expected_type}, got {actual_type}") + all_initialized = False + + print() + if all_initialized: + print("✅ All parser instances are correctly initialized!") + else: + print("❌ Some parser instances are missing or incorrect.") + print() + + +if __name__ == "__main__": + demo = MultiModalParserDemo() + demo.run() diff --git a/examples/mem_reader/parser_demos/demo_string.py b/examples/mem_reader/parser_demos/demo_string.py new file mode 100644 index 000000000..e6bf99f83 --- /dev/null +++ b/examples/mem_reader/parser_demos/demo_string.py @@ -0,0 +1,30 @@ +"""Demo for StringParser.""" + +from examples.mem_reader.samples import STRING_MESSAGE_CASES +from memos.mem_reader.read_multi_modal.string_parser import StringParser + +from ._base import BaseParserDemo + + +class StringParserDemo(BaseParserDemo): + def create_parser(self): + return StringParser(embedder=self.embedder, llm=self.llm) + + def run(self): + print("=== StringParser Demo ===") + + info = {"user_id": "user1", "session_id": "session1"} + + for case in STRING_MESSAGE_CASES: + print(f"\n--- Case: {case.description} ---") + print("📝 Processing string messages:\n") + for i, msg in enumerate(case.scene_data, 1): + print(f"Message {i}: {msg[:50]}...") + source = self.demo_source_creation(msg, info) + self.demo_rebuild(source) + print() + + +if __name__ == "__main__": + demo = StringParserDemo() + demo.run() diff --git a/examples/mem_reader/parser_demos/demo_system.py b/examples/mem_reader/parser_demos/demo_system.py new file mode 100644 index 000000000..8655833ef --- /dev/null +++ b/examples/mem_reader/parser_demos/demo_system.py @@ -0,0 +1,41 @@ +"""Demo for SystemParser.""" + +from examples.mem_reader.samples import SYSTEM_MESSAGE_CASES +from memos.mem_reader.read_multi_modal.system_parser import SystemParser + +from ._base import BaseParserDemo + + +class SystemParserDemo(BaseParserDemo): + def create_parser(self): + return SystemParser(embedder=self.embedder, llm=self.llm) + + def run(self): + print("=== SystemParser Demo ===") + + info = {"user_id": "user1", "session_id": "session1"} + + for case in SYSTEM_MESSAGE_CASES: + print(f"\n--- Case: {case.description} ---") + for msg in case.scene_data: + # Workaround: SystemParser in src only supports str/dict content, not list. + # Since we cannot modify src, we flatten list content here. + msg_to_process = msg + if isinstance(msg.get("content"), list): + msg_to_process = msg.copy() + content_list = msg["content"] + merged_text = "".join( + part.get("text", "") + for part in content_list + if isinstance(part, dict) and part.get("type") == "text" + ) + msg_to_process["content"] = merged_text + + source = self.demo_source_creation(msg_to_process, info) + self.demo_rebuild(source) + self.demo_parse_fast(msg_to_process, info) + + +if __name__ == "__main__": + demo = SystemParserDemo() + demo.run() diff --git a/examples/mem_reader/parser_demos/demo_text_content.py b/examples/mem_reader/parser_demos/demo_text_content.py new file mode 100644 index 000000000..4c27c6422 --- /dev/null +++ b/examples/mem_reader/parser_demos/demo_text_content.py @@ -0,0 +1,40 @@ +"""Demo for TextContentParser.""" + +from examples.mem_reader.samples import TEXT_CONTENT_PARTS +from memos.mem_reader.read_multi_modal.text_content_parser import TextContentParser + +from ._base import BaseParserDemo + + +class TextContentParserDemo(BaseParserDemo): + def create_parser(self): + return TextContentParser(embedder=self.embedder, llm=self.llm) + + def run(self): + print("=== TextContentParser Demo ===") + + info = {"user_id": "user1", "session_id": "session1"} + + for i, part in enumerate(TEXT_CONTENT_PARTS, 1): + print(f"\n--- Part {i} ---") + source = self.demo_source_creation(part, info) + + # Legacy example attempts to rebuild and access dict keys directly. + # Since current source returns None, we must handle it safely in the demo. + print("\n🔄 Rebuilding from source...") + rebuilt = self.parser.rebuild_from_source(source) + if rebuilt: + print(" ✅ Rebuilt result:") + if isinstance(rebuilt, dict): + from examples.mem_reader.utils import pretty_print_dict + + pretty_print_dict(rebuilt) + else: + print(f" {rebuilt}") + else: + print(" ⚠️ Rebuilt result is None (not implemented in source)") + + +if __name__ == "__main__": + demo = TextContentParserDemo() + demo.run() diff --git a/examples/mem_reader/parser_demos/demo_tool.py b/examples/mem_reader/parser_demos/demo_tool.py new file mode 100644 index 000000000..991f21751 --- /dev/null +++ b/examples/mem_reader/parser_demos/demo_tool.py @@ -0,0 +1,28 @@ +"""Demo for ToolParser.""" + +from examples.mem_reader.samples import TOOL_MESSAGE_CASES +from memos.mem_reader.read_multi_modal.tool_parser import ToolParser + +from ._base import BaseParserDemo + + +class ToolParserDemo(BaseParserDemo): + def create_parser(self): + return ToolParser(embedder=self.embedder, llm=self.llm) + + def run(self): + print("=== ToolParser Demo ===") + + info = {"user_id": "user1", "session_id": "session1"} + + for case in TOOL_MESSAGE_CASES: + print(f"\n--- Case: {case.description} ---") + for msg in case.scene_data: + source = self.demo_source_creation(msg, info) + self.demo_rebuild(source) + self.demo_parse_fast(msg, info) + + +if __name__ == "__main__": + demo = ToolParserDemo() + demo.run() diff --git a/examples/mem_reader/parser_demos/demo_user.py b/examples/mem_reader/parser_demos/demo_user.py new file mode 100644 index 000000000..16dafcb2d --- /dev/null +++ b/examples/mem_reader/parser_demos/demo_user.py @@ -0,0 +1,43 @@ +"""Demo for UserParser.""" + +from examples.mem_reader.samples import USER_MESSAGE_CASES +from memos.mem_reader.read_multi_modal.user_parser import UserParser + +from ._base import BaseParserDemo + + +class UserParserDemo(BaseParserDemo): + def create_parser(self): + return UserParser(embedder=self.embedder, llm=self.llm) + + def run(self): + print("=== UserParser Demo ===") + + info = {"user_id": "user1", "session_id": "session1"} + + for case in USER_MESSAGE_CASES: + print(f"\n--- Case: {case.description} ---") + for msg in case.scene_data: + sources = self.demo_source_creation(msg, info) + + # Rebuild all sources to show full multimodal support + if isinstance(sources, list): + for i, src in enumerate(sources): + print(f"\n🔄 Rebuilding source part {i + 1} ({src.type})...") + rebuilt = self.parser.rebuild_from_source(src) + print(" ✅ Rebuilt result:") + if isinstance(rebuilt, dict): + from examples.mem_reader.utils import pretty_print_dict + + pretty_print_dict(rebuilt) + else: + print(f" {rebuilt}") + else: + self.demo_rebuild(sources) + + self.demo_parse_fast(msg, info) + + +if __name__ == "__main__": + demo = UserParserDemo() + demo.run() diff --git a/examples/mem_reader/runners/__init__.py b/examples/mem_reader/runners/__init__.py new file mode 100644 index 000000000..e69de29bb diff --git a/examples/mem_reader/runners/run_multimodal.py b/examples/mem_reader/runners/run_multimodal.py new file mode 100644 index 000000000..e4176e898 --- /dev/null +++ b/examples/mem_reader/runners/run_multimodal.py @@ -0,0 +1,139 @@ +"""Runner for MultiModalStructMemReader.""" + +import argparse +import json +import time +import traceback + +from examples.mem_reader.builders import build_multimodal_reader +from examples.mem_reader.samples import ( + MULTIMODAL_MESSAGE_CASES, + RAW_INPUT_CASES, + STRING_MESSAGE_CASES, +) +from examples.mem_reader.utils import print_memory_item + + +# Map example names to test cases +EXAMPLE_MAP = { + "string_message": STRING_MESSAGE_CASES, + "multimodal": MULTIMODAL_MESSAGE_CASES, + "raw_input": RAW_INPUT_CASES, +} + + +def run_multimodal_reader(): + """Run MultiModalStructMemReader with sample data.""" + parser = argparse.ArgumentParser(description="MultiModalStructMemReader Example") + parser.add_argument( + "--example", + type=str, + default="all", + choices=[*list(EXAMPLE_MAP.keys()), "all"], + help="Example to run", + ) + parser.add_argument( + "--mode", + type=str, + default="fine", + choices=["fast", "fine"], + help="Processing mode (fast/fine)", + ) + parser.add_argument( + "--format", + type=str, + default="text", + choices=["text", "json"], + help="Output format", + ) + + args = parser.parse_args() + + print("🚀 Initializing MultiModalStructMemReader...") + reader = build_multimodal_reader() + print("✅ Initialization complete.") + + # Select test cases + if args.example == "all": + test_cases = [] + for cases in EXAMPLE_MAP.values(): + test_cases.extend(cases) + else: + test_cases = EXAMPLE_MAP[args.example] + + print(f"📋 Running {len(test_cases)} test cases in '{args.mode}' mode...\n") + + results = [] + + for i, case in enumerate(test_cases): + print(f"🔹 Case {i + 1}: {case.name} - {case.description}") + + info = case.get_info() + scene_data = case.scene_data + + # Data structure adaptation logic + # Ensure scene_data is List[List[dict]] if it looks like a single conversation + # Most samples in samples.py are wrapped in [], so they are List[List[dict]]. + # Except STRING_MESSAGE_CASES which are List[str]. + if ( + isinstance(scene_data, list) + and len(scene_data) > 0 + and not isinstance(scene_data[0], list) + and not isinstance(scene_data[0], str) + ): + scene_data = [scene_data] + + try: + start_time = time.time() + + # Determine input type + input_type = "chat" + if case in EXAMPLE_MAP["string_message"]: + input_type = "string" + elif case in EXAMPLE_MAP["raw_input"]: + input_type = "raw" + + memories = reader.get_memory( + scene_data, + type=input_type, + mode=args.mode, + info=info, + ) + duration = time.time() - start_time + + result_entry = { + "case": case.name, + "description": case.description, + "duration_seconds": round(duration, 4), + "memory_count": sum(len(m) for m in memories), + "memories": [], + } + + print( + f" ✅ Processed in {duration:.4f}s. Extracted {result_entry['memory_count']} memories." + ) + + # Flatten memories for display/output + flat_memories = [item for sublist in memories for item in sublist] + + if args.format == "json": + # Convert TextualMemoryItem to dict + result_entry["memories"] = [ + m.to_dict() if hasattr(m, "to_dict") else str(m) for m in flat_memories + ] + results.append(result_entry) + else: + for item in flat_memories: + print_memory_item(item, indent=6) + print() + + except Exception as e: + print(f" ❌ Error: {e}") + traceback.print_exc() + + if args.format == "json": + print(json.dumps(results, indent=2, ensure_ascii=False)) + + +if __name__ == "__main__": + run_multimodal_reader() diff --git a/examples/mem_reader/runners/run_simple.py b/examples/mem_reader/runners/run_simple.py new file mode 100644 index 000000000..21ecd73a9 --- /dev/null +++ b/examples/mem_reader/runners/run_simple.py @@ -0,0 +1,113 @@ +"""Runner for SimpleStructMemReader.""" + +import time + +from examples.mem_reader.samples import SIMPLE_CHAT_SCENE +from examples.mem_reader.settings import get_reader_config +from examples.mem_reader.utils import print_memory_item +from memos.configs.mem_reader import SimpleStructMemReaderConfig +from memos.mem_reader.simple_struct import SimpleStructMemReader + + +def _print_memory_sets(title: str, memories): + """memories: list[list[TextualMemoryItem]]""" + total = sum(len(mem_list) for mem_list in memories) + print(f"\n{title}") + print(f"📊 Total memory items: {total}") + print(f"✅ Extracted {len(memories)} memory sets.") + for i, memory_list in enumerate(memories): + print(f"\n--- Window/Conversation {i + 1} Memories ({len(memory_list)} items) ---") + for item in memory_list: + print_memory_item(item, indent=2) + + +def run_simple_reader(): + """Run SimpleStructMemReader with sample data.""" + print("🚀 Initializing SimpleStructMemReader from JSON config...") + + # Use settings config instead of hardcoded JSON + reader_config = SimpleStructMemReaderConfig(**get_reader_config()) + reader = SimpleStructMemReader(reader_config) + print("✅ Initialization complete.") + + info = {"user_id": "simple_user", "session_id": "simple_session"} + + print("\n📝 Processing Simple Chat Scene...") + # SIMPLE_CHAT_SCENE: list[list[dict]] (multiple conversations) + + try: + # 1) FINE + print("\n🔄 Testing FINE mode (with LLM)...") + t0 = time.time() + fine_memory = reader.get_memory( + SIMPLE_CHAT_SCENE, + type="chat", + info=info, + mode="fine", + ) + fine_time = time.time() - t0 + print(f"⏱️ Fine mode time: {fine_time:.2f}s") + _print_memory_sets("=== FINE Mode Results ===", fine_memory) + + # 2) FAST + print("\n⚡ Testing FAST mode (no LLM)...") + t0 = time.time() + fast_memory = reader.get_memory( + SIMPLE_CHAT_SCENE, + type="chat", + info=info, + mode="fast", + ) + fast_time = time.time() - t0 + print(f"⏱️ Fast mode time: {fast_time:.2f}s") + _print_memory_sets("=== FAST Mode Results ===", fast_memory) + + # 3) Transfer: FAST -> FINE + # fine_transfer_simple_mem expects a flat list[TextualMemoryItem] + print("\n🔁 Transfer FAST memories -> FINE...") + flat_fast_items = [item for mem_list in fast_memory for item in mem_list] + + t0 = time.time() + transferred = reader.fine_transfer_simple_mem(flat_fast_items, type="chat") + transfer_time = time.time() - t0 + + print(f"⏱️ Transfer time: {transfer_time:.2f}s") + _print_memory_sets("=== TRANSFER Results (FAST -> FINE) ===", transferred) + + # 4) Documents (Fine only) + print("\n📄 Processing Documents (Fine Mode Only)...") + doc_paths = [ + "text1.txt", + "text2.txt", + ] + + try: + t0 = time.time() + doc_memory = reader.get_memory( + doc_paths, + type="doc", + info={"user_id": "doc_user", "session_id": "doc_session"}, + mode="fine", + ) + doc_time = time.time() - t0 + print(f"⏱️ Doc fine mode time: {doc_time:.2f}s") + _print_memory_sets("=== DOC Mode Results (FINE) ===", doc_memory) + except Exception as e: + print(f"⚠️ Document processing failed: {e}") + print(" (This is expected if document files don't exist)") + + # 5) Summary (no speedup) + print("\n📈 Summary") + print(f" Fine: {fine_time:.2f}s") + print(f" Fast: {fast_time:.2f}s") + print(f" Transfer: {transfer_time:.2f}s") + + except Exception as e: + print(f"❌ Error during processing: {e}") + import traceback + + traceback.print_exc() + + +if __name__ == "__main__": + run_simple_reader() diff --git a/examples/mem_reader/samples.py b/examples/mem_reader/samples.py new file mode 100644 index 000000000..c29177c58 --- /dev/null +++ b/examples/mem_reader/samples.py @@ -0,0 +1,773 @@ +"""Sample data for MemReader examples. + +This module contains test cases and sample data for various MemReader scenarios, +including simple chat, multimodal messages, file content, and tool usage. +""" + +from dataclasses import dataclass, field +from typing import Any + + +@dataclass +class TestCase: + """Base class for test cases.""" + + name: str + description: str + scene_data: Any + expected_count: dict[str, int] = field(default_factory=dict) + + def get_info(self) -> dict[str, Any]: + """Get info dict for this test case.""" + return { + "user_id": "test_user", + "session_id": f"session_{self.name}", + "test_case": self.name, + } + + +# ============================================================================ +# 1. Simple Chat Samples (for SimpleStructMemReader) +# ============================================================================ + +SIMPLE_CHAT_SCENE = [ + [ + {"role": "user", "chat_time": "3 May 2025", "content": "I'm feeling a bit down today."}, + { + "role": "assistant", + "chat_time": "3 May 2025", + "content": "I'm sorry to hear that. Do you want to talk about what's been going on?", + }, + { + "role": "user", + "chat_time": "3 May 2025", + "content": "It's just been a tough couple of days, you know? Everything feels a bit overwhelming, and I just can't seem to shake it off.", + }, + { + "role": "assistant", + "chat_time": "3 May 2025", + "content": "It sounds like you're going through a lot right now. Sometimes it helps to talk things out. Is there something specific that's been weighing on you, or is it more of a general feeling?", + }, + { + "role": "user", + "chat_time": "3 May 2025", + "content": "It's a mix, I guess. Work's been really stressful, and on top of that, I've been feeling kinda disconnected from the people around me.", + }, + { + "role": "assistant", + "chat_time": "3 May 2025", + "content": "That can be really draining, especially when you're feeling isolated on top of the stress. Do you think there's something from your past that's contributing to how you're feeling now? Sometimes our emotions are tied to older experiences.", + }, + { + "role": "user", + "chat_time": "3 May 2025", + "content": "Yeah, now that you mention it… I've been thinking a lot about my childhood lately. I didn't have the easiest upbringing. My parents were always busy, and I often felt like I had to figure things out on my own. I remember one winter, I got really sick with a high fever. My parents were both working, and I didn't want to bother them. I just lay in bed, crying by myself, not knowing what to do.", + }, + { + "role": "assistant", + "chat_time": "3 May 2025", + "content": "That must have been very lonely. Growing up without that support can really make you feel like you have to handle everything on your own. How do you think that experience has affected you now?", + }, + { + "role": "user", + "chat_time": "3 May 2025", + "content": "Yeah, it was really lonely. I remember spending a lot of time doing things on my own, like drawing, because it was my only way to escape. I used to draw animals, and I created my own little world on paper. It was peaceful in a way, just me and my imagination, with no outside distractions.", + }, + { + "role": "assistant", + "chat_time": "3 May 2025", + "content": "Drawing sounds like a beautiful way to express yourself and find some peace in that solitude. It must have been a very quiet and private world for you. Do you think that sense of loneliness from back then has shaped how you connect with others today?", + }, + { + "role": "user", + "chat_time": "3 May 2025", + "content": "I think it has. I've always found it hard to let others in. I feel like I have to handle everything on my own. If I rely on someone else, I feel vulnerable, and I start to feel that loneliness again.", + }, + { + "role": "assistant", + "chat_time": "3 May 2025", + "content": "It sounds like this was a protective mechanism that helped you cope with your feelings of loneliness back then. But as you've grown, it seems like this pattern might not be serving you as well anymore. Have you thought about trying to break this cycle, or finding new ways to connect with others?", + }, + ] +] + +# ============================================================================ +# 2. String Message Cases +# ============================================================================ + +STRING_MESSAGE_CASES = [ + TestCase( + name="string_simple", + description="Simple string message", + scene_data=[ + "This is a simple text message that needs to be parsed.", + "Another string message for processing.", + "StringParser handles plain text strings and converts them to SourceMessage objects.", + ], + expected_count={"fast": 1, "fine": 1}, + ), + TestCase( + name="string_multiple", + description="Multiple string messages", + scene_data=[ + "这是第一条消息。", + "这是第二条消息。", + "这是第三条消息。", + ], + ), +] + +# ============================================================================ +# 3. Chat Message Cases (Standard & Multimodal) +# ============================================================================ + +CHAT_MESSAGE_CASES = [ + TestCase( + name="chat_simple", + description="Simple chat conversation", + scene_data=[ + [ + { + "role": "user", + "content": "Hello, how are you? I'm planning to learn Python next week.", + "chat_time": "2025-01-01T10:00:00Z", + "message_id": "chat_simple_u1", + }, + { + "role": "assistant", + "content": "I'm doing well, thank you!", + "chat_time": "2025-01-01T10:00:01Z", + "message_id": "chat_simple_a1", + }, + ] + ], + ), + TestCase( + name="chat_with_system", + description="Chat with system message", + scene_data=[ + [ + { + "role": "system", + "content": "You are a helpful assistant.", + "chat_time": "2025-01-01T10:00:00Z", + "message_id": "chat_sys_s1", + }, + { + "role": "user", + "content": "What's the weather?", + "chat_time": "2025-01-01T10:00:01Z", + "message_id": "chat_sys_u1", + }, + { + "role": "assistant", + "content": "I don't have access to weather data.", + "chat_time": "2025-01-01T10:00:02Z", + "message_id": "chat_sys_a1", + }, + ] + ], + ), + TestCase( + name="chat_multimodal_complex", + description="Complex multimodal chat with text, file, and image", + scene_data=[ + [ + { + "role": "user", + "content": [ + {"type": "text", "text": "我是测试base64"}, + { + "type": "file", + "file": { + "file_data": "Hello World", + "filename": "example.txt", + "file_id": "file_123", + }, + }, + { + "type": "image_url", + "image_url": { + "url": "https://statics.memtensor.com.cn/memos/memos-banner.gif", + "detail": "auto", + }, + }, + ], + "chat_time": "2025-01-01T10:00:03Z", + "message_id": "chat_mm_u1", + } + ] + ], + ), +] + + +# ============================================================================ +# 4. Tool Message Cases +# ============================================================================ + +TOOL_MESSAGE_CASES = [ + TestCase( + name="tool_weather", + description="Weather tool result", + scene_data=[ + { + "role": "user", + "content": "I'm planning a hiking trip to New York this weekend, can you check the weather?", + "chat_time": "2025-01-15T10:00:00", + "message_id": "msg_000", + }, + { + "role": "tool", + "content": '{"result": "Weather in New York: 72°F, sunny"}', + "tool_call_id": "call_abc123", + "chat_time": "2025-01-15T10:00:30", + "message_id": "msg_001", + }, + ], + ), + TestCase( + name="tool_data", + description="Data API result", + scene_data=[ + { + "role": "user", + "content": "Please retrieve my saved reading list items.", + "chat_time": "2025-01-15T10:05:00", + "message_id": "msg_000_2", + }, + { + "role": "tool", + "content": '{"status": "success", "data": {"items": ["The Great Gatsby", "1984", "Python Crash Course"]}}', + "tool_call_id": "call_def456", + "chat_time": "2025-01-15T10:05:30", + "message_id": "msg_002", + }, + ], + ), + TestCase( + name="tool_db", + description="Database query result", + scene_data=[ + { + "role": "user", + "content": "Did I complete the registration for the upcoming workshop?", + "chat_time": "2025-01-15T10:10:00", + "message_id": "msg_000_3", + }, + { + "role": "tool", + "content": "Database query executed successfully. Found registration record for user_id=123: status=confirmed.", + "tool_call_id": "call_ghi789", + "chat_time": "2025-01-15T10:10:30", + "message_id": "msg_003", + }, + ], + ), +] + +# ============================================================================ +# 5. File Content Samples (for FileContentParser Demo) +# ============================================================================ + +FILE_CONTENT_PARTS = [ + { + "type": "file", + "file": { + "filename": "document.pdf", + "file_id": "file_123", + "file_data": "This is the content extracted from the PDF file...", + }, + }, + { + "type": "file", + "file": { + "filename": "report.docx", + "file_id": "file_456", + "file_data": "Report content: Analysis of Q4 performance...", + }, + }, + { + "type": "file", + "file": { + "filename": "data.csv", + "file_id": "file_789", + "path": "/path/to/data.csv", + }, + }, +] + +FILE_CONTENT_REAL_FILE_PART = { + "type": "file", + "file": { + "filename": "example.txt", + "path": "examples/mem_reader/text1.txt", + }, +} + +# ============================================================================ +# 6. Text Content Samples (for TextContentParser Demo) +# ============================================================================ + +TEXT_CONTENT_PARTS = [ + {"type": "text", "text": "This is a simple text content part."}, + {"type": "text", "text": "TextContentParser handles text parts in multimodal messages."}, +] + +# ============================================================================ +# 7. System Message Samples (for SystemParser Demo) +# ============================================================================ + +SYSTEM_MESSAGE_CASES = [ + TestCase( + name="system_simple", + description="Simple text system message", + scene_data=[ + { + "role": "system", + "content": "You are a helpful assistant that provides clear and concise answers.", + "chat_time": "2025-01-15T10:00:00", + "message_id": "msg_001", + } + ], + ), + TestCase( + name="system_multimodal", + description="Multimodal system message (multiple text parts)", + scene_data=[ + { + "role": "system", + "content": [ + {"type": "text", "text": "You are a helpful assistant."}, + {"type": "text", "text": "Always provide clear and concise answers."}, + {"type": "text", "text": "If you don't know something, say so."}, + ], + "chat_time": "2025-01-15T10:05:00", + "message_id": "msg_002", + } + ], + ), + TestCase( + name="system_structured", + description="Structured system instructions (multiple text parts)", + scene_data=[ + { + "role": "system", + "content": [ + { + "type": "text", + "text": "You are a coding assistant specialized in Python programming.", + }, + {"type": "text", "text": "Always write clean, well-documented code."}, + {"type": "text", "text": "Explain your reasoning when providing solutions."}, + ], + "chat_time": "2025-01-15T10:10:00", + "message_id": "msg_003", + } + ], + ), +] + +# ============================================================================ +# 8. User Message Samples (for UserParser Demo) +# ============================================================================ + +USER_MESSAGE_CASES = [ + TestCase( + name="user_simple", + description="Simple text user message", + scene_data=[ + { + "role": "user", + "content": "I'm feeling a bit down today. Can you help me?", + "chat_time": "2025-01-15T10:00:00", + "message_id": "msg_001", + } + ], + ), + TestCase( + name="user_multimodal", + description="Multimodal user message (text + file)", + scene_data=[ + { + "role": "user", + "content": [ + {"type": "text", "text": "Please analyze this document:"}, + { + "type": "file", + "file": { + "filename": "report.pdf", + "file_id": "file_123", + "file_data": "This is the content of the PDF file...", + }, + }, + ], + "chat_time": "2025-01-15T10:05:00", + "message_id": "msg_002", + } + ], + ), + TestCase( + name="user_image", + description="User message with image", + scene_data=[ + { + "role": "user", + "content": [ + {"type": "text", "text": "What's in this image?"}, + {"type": "image_url", "image_url": {"url": "https://example.com/image.jpg"}}, + ], + "chat_time": "2025-01-15T10:10:00", + "message_id": "msg_003", + } + ], + ), +] + +# ============================================================================ +# 9. Assistant Message Samples (for AssistantParser Demo) +# ============================================================================ + +ASSISTANT_MESSAGE_CASES = [ + TestCase( + name="assistant_simple", + description="Simple support message", + scene_data=[ + { + "role": "assistant", + "content": "I'm sorry to hear that you're feeling down. Would you like to talk about what's been going on?", + "chat_time": "2025-01-15T10:00:30", + "message_id": "msg_001", + } + ], + ), + TestCase( + name="assistant_analysis", + description="Document analysis response", + scene_data=[ + { + "role": "assistant", + "content": "Based on the document you provided, I can see several key points: 1) The project timeline, 2) Budget considerations, and 3) Resource allocation.", + "chat_time": "2025-01-15T10:05:30", + "message_id": "msg_002", + } + ], + ), + TestCase( + name="assistant_code", + description="Code solution", + scene_data=[ + { + "role": "assistant", + "content": "Here's a Python solution for your problem:\n```python\ndef solve_problem():\n return 'solution'\n```", + "chat_time": "2025-01-15T10:10:30", + "message_id": "msg_003", + } + ], + ), +] + +# ============================================================================ +# 10. Image Samples (for ImageParser Demo) +# ============================================================================ + +IMAGE_MESSAGE_CASES = [ + { + "type": "image_url", + "image_url": { + "url": "https://statics.memtensor.com.cn/memos/memos-banner.gif", + "detail": "auto", + }, + "_note": "Real Image (MemOS Banner)", + }, + { + "type": "image", + "image_url": "https://example.com/image2.png", + "_note": "Dummy Image (Negative Test)", + }, +] + +# ============================================================================ +# 11. Multimodal Message Cases (from Legacy) +# ============================================================================ + +MULTIMODAL_MESSAGE_CASES = [ + TestCase( + name="multimodal_text_image", + description="User message with text and image", + scene_data=[ + [ + { + "role": "user", + "content": [ + {"type": "text", "text": "帮我看看这张图片大概是什么内容？"}, + { + "type": "image_url", + "image_url": { + "url": "https://example.com/mountain_lake.jpg", + "detail": "high", + }, + }, + ], + "chat_time": "2025-11-24T10:20:00Z", + "message_id": "mm-img-1", + } + ] + ], + ), + TestCase( + name="multimodal_text_file", + description="User message with text and file", + scene_data=[ + [ + { + "role": "user", + "content": [ + {"type": "text", "text": "请阅读这个PDF，总结里面的要点。"}, + {"type": "file", "file": {"file_id": "file_123", "filename": "report.pdf"}}, + ], + "chat_time": "2025-11-24T10:21:00Z", + "message_id": "mm-file-1", + } + ] + ], + ), + TestCase( + name="oss_text_file", + description="User message with text and file", + scene_data=[ + [ + { + "role": "user", + "content": [ + {"type": "text", "text": "请阅读这个PDF，总结里面的要点。"}, + { + "type": "file", + "file": { + "file_id": "file_123", + "filename": "report.pdf", + "file_data": "@http://139.196.232.20:9090/graph-test/algorithm/2025_11_13/1763043889_1763043782_PM1%E8%BD%A6%E9%97%B4PMT%E9%9D%B4%E5%8E%8B%E8%BE%B9%E5%8E%8B%E5%8E%8B%E5%8A%9B%E6%97%A0%E6%B3%95%E5%BB%BA%E7%AB%8B%E6%95%85%E9%9A%9C%E6%8A%A5%E5%91%8A20240720.md", + }, + }, + ], + "chat_time": "2025-11-24T10:21:00Z", + "message_id": "mm-file-1", + } + ] + ], + ), + TestCase( + name="pure_data_file", + description="User message with text and file", + scene_data=[ + [ + { + "role": "user", + "content": [ + {"type": "text", "text": "请阅读这个PDF，总结里面的要点。"}, + { + "type": "file", + "file": { + "file_id": "file_123", + "filename": "report.pdf", + "file_data": "明文记忆是系统与用户对话、操作等交互中动态习得，以及外部提供的、可显式管理的结构化知识形态，通常以文档、提示模板、图结构或用户规则等形式存在。它具备编辑性、可共享性与治理友好性，适合存储需要频繁修改、可审计或多方协同使用的信息。 在 MemOS 中，明文记忆可用于动态生成推理上下文、个性化偏好注入、多代理协作共享等场景，成为连接人类输入与模型认知的关键桥梁。激活记忆是指模型在推理过程中产生的瞬时性认知状态，包括 KV cache、隐藏层激活、注意力权重等中间张量结构。它通常用于维持上下文连续性、对话一致性与行为风格控制。 MemOS 将激活记忆抽象为可调度资源，支持按需唤醒、延迟卸载与结构变换。例如，某些上下文状态可以被压缩为“半结构化记忆片段”用于未来复用，也可以在任务级别转化为参数化模块，支持短期记忆的长期化演进。这一机制为模型行为一致性、风格保持与状态持续性提供了基础。", + }, + }, + ], + "chat_time": "2025-11-24T10:21:00Z", + "message_id": "mm-file-1", + } + ] + ], + ), + TestCase( + name="local_data_file", + description="User message with text and file", + scene_data=[ + [ + { + "role": "user", + "content": [ + {"type": "text", "text": "请阅读这个PDF，总结里面的要点。"}, + { + "type": "file", + "file": { + "file_id": "file_123", + "filename": "report.pdf", + "file_data": "./my_local_file/report.pdf", + }, + }, + ], + "chat_time": "2025-11-24T10:21:00Z", + "message_id": "mm-file-1", + } + ] + ], + ), + TestCase( + name="internet_file", + description="User message with text and file", + scene_data=[ + [ + { + "role": "user", + "content": [ + {"type": "text", "text": "请阅读这个PDF，总结里面的要点。"}, + { + "type": "file", + "file": { + "file_id": "file_123", + "filename": "report.pdf", + "file_data": "https://upload.wikimedia.org/wikipedia/commons/c/cb/NLC416-16jh004830-88775_%E7%B4%85%E6%A8%93%E5%A4%A2.pdf", + }, + }, + ], + "chat_time": "2025-11-24T10:21:00Z", + "message_id": "mm-file-1", + } + ] + ], + ), + TestCase( + name="multimodal_mixed", + description="Mixed multimodal message (text + file + image)", + scene_data=[ + [ + { + "role": "user", + "content": [ + {"type": "text", "text": "请同时分析这个报告和图表。"}, + { + "type": "file", + "file": {"file_id": "file_789", "filename": "analysis_report.pdf"}, + }, + { + "type": "image_url", + "image_url": {"url": "https://example.com/chart.png", "detail": "auto"}, + }, + ], + "chat_time": "2025-11-24T10:23:00Z", + "message_id": "mixed-1", + } + ] + ], + ), + TestCase( + name="multimodal_audio", + description="Audio-only message", + scene_data=[ + [ + { + "role": "user", + "content": [ + { + "type": "input_audio", + "input_audio": {"data": "base64_encoded_audio_here", "format": "mp3"}, + } + ], + "chat_time": "2025-11-24T10:22:00Z", + "message_id": "audio-1", + } + ] + ], + ), +] + +# ============================================================================ +# 12. Raw Input Cases (from Legacy) +# ============================================================================ + +RAW_INPUT_CASES = [ + TestCase( + name="raw_text_items", + description="Pure text input items without dialog context", + scene_data=[ + [ + {"type": "text", "text": "这是一段独立的文本输入，没有明确的对话上下文。"}, + {"type": "text", "text": "它依然会被抽取和写入明文记忆。"}, + ] + ], + ), + TestCase( + name="raw_file_item", + description="Pure file input by file_id", + scene_data=[ + [{"type": "file", "file": {"file_id": "file_uploaded_123", "filename": "document.pdf"}}] + ], + ), + TestCase( + name="file_only_file_id", + description="File with only file_id parameter", + scene_data=[[{"type": "file", "file": {"file_id": "file_only_id_123"}}]], + ), + TestCase( + name="file_only_filename", + description="File with only filename parameter", + scene_data=[[{"type": "file", "file": {"filename": "document_only.pdf"}}]], + ), + TestCase( + name="file_only_file_data_base64", + description="File with only file_data (base64 encoded)", + scene_data=[ + [ + { + "type": "file", + "file": { + "file_data": "data:application/pdf;base64,JVBERi0xLjQKJdPr6eEKMSAwIG9iago8PAovVHlwZSAvQ2F0YWxvZwovUGFnZXMgMiAwIFIKPj4KZW5kb2JqCjIgMCBvYmoKPDwKL1R5cGUgL1BhZ2VzCi9LaWRzIFszIDAgUl0KL0NvdW50IDEKPD4KZW5kb2JqCjMgMCBvYmoKPDwKL1R5cGUgL1BhZ2UKL1BhcmVudCAyIDAgUgovTWVkaWFCb3ggWzAgMCA2MTIgNzkyXQovUmVzb3VyY2VzIDw8Ci9Gb250IDw8Ci9GMSA0IDAgUgo+Pgo+PgovQ29udGVudHMgNSAwIFIKPj4KZW5kb2JqCjQgMCBvYmoKPDwKL1R5cGUgL0ZvbnQKL1N1YnR5cGUgL1R5cGUxCi9CYXNlRm9udCAvSGVsdmV0aWNhCj4+CmVuZG9iag==" + }, + } + ] + ], + ), + TestCase( + name="file_only_file_data_url", + description="File with only file_data (URL)", + scene_data=[ + [{"type": "file", "file": {"file_data": "https://example.com/documents/report.pdf"}}] + ], + ), + TestCase( + name="file_only_file_data_text", + description="File with only file_data (plain text content)", + scene_data=[ + [ + { + "type": "file", + "file": { + "file_data": "This is a plain text file content. It contains multiple lines.\nLine 2 of the file.\nLine 3 of the file." + }, + } + ] + ], + ), + TestCase( + name="file_file_data_and_file_id", + description="File with file_data and file_id", + scene_data=[ + [ + { + "type": "file", + "file": { + "file_data": "https://example.com/documents/data.pdf", + "file_id": "file_with_data_123", + }, + } + ] + ], + ), + TestCase( + name="file_file_data_and_filename", + description="File with file_data and filename", + scene_data=[ + [ + { + "type": "file", + "file": { + "file_data": "This is file content with filename.", + "filename": "content_file.txt", + }, + } + ] + ], + ), +] diff --git a/examples/mem_reader/settings.py b/examples/mem_reader/settings.py new file mode 100644 index 000000000..0623a91de --- /dev/null +++ b/examples/mem_reader/settings.py @@ -0,0 +1,107 @@ +"""Configuration settings for MemReader examples. + +This module handles environment variables and default configurations for +LLMs, Embedders, and Chunkers used in the examples. +""" + +import os + +from typing import Any + +from dotenv import load_dotenv + + +# Load environment variables from .env file +load_dotenv() + + +def get_llm_config() -> dict[str, Any]: + """Get LLM configuration from environment variables.""" + openai_api_key = os.getenv("OPENAI_API_KEY") + openai_base_url = os.getenv("OPENAI_API_BASE", "https://api.openai.com/v1") + ollama_api_base = os.getenv("OLLAMA_API_BASE", "http://localhost:11434") + + # Use MEMRADER_ variables from .env as primary source + reader_model = os.getenv("MEMRADER_MODEL", os.getenv("MOS_CHAT_MODEL", "gpt-4o-mini")) + reader_api_key = os.getenv("MEMRADER_API_KEY", openai_api_key) + reader_api_base = os.getenv("MEMRADER_API_BASE", openai_base_url) + + # Check for specific MemReader backend override, otherwise assume openai if keys present + llm_backend = os.getenv("MEMRADER_LLM_BACKEND", "openai") + + if llm_backend == "ollama": + return { + "backend": "ollama", + "config": { + "model_name_or_path": reader_model, + "api_base": ollama_api_base, + "temperature": float(os.getenv("MEMRADER_TEMPERATURE", "0.0")), + "remove_think_prefix": os.getenv("MEMRADER_REMOVE_THINK_PREFIX", "true").lower() + == "true", + "max_tokens": int(os.getenv("MEMRADER_MAX_TOKENS", "8192")), + }, + } + else: # openai + return { + "backend": "openai", + "config": { + "model_name_or_path": reader_model, + "api_key": reader_api_key or "EMPTY", + "api_base": reader_api_base, + "temperature": float(os.getenv("MEMRADER_TEMPERATURE", "0.5")), + "remove_think_prefix": os.getenv("MEMRADER_REMOVE_THINK_PREFIX", "true").lower() + == "true", + "max_tokens": int(os.getenv("MEMRADER_MAX_TOKENS", "8192")), + }, + } + + +def get_embedder_config() -> dict[str, Any]: + """Get Embedder configuration from environment variables.""" + openai_api_key = os.getenv("OPENAI_API_KEY") + openai_base_url = os.getenv("OPENAI_API_BASE", "https://api.openai.com/v1") + ollama_api_base = os.getenv("OLLAMA_API_BASE", "http://localhost:11434") + + # .env uses MOS_EMBEDDER_BACKEND + embedder_backend = os.getenv("MOS_EMBEDDER_BACKEND", "ollama") + + if embedder_backend == "universal_api": + return { + "backend": "universal_api", + "config": { + "provider": os.getenv("MOS_EMBEDDER_PROVIDER", "openai"), + "api_key": os.getenv("MOS_EMBEDDER_API_KEY", openai_api_key or "sk-xxxx"), + "model_name_or_path": os.getenv("MOS_EMBEDDER_MODEL", "text-embedding-3-large"), + "base_url": os.getenv("MOS_EMBEDDER_API_BASE", openai_base_url), + }, + } + else: # ollama + return { + "backend": "ollama", + "config": { + "model_name_or_path": os.getenv("MOS_EMBEDDER_MODEL", "nomic-embed-text:latest"), + "api_base": ollama_api_base, + }, + } + + +def get_chunker_config() -> dict[str, Any]: + """Get Chunker configuration from environment variables.""" + return { + "backend": "sentence", + "config": { + "tokenizer_or_token_counter": "gpt2", + "chunk_size": 512, + "chunk_overlap": 128, + "min_sentences_per_chunk": 1, + }, + } + + +def get_reader_config() -> dict[str, Any]: + """Get full reader configuration.""" + return { + "llm": get_llm_config(), + "embedder": get_embedder_config(), + "chunker": get_chunker_config(), + } diff --git a/examples/mem_reader/simple_struct_reader.py b/examples/mem_reader/simple_struct_reader.py deleted file mode 100644 index 72dc5fd05..000000000 --- a/examples/mem_reader/simple_struct_reader.py +++ /dev/null @@ -1,831 +0,0 @@ -import argparse -import json -import os -import time - -from typing import Any - -from dotenv import load_dotenv - -from memos.configs.mem_reader import SimpleStructMemReaderConfig -from memos.mem_reader.simple_struct import SimpleStructMemReader -from memos.memories.textual.item import ( - SourceMessage, - TextualMemoryItem, - TreeNodeTextualMemoryMetadata, -) - - -# Load environment variables from .env file -load_dotenv() - - -def print_textual_memory_item( - item: TextualMemoryItem, max_memory_length: int = 200, indent: int = 0 -): - """ - Print a TextualMemoryItem in a structured format. - - Args: - item: The TextualMemoryItem to print - max_memory_length: Maximum length of memory content to display - indent: Number of spaces for indentation - """ - indent_str = " " * indent - print(f"{indent_str}{'=' * 80}") - print(f"{indent_str}TextualMemoryItem") - print(f"{indent_str}{'=' * 80}") - print(f"{indent_str}ID: {item.id}") - print( - f"{indent_str}Memory: {item.memory[:max_memory_length]}{'...' if len(item.memory) > max_memory_length else ''}" - ) - print(f"{indent_str}Memory Length: {len(item.memory)} characters") - - # Print metadata - if hasattr(item.metadata, "user_id"): - print(f"{indent_str}User ID: {item.metadata.user_id}") - if hasattr(item.metadata, "session_id"): - print(f"{indent_str}Session ID: {item.metadata.session_id}") - if hasattr(item.metadata, "memory_type"): - print(f"{indent_str}Memory Type: {item.metadata.memory_type}") - if hasattr(item.metadata, "type"): - print(f"{indent_str}Type: {item.metadata.type}") - if hasattr(item.metadata, "key") and item.metadata.key: - print(f"{indent_str}Key: {item.metadata.key}") - if hasattr(item.metadata, "tags") and item.metadata.tags: - print(f"{indent_str}Tags: {', '.join(item.metadata.tags)}") - if hasattr(item.metadata, "confidence"): - print(f"{indent_str}Confidence: {item.metadata.confidence}") - if hasattr(item.metadata, "status"): - print(f"{indent_str}Status: {item.metadata.status}") - if hasattr(item.metadata, "background") and item.metadata.background: - bg_preview = ( - item.metadata.background[:100] + "..." - if len(item.metadata.background) > 100 - else item.metadata.background - ) - print(f"{indent_str}Background: {bg_preview}") - if hasattr(item.metadata, "sources") and item.metadata.sources: - print(f"{indent_str}Sources ({len(item.metadata.sources)}):") - for i, source in enumerate(item.metadata.sources): - source_info = [] - if hasattr(source, "type"): - source_info.append(f"type={source.type}") - if hasattr(source, "role"): - source_info.append(f"role={source.role}") - if hasattr(source, "doc_path"): - source_info.append(f"doc_path={source.doc_path}") - if hasattr(source, "chat_time"): - source_info.append(f"chat_time={source.chat_time}") - if hasattr(source, "index") and source.index is not None: - source_info.append(f"index={source.index}") - print(f"{indent_str} [{i + 1}] {', '.join(source_info)}") - if hasattr(item.metadata, "created_at"): - print(f"{indent_str}Created At: {item.metadata.created_at}") - if hasattr(item.metadata, "updated_at"): - print(f"{indent_str}Updated At: {item.metadata.updated_at}") - if hasattr(item.metadata, "embedding") and item.metadata.embedding: - print(f"{indent_str}Embedding: [vector of {len(item.metadata.embedding)} dimensions]") - print(f"{indent_str}{'=' * 80}\n") - - -def print_textual_memory_item_json(item: TextualMemoryItem, indent: int = 2): - """ - Print a TextualMemoryItem as formatted JSON. - - Args: - item: The TextualMemoryItem to print - indent: JSON indentation level - """ - # Convert to dict and exclude embedding for readability - data = item.to_dict() - if "metadata" in data and "embedding" in data["metadata"]: - embedding = data["metadata"]["embedding"] - if embedding: - data["metadata"]["embedding"] = f"[vector of {len(embedding)} dimensions]" - - print(json.dumps(data, indent=indent, ensure_ascii=False)) - - -def get_reader_config() -> dict[str, Any]: - """ - Get reader configuration from environment variables. - - Returns a dictionary that can be used to create SimpleStructMemReaderConfig. - Similar to APIConfig.get_reader_config() in server_router_api.py. - - Returns: - Configuration dictionary for SimpleStructMemReaderConfig - """ - openai_api_key = os.getenv("OPENAI_API_KEY") - openai_base_url = os.getenv("OPENAI_API_BASE", "https://api.openai.com/v1") - ollama_api_base = os.getenv("OLLAMA_API_BASE", "http://localhost:11434") - - # Get LLM backend and config - llm_backend = os.getenv("MEM_READER_LLM_BACKEND", "openai") - if llm_backend == "ollama": - llm_config = { - "backend": "ollama", - "config": { - "model_name_or_path": os.getenv("MEM_READER_LLM_MODEL", "qwen3:0.6b"), - "api_base": ollama_api_base, - "temperature": float(os.getenv("MEM_READER_LLM_TEMPERATURE", "0.0")), - "remove_think_prefix": os.getenv( - "MEM_READER_LLM_REMOVE_THINK_PREFIX", "true" - ).lower() - == "true", - "max_tokens": int(os.getenv("MEM_READER_LLM_MAX_TOKENS", "8192")), - }, - } - else: # openai - llm_config = { - "backend": "openai", - "config": { - "model_name_or_path": os.getenv("MEM_READER_LLM_MODEL", "gpt-4o-mini"), - "api_key": openai_api_key or os.getenv("MEMRADER_API_KEY", "EMPTY"), - "api_base": openai_base_url, - "temperature": float(os.getenv("MEM_READER_LLM_TEMPERATURE", "0.5")), - "remove_think_prefix": os.getenv( - "MEM_READER_LLM_REMOVE_THINK_PREFIX", "true" - ).lower() - == "true", - "max_tokens": int(os.getenv("MEM_READER_LLM_MAX_TOKENS", "8192")), - }, - } - - # Get embedder backend and config - embedder_backend = os.getenv( - "MEM_READER_EMBEDDER_BACKEND", os.getenv("MOS_EMBEDDER_BACKEND", "ollama") - ) - if embedder_backend == "universal_api": - embedder_config = { - "backend": "universal_api", - "config": { - "provider": os.getenv( - "MEM_READER_EMBEDDER_PROVIDER", os.getenv("MOS_EMBEDDER_PROVIDER", "openai") - ), - "api_key": os.getenv( - "MEM_READER_EMBEDDER_API_KEY", - os.getenv("MOS_EMBEDDER_API_KEY", openai_api_key or "sk-xxxx"), - ), - "model_name_or_path": os.getenv( - "MEM_READER_EMBEDDER_MODEL", - os.getenv("MOS_EMBEDDER_MODEL", "text-embedding-3-large"), - ), - "base_url": os.getenv( - "MEM_READER_EMBEDDER_API_BASE", - os.getenv("MOS_EMBEDDER_API_BASE", openai_base_url), - ), - }, - } - else: # ollama - embedder_config = { - "backend": "ollama", - "config": { - "model_name_or_path": os.getenv( - "MEM_READER_EMBEDDER_MODEL", - os.getenv("MOS_EMBEDDER_MODEL", "nomic-embed-text:latest"), - ), - "api_base": ollama_api_base, - }, - } - - return { - "llm": llm_config, - "embedder": embedder_config, - "chunker": { - "backend": "sentence", - "config": { - "tokenizer_or_token_counter": "gpt2", - "chunk_size": 512, - "chunk_overlap": 128, - "min_sentences_per_chunk": 1, - }, - }, - } - - -def main(): - # Parse command line arguments - parser = argparse.ArgumentParser(description="Test Mem-Reader with structured output") - parser.add_argument( - "--format", - choices=["text", "json"], - default="text", - help="Output format: 'text' for structured text, 'json' for JSON format (default: text)", - ) - parser.add_argument( - "--max-memory-length", - type=int, - default=200, - help="Maximum length of memory content to display in text format (default: 200)", - ) - args = parser.parse_args() - - # 1. Create Configuration from environment variables or JSON file - # Try to get config from environment variables first - openai_api_key = os.getenv("OPENAI_API_KEY") - if openai_api_key: - # Use environment variables (similar to server_router_api.py) - config_dict = get_reader_config() - reader_config = SimpleStructMemReaderConfig.model_validate(config_dict) - else: - # Fall back to JSON file - reader_config = SimpleStructMemReaderConfig.from_json_file( - "examples/data/config/simple_struct_reader_config.json" - ) - reader = SimpleStructMemReader(reader_config) - - # 2. Define scene data - scene_data = [ - [ - {"role": "user", "chat_time": "3 May 2025", "content": "I'm feeling a bit down today."}, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "I'm sorry to hear that. Do you want to talk about what's been going on?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "It's just been a tough couple of days, you know? Everything feels a bit overwhelming, and I just can't seem to shake it off.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "It sounds like you're going through a lot right now. Sometimes it helps to talk things out. Is there something specific that's been weighing on you, or is it more of a general feeling?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "It's a mix, I guess. Work's been really stressful, and on top of that, I've been feeling kinda disconnected from the people around me.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "That can be really draining, especially when you're feeling isolated on top of the stress. Do you think there's something from your past that's contributing to how you're feeling now? Sometimes our emotions are tied to older experiences.", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "Yeah, now that you mention it… I've been thinking a lot about my childhood lately. I didn't have the easiest upbringing. My parents were always busy, and I often felt like I had to figure things out on my own. I remember one winter, I got really sick with a high fever. My parents were both working, and I didn't want to bother them. I just lay in bed, crying by myself, not knowing what to do.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "That must have been very lonely. Growing up without that support can really make you feel like you have to handle everything on your own. How do you think that experience has affected you now?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "Yeah, it was really lonely. I remember spending a lot of time doing things on my own, like drawing, because it was my only way to escape. I used to draw animals, and I created my own little world on paper. It was peaceful in a way, just me and my imagination, with no outside distractions.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "Drawing sounds like a beautiful way to express yourself and find some peace in that solitude. It must have been a very quiet and private world for you. Do you think that sense of loneliness from back then has shaped how you connect with others today?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "I think it has. I've always found it hard to let others in. I feel like I have to handle everything on my own. If I rely on someone else, I feel vulnerable, and I start to feel that loneliness again.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "It sounds like this was a protective mechanism that helped you cope with your feelings of loneliness back then. But as you've grown, it seems like this pattern might not be serving you as well anymore. Have you thought about trying to break this cycle, or finding new ways to connect with others?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "I've thought about it, though it's hard. But I also know I can't keep going like this forever. Maybe it's time to seek some help.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "It's a big step, but a positive one. Exploring your emotions and past experiences could really help you understand how they've shaped your present. Therapy could be a great place to start, as it would help you connect with both yourself and others. Would you be open to trying it?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "I'm open to trying. Thank you for encouraging me. It feels like it might be a good starting point for me.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "I'm really glad to hear that. Taking that first step is always difficult, but you've made an important decision. If you need help finding the right therapist, or just want to talk, I'm here for you.", - }, - ], - [ - { - "role": "user", - "chat_time": "3 May 2025", - "content": "This week I've been feeling a bit off, especially when playing tennis. My body just doesn't feel right.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "It sounds like you've been having some physical discomfort lately. Where specifically are you feeling unwell? Is there something particular about tennis that's been difficult?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "I just feel a bit fatigued, and every time I play tennis, my movements feel off. I used to play pretty well, but lately, I've been feeling weak and my movements aren't as coordinated as they used to be.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "Physical discomfort can definitely affect performance, especially in a sport like tennis that requires quick reactions and coordination. Have you noticed anything specific that might be causing these changes? Could it be overtraining, or is there another physical issue making you feel off?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "I think it might be due to stress and lack of sleep recently. I just feel drained all the time. Plus, I've been frustrated with my tennis performance, which makes me feel even worse, like I'm stuck in a vicious cycle.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "Stress and poor sleep quality can definitely affect your physical performance, especially in sports. When you're also feeling down about your performance, it's easy to fall into a negative cycle. Have you tried anything to relieve some of the stress, or to improve your sleep?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "I've tried some relaxation techniques, like deep breathing and simple meditation, but it doesn't seem to help much. I still can't focus during tennis, and I feel like my mind is distracted by other things.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "Relaxation techniques like deep breathing and meditation can help reduce stress, but sometimes we need a more structured approach to improve both our physical and mental state. For example, you might want to focus on specific aspects of your tennis training, or adjust your rest and recovery time. Have you thought about setting smaller goals for yourself? Like focusing on a specific tennis move each day, rather than expecting perfection right away? That might help you gradually regain confidence.", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "That's a good idea. I think I've been expecting too much from myself, which just makes me more anxious. Maybe I can start with small practice sessions and gradually find my rhythm again.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "Exactly! Setting small and clear goals will help reduce pressure and give you a sense of accomplishment. And don't forget that rest is just as important—taking time to recover will help you feel better physically and mentally. Both physical and mental recovery are key to improving your performance and health.", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "Thanks, I feel like I need to rethink a few things. Maybe I really need to be kinder to myself and take things step by step.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "You're on the right track. Rethinking things and giving yourself space and patience is so important. You've already taken the first step, now just keep moving forward, one step at a time. If you need anything, I'm always here to help.", - }, - ], - [ - { - "role": "user", - "chat_time": "3 May 2025", - "content": "Yesterday at 3 PM I had a meeting with my team about the new project. We discussed the timeline and I'm concerned we won't meet the deadline of December 15th.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "I understand your concern about the project timeline. What specific challenges do you think might prevent your team from meeting the December 15th deadline?", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "We have too many requirements to implement in just two weeks. The backend team needs at least until December 10th to finish their part, which leaves almost no time for testing before the deadline.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "That's definitely a tight timeline. Have you considered prioritizing the requirements to focus on the most critical features first? You might also want to discuss with stakeholders about possibly extending the deadline or reducing the scope for this initial release.", - }, - { - "role": "user", - "chat_time": "3 May 2025", - "content": "That's a good point. We have a follow-up meeting scheduled for tomorrow morning at 9:30 AM. I'll suggest we prioritize features and maybe propose pushing the deadline to January 5th next year.", - }, - { - "role": "assistant", - "chat_time": "3 May 2025", - "content": "That sounds like a solid approach. Coming to the meeting with both a prioritization strategy and a specific alternative deadline shows you've thought through the problem carefully. It might also help to bring some data about similar past projects and how long they took to complete, if you have that information available.", - }, - ], - ] - - print("=== Mem-Reader Fast vs Fine Mode Comparison ===\n") - - # 3. Test Fine Mode (default) - print("🔄 Testing FINE mode (default, with LLM processing)...") - start_time = time.time() - fine_memory = reader.get_memory( - scene_data, type="chat", info={"user_id": "user1", "session_id": "session1"}, mode="fine" - ) - fine_time = time.time() - start_time - print(f"✅ Fine mode completed in {fine_time:.2f} seconds") - print(f"📊 Fine mode generated {sum(len(mem_list) for mem_list in fine_memory)} memory items") - - # 4. Test Fast Mode - print("\n⚡ Testing FAST mode (quick processing, no LLM calls)...") - start_time = time.time() - fast_memory = reader.get_memory( - scene_data, type="chat", info={"user_id": "user1", "session_id": "session1"}, mode="fast" - ) - fast_time = time.time() - start_time - print(f"✅ Fast mode completed in {fast_time:.2f} seconds") - print(f"📊 Fast mode generated {sum(len(mem_list) for mem_list in fast_memory)} memory items") - - # 5. Performance Comparison - print("\n📈 Performance Comparison:") - print(f" Fine mode: {fine_time:.2f}s") - print(f" Fast mode: {fast_time:.2f}s") - print(f" Speed improvement: {fine_time / fast_time:.1f}x faster") - - # 6. Show sample results from both modes - print("\n🔍 Sample Results Comparison:") - print("\n--- FINE Mode Results (first 3 items) ---") - for i, mem_list in enumerate(fine_memory[:3]): - for j, mem_item in enumerate(mem_list[:2]): # Show first 2 items from each list - print(f"\n[Scene {i}][Item {j}]") - if args.format == "json": - print_textual_memory_item_json(mem_item, indent=2) - else: - print_textual_memory_item( - mem_item, max_memory_length=args.max_memory_length, indent=2 - ) - - print("\n--- FAST Mode Results (first 3 items) ---") - for i, mem_list in enumerate(fast_memory[:3]): - for j, mem_item in enumerate(mem_list[:2]): # Show first 2 items from each list - print(f"\n[Scene {i}][Item {j}]") - if args.format == "json": - print_textual_memory_item_json(mem_item, indent=2) - else: - print_textual_memory_item( - mem_item, max_memory_length=args.max_memory_length, indent=2 - ) - - # 7. Example of transfer fast mode result into fine result - fast_mode_memories = [ - TextualMemoryItem( - id="4553141b-3a33-4548-b779-e677ec797a9f", - memory="user: Nate:Oh cool! I might check that one out some time soon! I do love watching classics.\nassistant: Joanna:Yep, that movie is awesome. I first watched it around 3 years ago. I even went out and got a physical copy!\nuser: Nate:Sounds cool! Have you seen it a lot? sounds like you know the movie well!\nassistant: Joanna:A few times. It's one of my favorites! I really like the idea and the acting.\nuser: Nate:Cool! I'll definitely check it out. Thanks for the recommendation!\nassistant: Joanna:No problem, Nate! Let me know if you like it!\n", - metadata=TreeNodeTextualMemoryMetadata( - user_id="nate_test", - session_id="root_session", - status="activated", - type="fact", - key="user: Nate:Oh cool", - confidence=0.9900000095367432, - source=None, - tags=["mode:fast", "lang:en", "role:assistant", "role:user"], - visibility=None, - updated_at="2025-10-16T17:16:30.094877+08:00", - memory_type="LongTermMemory", - sources=[ - SourceMessage( - type="chat", - role="user", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=0, - ), - SourceMessage( - type="chat", - role="assistant", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=1, - ), - SourceMessage( - type="chat", - role="user", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=2, - ), - SourceMessage( - type="chat", - role="assistant", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=3, - ), - SourceMessage( - type="chat", - role="user", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=4, - ), - SourceMessage( - type="chat", - role="assistant", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=5, - ), - ], - embedding=None, - created_at="2025-10-16T17:16:30.094919+08:00", - usage=[], - background="", - ), - ), - TextualMemoryItem( - id="752e42fa-92b6-491a-a430-6864a7730fba", - memory="user: Nate:It was! How about you? Do you have any hobbies you love?\nassistant: Joanna:Yeah! Besides writing, I also enjoy reading, watching movies, and exploring nature. Anything else you enjoy doing, Nate?\nuser: Nate:Playing video games and watching movies are my main hobbies.\nassistant: Joanna:Cool, Nate! So we both have similar interests. What type of movies do you like best?\nuser: Nate:I love action and sci-fi movies, the effects are so cool! What about you, what's your favorite genre?\nassistant: Joanna:I'm all about dramas and romcoms. I love getting immersed in the feelings and plots.\nuser: Nate:Wow, movies can be so powerful! Do you have any recommendations for me?\nassistant: Joanna:Yeah, totally! Have you seen this romantic drama that's all about memory and relationships? It's such a good one.\nuser: Nate:Oh cool! I might check that one out some time soon! I do love watching classics.\nassistant: Joanna:Yep, that movie is awesome. I first watched it around 3 years ago. I even went out and got a physical copy!\n", - metadata=TreeNodeTextualMemoryMetadata( - user_id="nate_test", - session_id="root_session", - status="activated", - type="fact", - key="user: Nate:It was", - confidence=0.9900000095367432, - source=None, - tags=["mode:fast", "lang:en", "role:assistant", "role:user"], - visibility=None, - updated_at="2025-10-16T17:16:30.095726+08:00", - memory_type="LongTermMemory", - sources=[ - SourceMessage( - type="chat", - role="user", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=0, - ), - SourceMessage( - type="chat", - role="assistant", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=1, - ), - SourceMessage( - type="chat", - role="user", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=2, - ), - SourceMessage( - type="chat", - role="assistant", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=3, - ), - SourceMessage( - type="chat", - role="user", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=4, - ), - SourceMessage( - type="chat", - role="assistant", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=5, - ), - SourceMessage( - type="chat", - role="user", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=6, - ), - SourceMessage( - type="chat", - role="assistant", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=7, - ), - SourceMessage( - type="chat", - role="user", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=8, - ), - SourceMessage( - type="chat", - role="assistant", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=9, - ), - ], - embedding=None, - created_at="2025-10-16T17:16:30.095767+08:00", - usage=[], - background="", - ), - ), - TextualMemoryItem( - id="c9cf448c-deee-43a8-bafd-eb15fde535b2", - memory="user: Nate:Hey Joanna! Long time no see! What's up? Anything fun going on?\nassistant: Joanna:Hey Nate! Long time no see! I've been working on a project lately - it's been pretty cool. What about you - any fun projects or hobbies?\nuser: Nate:Hey Joanna! That's cool! I won my first video game tournament last week - so exciting!\nassistant: Joanna:Wow Nate! Congrats on winning! Tell me more - what game was it?\nuser: Nate:Thanks! it's a team shooter game.\nassistant: Joanna:Wow, great job! What was is called?\nuser: Nate:The game was called Counter-Strike: Global Offensive, and me and my team had a blast to the very end!\nassistant: Joanna:Cool, Nate! Sounds like a fun experience, even if I'm not into games.\nuser: Nate:It was! How about you? Do you have any hobbies you love?\nassistant: Joanna:Yeah! Besides writing, I also enjoy reading, watching movies, and exploring nature. Anything else you enjoy doing, Nate?\n", - metadata=TreeNodeTextualMemoryMetadata( - user_id="nate_test", - session_id="root_session", - status="activated", - type="fact", - key="user: Nate:Hey Joanna", - confidence=0.9900000095367432, - source=None, - tags=["mode:fast", "lang:en", "role:assistant", "role:user"], - visibility=None, - updated_at="2025-10-16T17:16:30.098208+08:00", - memory_type="LongTermMemory", - sources=[ - SourceMessage( - type="chat", - role="user", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=0, - ), - SourceMessage( - type="chat", - role="assistant", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=1, - ), - SourceMessage( - type="chat", - role="user", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=2, - ), - SourceMessage( - type="chat", - role="assistant", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=3, - ), - SourceMessage( - type="chat", - role="user", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=4, - ), - SourceMessage( - type="chat", - role="assistant", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=5, - ), - SourceMessage( - type="chat", - role="user", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=6, - ), - SourceMessage( - type="chat", - role="assistant", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=7, - ), - SourceMessage( - type="chat", - role="user", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=8, - ), - SourceMessage( - type="chat", - role="assistant", - chat_time="7:31 pm on 21 January, 2022", - message_id=None, - content=None, - doc_path=None, - index=9, - ), - ], - embedding=None, - created_at="2025-10-16T17:16:30.098246+08:00", - usage=[], - background="", - ), - ), - ] - fine_memories = reader.fine_transfer_simple_mem(fast_mode_memories, type="chat") - print("\n--- Transfer Mode Results (first 3 items) ---") - for i, mem_list in enumerate(fine_memories[:3]): - for j, mem_item in enumerate(mem_list[:2]): # Show first 2 items from each list - print(f"\n[Scene {i}][Item {j}]") - if args.format == "json": - print_textual_memory_item_json(mem_item, indent=2) - else: - print_textual_memory_item( - mem_item, max_memory_length=args.max_memory_length, indent=2 - ) - - # 7. Example of processing documents (only in fine mode) - print("\n=== Processing Documents (Fine Mode Only) ===") - # Example document paths (you should replace these with actual document paths) - doc_paths = [ - "text1.txt", - "text2.txt", - ] - - try: - # 6. Acquiring memories from documents - doc_memory = reader.get_memory( - doc_paths, - "doc", - info={ - "user_id": "1111", - "session_id": "2222", - }, - mode="fine", - ) - total_items = sum(len(mem_list) for mem_list in doc_memory) - print(f"\n📄 Document Memory generated {total_items} items") - - # Print structured document memory items - if doc_memory: - print("\n--- Document Memory Items (first 3) ---") - for i, mem_list in enumerate(doc_memory[:3]): - for j, mem_item in enumerate(mem_list[:3]): # Show first 3 items from each document - print(f"\n[Document {i}][Item {j}]") - if args.format == "json": - print_textual_memory_item_json(mem_item, indent=2) - else: - print_textual_memory_item( - mem_item, max_memory_length=args.max_memory_length, indent=2 - ) - except Exception as e: - print(f"⚠️ Document processing failed: {e}") - print(" (This is expected if document files don't exist)") - - print("\n🎯 Summary:") - print(f" • Fast mode: {fast_time:.2f}s - Quick processing, no LLM calls") - print(f" • Fine mode: {fine_time:.2f}s - Full LLM processing for better understanding") - print(" • Use fast mode for: Real-time applications, high-throughput scenarios") - print(" • Use fine mode for: Quality analysis, detailed memory extraction") - - -if __name__ == "__main__": - main() diff --git a/examples/mem_reader/utils.py b/examples/mem_reader/utils.py new file mode 100644 index 000000000..0ac3e4d6a --- /dev/null +++ b/examples/mem_reader/utils.py @@ -0,0 +1,84 @@ +"""Utility functions for MemReader examples.""" + +import json +import pprint + +from typing import Any + +from memos.memories.textual.item import TextualMemoryItem + + +def _truncate(s: str, max_len: int | None) -> str: + if max_len is None or len(s) <= max_len: + return s + return s[:max_len] + "..." + + +def sanitize_for_print(obj: Any, *, max_str_len: int | None = 500) -> Any: + """ + Recursively sanitize data for pretty printing: + - Long strings are truncated + - Strings keep real newlines (so box printer can render multi-line) + """ + if isinstance(obj, str): + return _truncate(obj, max_str_len) + if isinstance(obj, dict): + return {k: sanitize_for_print(v, max_str_len=max_str_len) for k, v in obj.items()} + if isinstance(obj, list): + return [sanitize_for_print(v, max_str_len=max_str_len) for v in obj] + if isinstance(obj, tuple): + return tuple(sanitize_for_print(v, max_str_len=max_str_len) for v in obj) + return obj + + +def pretty_print_dict(d: dict, *, max_str_len: int | None = 500): + """Print a dictionary in a pretty bordered box (handles multiline strings).""" + d2 = sanitize_for_print(d, max_str_len=max_str_len) + + # Prefer JSON formatting if possible, fallback to pprint + try: + text = json.dumps(d2, indent=2, ensure_ascii=False) + except (TypeError, ValueError): + text = pprint.pformat(d2, indent=2, width=120) + + # Expand the JSON/pprint output into lines + lines: list[str] = [] + for line in text.splitlines(): + # If a line itself contains literal "\n" sequences (rare), leave it; + # real newlines are already split by splitlines(). + lines.append(line) + + # Prevent extremely wide boxes (optional safety) + max_len = max(len(line) for line in lines) if lines else 0 + border = "═" * (max_len + 4) + + print(f"╔{border}╗") + for line in lines: + print(f"║ {line.ljust(max_len)} ║") + print(f"╚{border}╝") + + +def print_memory_item( + item: TextualMemoryItem, + indent: int = 0, + max_memory_length: int | None = 300, # None = 不截断 +): + """Print a TextualMemoryItem in a structured format.""" + prefix = " " * indent + print(f"{prefix}--- Memory Item ---") + print(f"{prefix}Type: {item.metadata.memory_type}") + + mem = item.memory or "" + mem_preview = mem if max_memory_length is None else _truncate(mem, max_memory_length) + print(f"{prefix}Memory: {mem_preview}") + + if item.metadata.tags: + print(f"{prefix}Tags: {item.metadata.tags}") + + if item.metadata.confidence is not None: + print(f"{prefix}Confidence: {item.metadata.confidence}") + + if hasattr(item.metadata, "sources") and item.metadata.sources: + print(f"{prefix}Sources ({len(item.metadata.sources)}):") + for source in item.metadata.sources: + print(f"{prefix} - {source.type} (role: {getattr(source, 'role', 'N/A')})") diff --git a/examples/mem_scheduler/api_w_scheduler.py b/examples/mem_scheduler/api_w_scheduler.py index 871dd0258..b02161c12 100644 --- a/examples/mem_scheduler/api_w_scheduler.py +++ b/examples/mem_scheduler/api_w_scheduler.py @@ -1,85 +1,112 @@ +""" +# Prerequisites & Configuration +# To run this script, you must have the following services +# running and configured in your .env file (or environment variables): +# 1. Redis (Required for TaskStatusTracker and Scheduler Queue) +# 2. Graph Database (Required for Memory Storage) +# 3. Vector Database (Required if using Neo4j Community or Preference Memory) +""" + +import sys + +from pathlib import Path from time import sleep -from memos.api.handlers.scheduler_handler import ( + +FILE_PATH = Path(__file__).absolute() +BASE_DIR = FILE_PATH.parent.parent.parent +sys.path.insert(0, str(BASE_DIR)) # Enable execution from any working directory + +from memos.api.handlers.scheduler_handler import ( # noqa: E402 handle_scheduler_status, handle_scheduler_wait, ) -from memos.api.routers.server_router import mem_scheduler, status_tracker -from memos.mem_scheduler.schemas.message_schemas import ScheduleMessageItem - - -# Debug: Print scheduler configuration -print("=== Scheduler Configuration Debug ===") -print(f"Scheduler type: {type(mem_scheduler).__name__}") -print(f"Config: {mem_scheduler.config}") -print(f"use_redis_queue: {mem_scheduler.use_redis_queue}") -print(f"Queue type: {type(mem_scheduler.memos_message_queue).__name__}") -print(f"Queue maxsize: {getattr(mem_scheduler.memos_message_queue, 'maxsize', 'N/A')}") -print("=====================================\n") - -queue = mem_scheduler.memos_message_queue -queue.clear() - - -# 1. Define a handler function -def my_test_handler(messages: list[ScheduleMessageItem]): - print(f"My test handler received {len(messages)} messages:") - for msg in messages: - print(f" my_test_handler - {msg.item_id}: {msg.content}") - user_status_running = handle_scheduler_status( - user_id=msg.user_id, status_tracker=status_tracker - ) - print("[Monitor] Status after submit:", user_status_running) +from memos.api.routers.server_router import mem_scheduler, status_tracker # noqa: E402 +from memos.mem_scheduler.schemas.message_schemas import ScheduleMessageItem # noqa: E402 -# 2. Register the handler TEST_HANDLER_LABEL = "test_handler" TEST_USER_ID = "test_user" -mem_scheduler.register_handlers({TEST_HANDLER_LABEL: my_test_handler}) - -# 2.1 Monitor global scheduler status before submitting tasks -global_status_before = handle_scheduler_status(user_id=TEST_USER_ID, status_tracker=status_tracker) -print("[Monitor] Global status before submit:", global_status_before) - -# 3. Create messages -messages_to_send = [ - ScheduleMessageItem( - item_id=f"test_item_{i}", - user_id=TEST_USER_ID, - mem_cube_id="test_mem_cube", - label=TEST_HANDLER_LABEL, - content=f"This is test message {i}", +USER_MEM_CUBE = "test_mem_cube" + + +def run_with_scheduler_api(): + # Debug: Print scheduler configuration + print("=== Scheduler Configuration Debug ===") + print(f"Scheduler type: {type(mem_scheduler).__name__}") + print(f"Config: {mem_scheduler.config}") + print(f"use_redis_queue: {mem_scheduler.use_redis_queue}") + print(f"Queue type: {type(mem_scheduler.memos_message_queue).__name__}") + print(f"Queue maxsize: {getattr(mem_scheduler.memos_message_queue, 'maxsize', 'N/A')}") + print("=====================================\n") + + queue = mem_scheduler.memos_message_queue + queue.clear() + + # 1. Define a handler function + def my_test_handler(messages: list[ScheduleMessageItem]): + print(f"My test handler received {len(messages)} messages:") + for msg in messages: + print(f" my_test_handler - {msg.item_id}: {msg.content}") + user_status_running = handle_scheduler_status( + user_id=msg.user_id, status_tracker=status_tracker + ) + print("[Monitor] Status after submit:", user_status_running) + + # 2. Register the handler + mem_scheduler.register_handlers({TEST_HANDLER_LABEL: my_test_handler}) + + # 2.1 Monitor global scheduler status before submitting tasks + global_status_before = handle_scheduler_status( + user_id=TEST_USER_ID, status_tracker=status_tracker ) - for i in range(5) -] + print("[Monitor] Global status before submit:", global_status_before) + + # 3. Create messages + messages_to_send = [ + ScheduleMessageItem( + item_id=f"test_item_{i}", + user_id=TEST_USER_ID, + mem_cube_id="test_mem_cube", + label=TEST_HANDLER_LABEL, + content=f"This is test message {i}", + ) + for i in range(5) + ] -# 5. Submit messages -for mes in messages_to_send: - print(f"Submitting message {mes.item_id} to the scheduler...") - mem_scheduler.submit_messages([mes]) - sleep(1) + # 5. Submit messages + for mes in messages_to_send: + print(f"Submitting message {mes.item_id} to the scheduler...") + mem_scheduler.submit_messages([mes]) + sleep(1) -# 5.1 Monitor status for specific mem_cube while running -USER_MEM_CUBE = "test_mem_cube" + # 5.1 Monitor status for specific mem_cube while running + # 6. Wait for messages to be processed (limited to 100 checks) -# 6. Wait for messages to be processed (limited to 100 checks) + user_status_running = handle_scheduler_status( + user_id=TEST_USER_ID, status_tracker=status_tracker + ) + print(f"[Monitor] Status for {USER_MEM_CUBE} after submit:", user_status_running) + + # 6.1 Wait until idle for specific mem_cube via handler + wait_result = handle_scheduler_wait( + user_name=TEST_USER_ID, + status_tracker=status_tracker, + timeout_seconds=120.0, + poll_interval=0.5, + ) + print(f"[Monitor] Wait result for {USER_MEM_CUBE}:", wait_result) -user_status_running = handle_scheduler_status(user_id=TEST_USER_ID, status_tracker=status_tracker) -print(f"[Monitor] Status for {USER_MEM_CUBE} after submit:", user_status_running) + # 6.2 Monitor global scheduler status after processing + global_status_after = handle_scheduler_status( + user_id=TEST_USER_ID, status_tracker=status_tracker + ) + print("[Monitor] Global status after processing:", global_status_after) -# 6.1 Wait until idle for specific mem_cube via handler -wait_result = handle_scheduler_wait( - user_name=TEST_USER_ID, - status_tracker=status_tracker, - timeout_seconds=120.0, - poll_interval=0.5, -) -print(f"[Monitor] Wait result for {USER_MEM_CUBE}:", wait_result) + # 7. Stop the scheduler + print("Stopping the scheduler...") + mem_scheduler.stop() -# 6.2 Monitor global scheduler status after processing -global_status_after = handle_scheduler_status(user_id=TEST_USER_ID, status_tracker=status_tracker) -print("[Monitor] Global status after processing:", global_status_after) -# 7. Stop the scheduler -print("Stopping the scheduler...") -mem_scheduler.stop() +if __name__ == "__main__": + run_with_scheduler_api() diff --git a/examples/mem_scheduler/debug_text_mem_replace.py b/examples/mem_scheduler/debug_text_mem_replace.py deleted file mode 100644 index a5de8e572..000000000 --- a/examples/mem_scheduler/debug_text_mem_replace.py +++ /dev/null @@ -1,109 +0,0 @@ -import json -import shutil -import sys - -from pathlib import Path - -from memos_w_scheduler_for_test import init_task - -from memos.configs.mem_cube import GeneralMemCubeConfig -from memos.configs.mem_os import MOSConfig -from memos.configs.mem_scheduler import AuthConfig -from memos.log import get_logger -from memos.mem_cube.general import GeneralMemCube -from memos.mem_scheduler.analyzer.mos_for_test_scheduler import MOSForTestScheduler - - -FILE_PATH = Path(__file__).absolute() -BASE_DIR = FILE_PATH.parent.parent.parent -sys.path.insert(0, str(BASE_DIR)) - -# Enable execution from any working directory - -logger = get_logger(__name__) - -if __name__ == "__main__": - # set up data - conversations, questions = init_task() - - # set configs - mos_config = MOSConfig.from_yaml_file( - f"{BASE_DIR}/examples/data/config/mem_scheduler/memos_config_w_optimized_scheduler.yaml" - ) - - mem_cube_config = GeneralMemCubeConfig.from_yaml_file( - f"{BASE_DIR}/examples/data/config/mem_scheduler/mem_cube_config_neo4j.yaml" - ) - - # default local graphdb uri - if AuthConfig.default_config_exists(): - auth_config = AuthConfig.from_local_config() - - mos_config.mem_reader.config.llm.config.api_key = auth_config.openai.api_key - mos_config.mem_reader.config.llm.config.api_base = auth_config.openai.base_url - - mem_cube_config.text_mem.config.graph_db.config.uri = auth_config.graph_db.uri - mem_cube_config.text_mem.config.graph_db.config.user = auth_config.graph_db.user - mem_cube_config.text_mem.config.graph_db.config.password = auth_config.graph_db.password - mem_cube_config.text_mem.config.graph_db.config.db_name = auth_config.graph_db.db_name - mem_cube_config.text_mem.config.graph_db.config.auto_create = ( - auth_config.graph_db.auto_create - ) - - # Initialization - mos = MOSForTestScheduler(mos_config) - - user_id = "user_1" - mos.create_user(user_id) - - mem_cube_id = "mem_cube_5" - mem_cube_name_or_path = f"{BASE_DIR}/outputs/mem_scheduler/{user_id}/{mem_cube_id}" - - if Path(mem_cube_name_or_path).exists(): - shutil.rmtree(mem_cube_name_or_path) - print(f"{mem_cube_name_or_path} is not empty, and has been removed.") - - mem_cube = GeneralMemCube(mem_cube_config) - mem_cube.dump(mem_cube_name_or_path) - mos.register_mem_cube( - mem_cube_name_or_path=mem_cube_name_or_path, mem_cube_id=mem_cube_id, user_id=user_id - ) - - mos.add(conversations, user_id=user_id, mem_cube_id=mem_cube_id) - - # Add interfering conversations - file_path = Path(f"{BASE_DIR}/examples/data/mem_scheduler/scene_data.json") - scene_data = json.load(file_path.open("r", encoding="utf-8")) - mos.add(scene_data[0], user_id=user_id, mem_cube_id=mem_cube_id) - mos.add(scene_data[1], user_id=user_id, mem_cube_id=mem_cube_id) - - # Test the replace_working_memory functionality - print("\n--- Testing replace_working_memory ---") - - # Get current working memories - text_mem_base = mem_cube.text_mem - if text_mem_base is not None: - working_memories_before = text_mem_base.get_working_memory() - print(f"Working memories before replacement: {len(working_memories_before)}") - - # Create filtered memories (simulate what the scheduler would do) - # Keep only memories related to Max - filtered_memories = [working_memories_before[1], working_memories_before[4]] - - text_mem_base.replace_working_memory(memories=filtered_memories) - - # Check working memory after replacement - working_memories_after = text_mem_base.get_working_memory() - print(f"Working memories after replacement: {len(working_memories_after)}") - - if len(working_memories_after) == len(filtered_memories): - print("✅ SUCCESS: Working memory count matches filtered memories") - else: - print( - f"❌ FAILED: Expected {len(filtered_memories)}, got {len(working_memories_after)}" - ) - - else: - print("❌ text_mem is None - not properly initialized") - - mos.mem_scheduler.stop() diff --git a/examples/mem_scheduler/memos_w_scheduler.py b/examples/mem_scheduler/memos_w_scheduler.py index ef7d853df..b7250a677 100644 --- a/examples/mem_scheduler/memos_w_scheduler.py +++ b/examples/mem_scheduler/memos_w_scheduler.py @@ -1,32 +1,43 @@ -import re -import shutil +# Prerequisites & Configuration +# To run this script, you must have the following services +# running and configured in your .env file (or environment variables): +# 1. Redis (Required for TaskStatusTracker and Scheduler Queue) +# 2. Graph Database (Required for Memory Storage) +# 3. Vector Database (Required if using Neo4j Community or Preference Memory) + +import asyncio +import json +import os import sys +import time -from datetime import datetime from pathlib import Path -from memos.configs.mem_cube import GeneralMemCubeConfig -from memos.configs.mem_os import MOSConfig -from memos.configs.mem_scheduler import AuthConfig -from memos.log import get_logger -from memos.mem_cube.general import GeneralMemCube -from memos.mem_os.main import MOS -from memos.mem_scheduler.schemas.message_schemas import ScheduleLogForWebItem -from memos.mem_scheduler.schemas.task_schemas import ( - ADD_TASK_LABEL, - ANSWER_TASK_LABEL, - MEM_ARCHIVE_TASK_LABEL, - MEM_ORGANIZE_TASK_LABEL, - MEM_UPDATE_TASK_LABEL, - QUERY_TASK_LABEL, -) -from memos.mem_scheduler.utils.filter_utils import transform_name_to_key - +# Setup paths before imports that depend on them FILE_PATH = Path(__file__).absolute() BASE_DIR = FILE_PATH.parent.parent.parent sys.path.insert(0, str(BASE_DIR)) # Enable execution from any working directory +# Set environment variables before importing server_router to ensure components are initialized correctly +os.environ["ENABLE_CHAT_API"] = "true" + +from memos.api.product_models import APIADDRequest, ChatPlaygroundRequest # noqa: E402 + +# Import from server_router for initialization +from memos.api.routers.server_router import ( # noqa: E402 + add_handler, + chat_stream_playground, + mem_scheduler, +) +from memos.log import get_logger # noqa: E402 +from memos.mem_scheduler.schemas.message_schemas import ScheduleMessageItem # noqa: E402 +from memos.mem_scheduler.schemas.task_schemas import ( # noqa: E402 + MEM_UPDATE_TASK_LABEL, + QUERY_TASK_LABEL, +) + + logger = get_logger(__name__) @@ -73,150 +84,106 @@ def init_task(): return conversations, questions -def _truncate_with_rules(text: str) -> str: - has_cjk = bool(re.search(r"[\u4e00-\u9fff]", text)) - limit = 32 if has_cjk else 64 - normalized = text.strip().replace("\n", " ") - if len(normalized) <= limit: - return normalized - return normalized[:limit] + "..." - - -def _format_title(ts: datetime, title_text: str) -> str: - return f"{ts.astimezone().strftime('%H:%M:%S')} {title_text}" - - -def _cube_display_from(mem_cube_id: str) -> str: - if "public" in (mem_cube_id or "").lower(): - return "PublicMemCube" - return "UserMemCube" - - -_TYPE_SHORT = { - "LongTermMemory": "LTM", - "UserMemory": "User", - "WorkingMemory": "Working", - "ActivationMemory": "Activation", - "ParameterMemory": "Parameter", - "TextMemory": "Text", - "UserInput": "Input", - "NotApplicable": "NA", -} - - -def _format_entry(item: ScheduleLogForWebItem) -> tuple[str, str]: - cube_display = getattr(item, "memcube_name", None) or _cube_display_from(item.mem_cube_id) - label = item.label - content = item.log_content or "" - memcube_content = getattr(item, "memcube_log_content", None) or [] - memory_len = getattr(item, "memory_len", None) or len(memcube_content) or 1 - - def _first_content() -> str: - if memcube_content: - return memcube_content[0].get("content", "") or content - return content - - if label in ("addMessage", QUERY_TASK_LABEL, ANSWER_TASK_LABEL): - target_cube = cube_display.replace("MemCube", "") - title = _format_title(item.timestamp, f"addMessages to {target_cube} MemCube") - return title, _truncate_with_rules(_first_content()) - - if label in ("addMemory", ADD_TASK_LABEL): - title = _format_title(item.timestamp, f"{cube_display} added {memory_len} memories") - return title, _truncate_with_rules(_first_content()) - - if label in ("updateMemory", MEM_UPDATE_TASK_LABEL): - title = _format_title(item.timestamp, f"{cube_display} updated {memory_len} memories") - return title, _truncate_with_rules(_first_content()) - - if label in ("archiveMemory", MEM_ARCHIVE_TASK_LABEL): - title = _format_title(item.timestamp, f"{cube_display} archived {memory_len} memories") - return title, _truncate_with_rules(_first_content()) - - if label in ("mergeMemory", MEM_ORGANIZE_TASK_LABEL): - title = _format_title(item.timestamp, f"{cube_display} merged {memory_len} memories") - merged = [c for c in memcube_content if c.get("type") == "merged"] - post = [c for c in memcube_content if c.get("type") == "postMerge"] - parts = [] - if merged: - parts.append("Merged: " + " | ".join(c.get("content", "") for c in merged)) - if post: - parts.append("Result: " + " | ".join(c.get("content", "") for c in post)) - detail = " ".join(parts) if parts else _first_content() - return title, _truncate_with_rules(detail) - - if label == "scheduleMemory": - title = _format_title(item.timestamp, f"{cube_display} scheduled {memory_len} memories") - if memcube_content: - return title, _truncate_with_rules(memcube_content[0].get("content", "")) - key = transform_name_to_key(content) - from_short = _TYPE_SHORT.get(item.from_memory_type, item.from_memory_type) - to_short = _TYPE_SHORT.get(item.to_memory_type, item.to_memory_type) - return title, _truncate_with_rules(f"[{from_short}→{to_short}] {key}: {content}") - - title = _format_title(item.timestamp, f"{cube_display} event") - return title, _truncate_with_rules(_first_content()) - - -def run_with_scheduler_init(): - print("==== run_with_automatic_scheduler_init ====") - conversations, questions = init_task() - - # set configs - mos_config = MOSConfig.from_yaml_file( - f"{BASE_DIR}/examples/data/config/mem_scheduler/memos_config_w_optimized_scheduler.yaml" - ) - - mem_cube_config = GeneralMemCubeConfig.from_yaml_file( - f"{BASE_DIR}/examples/data/config/mem_scheduler/mem_cube_config_neo4j.yaml" - ) - - # default local graphdb uri - if AuthConfig.default_config_exists(): - auth_config = AuthConfig.from_local_config() +working_memories = [] + + +# Define custom query handler function +def custom_query_handler(messages: list[ScheduleMessageItem]): + for msg in messages: + # Print user input content + print(f"\n[scheduler] User input query: {msg.content}") + # Manually construct a new message with MEM_UPDATE label to trigger memory update + new_msg = msg.model_copy(update={"label": MEM_UPDATE_TASK_LABEL}) + # Submit the message to the scheduler for processing + mem_scheduler.submit_messages([new_msg]) + + +# Define custom memory update handler function +def custom_mem_update_handler(messages: list[ScheduleMessageItem]): + global working_memories + search_args = {} + top_k = 2 + for msg in messages: + # Search for memories relevant to the current content in text memory (return top_k=2) + results = mem_scheduler.retriever.search( + query=msg.content, + user_id=msg.user_id, + mem_cube_id=msg.mem_cube_id, + mem_cube=mem_scheduler.current_mem_cube, + top_k=top_k, + method=mem_scheduler.search_method, + search_args=search_args, + ) + working_memories.extend(results) + working_memories = working_memories[-5:] + for mem in results: + print(f"\n[scheduler] Retrieved memory: {mem.memory}") - mos_config.mem_reader.config.llm.config.api_key = auth_config.openai.api_key - mos_config.mem_reader.config.llm.config.api_base = auth_config.openai.base_url - mem_cube_config.text_mem.config.graph_db.config.uri = auth_config.graph_db.uri - mem_cube_config.text_mem.config.graph_db.config.user = auth_config.graph_db.user - mem_cube_config.text_mem.config.graph_db.config.password = auth_config.graph_db.password - mem_cube_config.text_mem.config.graph_db.config.db_name = auth_config.graph_db.db_name - mem_cube_config.text_mem.config.graph_db.config.auto_create = ( - auth_config.graph_db.auto_create - ) +async def run_with_scheduler(): + print("==== run_with_automatic_scheduler_init ====") + conversations, questions = init_task() - # Initialization - mos = MOS(mos_config) + # Initialization using server_router components + # Configs are loaded via environment variables in init_server() user_id = "user_1" - mos.create_user(user_id) - mem_cube_id = "mem_cube_5" - mem_cube_name_or_path = f"{BASE_DIR}/outputs/mem_scheduler/{user_id}/{mem_cube_id}" - if Path(mem_cube_name_or_path).exists(): - shutil.rmtree(mem_cube_name_or_path) - print(f"{mem_cube_name_or_path} is not empty, and has been removed.") + print(f"Adding conversations for user {user_id}...") - mem_cube = GeneralMemCube(mem_cube_config) - mem_cube.dump(mem_cube_name_or_path) - mos.register_mem_cube( - mem_cube_name_or_path=mem_cube_name_or_path, mem_cube_id=mem_cube_id, user_id=user_id + # Use add_handler to add memories + add_req = APIADDRequest( + user_id=user_id, + writable_cube_ids=[mem_cube_id], + messages=conversations, + async_mode="sync", # Use sync mode for immediate addition in this example ) - - mos.add(conversations, user_id=user_id, mem_cube_id=mem_cube_id) - mos.mem_scheduler.current_mem_cube = mem_cube + add_handler.handle_add_memories(add_req) for item in questions: print("===== Chat Start =====") query = item["question"] print(f"Query:\n {query}\n") - response = mos.chat(query=query, user_id=user_id) - print(f"Answer:\n {response}\n") - mos.mem_scheduler.stop() + # Use chat_handler to chat + chat_req = ChatPlaygroundRequest( + user_id=user_id, + query=query, + readable_cube_ids=[mem_cube_id], + writable_cube_ids=[mem_cube_id], + ) + response = chat_stream_playground(chat_req) + + answer = "" + buffer = "" + async for chunk in response.body_iterator: + if isinstance(chunk, bytes): + chunk = chunk.decode("utf-8") + buffer += chunk + while "\n\n" in buffer: + msg, buffer = buffer.split("\n\n", 1) + for line in msg.split("\n"): + if line.startswith("data: "): + json_str = line[6:] + try: + data = json.loads(json_str) + if data.get("type") == "text": + answer += data["data"] + except json.JSONDecodeError: + pass + print(f"\nAnswer: {answer}") if __name__ == "__main__": - run_with_scheduler_init() + mem_scheduler.register_handlers( + { + QUERY_TASK_LABEL: custom_query_handler, # Query task + MEM_UPDATE_TASK_LABEL: custom_mem_update_handler, # Memory update task + } + ) + + asyncio.run(run_with_scheduler()) + + time.sleep(20) + mem_scheduler.stop() diff --git a/examples/mem_scheduler/memos_w_scheduler_for_test.py b/examples/mem_scheduler/memos_w_scheduler_for_test.py deleted file mode 100644 index 2e135f127..000000000 --- a/examples/mem_scheduler/memos_w_scheduler_for_test.py +++ /dev/null @@ -1,88 +0,0 @@ -import json -import shutil -import sys - -from pathlib import Path - -from memos_w_scheduler import init_task - -from memos.configs.mem_cube import GeneralMemCubeConfig -from memos.configs.mem_os import MOSConfig -from memos.configs.mem_scheduler import AuthConfig -from memos.log import get_logger -from memos.mem_cube.general import GeneralMemCube -from memos.mem_scheduler.analyzer.mos_for_test_scheduler import MOSForTestScheduler - - -FILE_PATH = Path(__file__).absolute() -BASE_DIR = FILE_PATH.parent.parent.parent -sys.path.insert(0, str(BASE_DIR)) - -# Enable execution from any working directory - -logger = get_logger(__name__) - -if __name__ == "__main__": - # set up data - conversations, questions = init_task() - - # set configs - mos_config = MOSConfig.from_yaml_file( - f"{BASE_DIR}/examples/data/config/mem_scheduler/memos_config_w_optimized_scheduler.yaml" - ) - - mem_cube_config = GeneralMemCubeConfig.from_yaml_file( - f"{BASE_DIR}/examples/data/config/mem_scheduler/mem_cube_config_neo4j.yaml" - ) - - # default local graphdb uri - if AuthConfig.default_config_exists(): - auth_config = AuthConfig.from_local_config() - - mos_config.mem_reader.config.llm.config.api_key = auth_config.openai.api_key - mos_config.mem_reader.config.llm.config.api_base = auth_config.openai.base_url - - mem_cube_config.text_mem.config.graph_db.config.uri = auth_config.graph_db.uri - mem_cube_config.text_mem.config.graph_db.config.user = auth_config.graph_db.user - mem_cube_config.text_mem.config.graph_db.config.password = auth_config.graph_db.password - mem_cube_config.text_mem.config.graph_db.config.db_name = auth_config.graph_db.db_name - mem_cube_config.text_mem.config.graph_db.config.auto_create = ( - auth_config.graph_db.auto_create - ) - - # Initialization - mos = MOSForTestScheduler(mos_config) - - user_id = "user_1" - mos.create_user(user_id) - - mem_cube_id = "mem_cube_5" - mem_cube_name_or_path = f"{BASE_DIR}/outputs/mem_scheduler/{user_id}/{mem_cube_id}" - - if Path(mem_cube_name_or_path).exists(): - shutil.rmtree(mem_cube_name_or_path) - print(f"{mem_cube_name_or_path} is not empty, and has been removed.") - - mem_cube = GeneralMemCube(mem_cube_config) - mem_cube.dump(mem_cube_name_or_path) - mos.register_mem_cube( - mem_cube_name_or_path=mem_cube_name_or_path, mem_cube_id=mem_cube_id, user_id=user_id - ) - mos.mem_scheduler.current_mem_cube = mem_cube - - mos.add(conversations, user_id=user_id, mem_cube_id=mem_cube_id) - - # Add interfering conversations - file_path = Path(f"{BASE_DIR}/examples/data/mem_scheduler/scene_data.json") - scene_data = json.load(file_path.open("r", encoding="utf-8")) - mos.add(scene_data[0], user_id=user_id, mem_cube_id=mem_cube_id) - mos.add(scene_data[1], user_id=user_id, mem_cube_id=mem_cube_id) - - for item in questions: - print("===== Chat Start =====") - query = item["question"] - print(f"Query:\n {query}\n") - response = mos.chat(query=query, user_id=user_id) - print(f"Answer:\n {response}\n") - - mos.mem_scheduler.stop() diff --git a/examples/mem_scheduler/quick_start_examples.py b/examples/mem_scheduler/quick_start_examples.py deleted file mode 100644 index 724663be6..000000000 --- a/examples/mem_scheduler/quick_start_examples.py +++ /dev/null @@ -1,312 +0,0 @@ -import json -import shutil -import sys -import uuid - -from pathlib import Path - -from transformers import DynamicCache - -from memos.configs.mem_cube import GeneralMemCubeConfig -from memos.configs.mem_os import MOSConfig -from memos.configs.memory import MemoryConfigFactory -from memos.mem_cube.general import GeneralMemCube -from memos.mem_os.main import MOS -from memos.mem_scheduler.schemas.message_schemas import ScheduleMessageItem -from memos.mem_scheduler.schemas.task_schemas import ( - ANSWER_TASK_LABEL, - MEM_UPDATE_TASK_LABEL, - QUERY_TASK_LABEL, -) -from memos.mem_scheduler.utils.db_utils import get_utc_now -from memos.mem_scheduler.utils.misc_utils import parse_yaml -from memos.memories.activation.item import KVCacheItem -from memos.memories.factory import MemoryFactory - - -FILE_PATH = Path(__file__).absolute() -BASE_DIR = FILE_PATH.parent.parent.parent -sys.path.insert(0, str(BASE_DIR)) # Enable execution from any working directory - - -def get_cache_info(cache): - if not cache: - return None - - num_layers = 0 - total_size_bytes = 0 - - if hasattr(cache, "layers"): - num_layers = len(cache.layers) - for layer in cache.layers: - if hasattr(layer, "key_cache") and layer.key_cache is not None: - total_size_bytes += layer.key_cache.nelement() * layer.key_cache.element_size() - if hasattr(layer, "value_cache") and layer.value_cache is not None: - total_size_bytes += layer.value_cache.nelement() * layer.value_cache.element_size() - - if hasattr(layer, "keys") and layer.keys is not None: - total_size_bytes += layer.keys.nelement() * layer.keys.element_size() - if hasattr(layer, "values") and layer.values is not None: - total_size_bytes += layer.values.nelement() * layer.values.element_size() - - elif hasattr(cache, "key_cache") and hasattr(cache, "value_cache"): - num_layers = len(cache.key_cache) - for k, v in zip(cache.key_cache, cache.value_cache, strict=False): - if k is not None: - total_size_bytes += k.nelement() * k.element_size() - if v is not None: - total_size_bytes += v.nelement() * v.element_size() - - return { - "num_layers": num_layers, - "size_bytes": total_size_bytes, - "size_mb": f"{total_size_bytes / (1024 * 1024):.2f} MB", - } - - -def serialize_item(obj): - if isinstance(obj, list): - return [serialize_item(x) for x in obj] - - if isinstance(obj, KVCacheItem): - return { - "id": obj.id, - "metadata": obj.metadata, - "records": obj.records.model_dump() - if hasattr(obj.records, "model_dump") - else obj.records, - "memory": get_cache_info(obj.memory), - } - - if isinstance(obj, DynamicCache): - return get_cache_info(obj) - - return str(obj) - - -def kv_cache_only(): - # 为 KVCacheMemory(HuggingFace 后端)创建配置 - config = MemoryConfigFactory( - backend="kv_cache", - config={ - "extractor_llm": { - "backend": "huggingface", - "config": { - "model_name_or_path": "Qwen/Qwen3-0.6B", - "max_tokens": 32, - "add_generation_prompt": True, - "remove_think_prefix": True, - }, - }, - }, - ) - - # 实例化 KVCacheMemory - kv_mem = MemoryFactory.from_config(config) - - # 提取一个 KVCacheItem(DynamicCache) - prompt = [ - {"role": "user", "content": "What is MemOS?"}, - {"role": "assistant", "content": "MemOS is a memory operating system for LLMs."}, - ] - print("===== Extract KVCacheItem =====") - cache_item = kv_mem.extract(prompt) - print(json.dumps(serialize_item(cache_item), indent=2, default=str)) - - # 将缓存添加到内存中 - kv_mem.add([cache_item]) - print("All caches:") - print(json.dumps(serialize_item(kv_mem.get_all()), indent=2, default=str)) - - # 通过 ID 获取 - retrieved = kv_mem.get(cache_item.id) - print("Retrieved:") - print(json.dumps(serialize_item(retrieved), indent=2, default=str)) - - # 合并缓存 - item2 = kv_mem.extract([{"role": "user", "content": "Tell me a joke."}]) - kv_mem.add([item2]) - merged = kv_mem.get_cache([cache_item.id, item2.id]) - print("Merged cache:") - print(json.dumps(serialize_item(merged), indent=2, default=str)) - - # 删除其中一个 - kv_mem.delete([cache_item.id]) - print("After delete:") - print(json.dumps(serialize_item(kv_mem.get_all()), indent=2, default=str)) - - # 导出和加载缓存 - kv_mem.dump("tmp/kv_mem") - print("Dumped to tmp/kv_mem") - kv_mem.delete_all() - kv_mem.load("tmp/kv_mem") - print("Loaded caches:") - print(json.dumps(serialize_item(kv_mem.get_all()), indent=2, default=str)) - - -def run_scheduler_example(): - # 使用 MemScheduler 加载主 MOS(Memory-Oriented System)配置文件 - config = parse_yaml( - f"{BASE_DIR}/examples/data/config/mem_scheduler/memos_config_w_scheduler.yaml" - ) - # 将解析出的配置字典传入 MOSConfig 构造器, 构建配置对象 - mos_config = MOSConfig(**config) - # 使用配置对象初始化 MOS 系统实例 - mos = MOS(mos_config) - - # 生成一个唯一的动态用户 ID(使用 UUID4) - user_id = str(uuid.uuid4()) - # 在 MOS 系统中为该用户创建账户 - mos.create_user(user_id=user_id) - - # 从 YAML 文件加载 MemCube(记忆立方体)的通用配置 - config = GeneralMemCubeConfig.from_yaml_file( - f"{BASE_DIR}/examples/data/config/mem_scheduler/mem_cube_config.yaml" - ) - # 定义 MemCube 的唯一标识符 - mem_cube_id = "mem_cube_5" - # 定义 MemCube 的本地存储路径(路径中包含用户 ID 和 MemCube ID) - mem_cube_name_or_path = f"{BASE_DIR}/outputs/mem_scheduler/{user_id}/{mem_cube_id}" - - # 如果该路径已存在, 则先删除旧目录 - if Path(mem_cube_name_or_path).exists(): - shutil.rmtree(mem_cube_name_or_path) - print(f"{mem_cube_name_or_path} 目录非空，已被删除。") - - # 根据加载的配置创建一个新的 MemCube 实例 - mem_cube = GeneralMemCube(config) - # 将该 MemCube 实例序列化并保存到指定路径 - mem_cube.dump(mem_cube_name_or_path) - - # 在 MOS 系统中为当前用户注册这个 MemCube - mos.register_mem_cube( - mem_cube_name_or_path=mem_cube_name_or_path, mem_cube_id=mem_cube_id, user_id=user_id - ) - - # 定义一个辅助函数, 用于获取缓存(如 KV Cache)的内存信息 - def get_cache_info(cache): - # 如果缓存为空, 则直接返回 None - if not cache: - return None - - num_layers = 0 # 记录缓存的层数 - total_size_bytes = 0 # 记录总字节数 - - # 情况一: 缓存结构包含 layers 属性(如 HuggingFace 的缓存格式) - if hasattr(cache, "layers"): - num_layers = len(cache.layers) - for layer in cache.layers: - # 统计 key_cache 的内存占用(如果存在) - if hasattr(layer, "key_cache") and layer.key_cache is not None: - total_size_bytes += layer.key_cache.nelement() * layer.key_cache.element_size() - # 统计 value_cache 的内存占用(如果存在) - if hasattr(layer, "value_cache") and layer.value_cache is not None: - total_size_bytes += ( - layer.value_cache.nelement() * layer.value_cache.element_size() - ) - - # 兼容其他可能的缓存命名方式(如 keys/values) - if hasattr(layer, "keys") and layer.keys is not None: - total_size_bytes += layer.keys.nelement() * layer.keys.element_size() - if hasattr(layer, "values") and layer.values is not None: - total_size_bytes += layer.values.nelement() * layer.values.element_size() - - # 情况二: 缓存结构直接包含 key_cache 和 value_cache 列表(如某些自定义格式) - elif hasattr(cache, "key_cache") and hasattr(cache, "value_cache"): - num_layers = len(cache.key_cache) - for k, v in zip(cache.key_cache, cache.value_cache, strict=False): - if k is not None: - total_size_bytes += k.nelement() * k.element_size() - if v is not None: - total_size_bytes += v.nelement() * v.element_size() - - # 返回结构化的缓存信息, 包括层数, 字节数和以 MB 为单位的可读格式 - return { - "num_layers": num_layers, - "size_bytes": total_size_bytes, - "size_mb": f"{total_size_bytes / (1024 * 1024):.2f} MB", - } - - # 定义自定义的查询(query)处理函数 - def custom_query_handler(messages: list[ScheduleMessageItem]): - for msg in messages: - # 打印用户输入内容 - print(f"\n[scheduler] 用户输入了查询：{msg.content}") - # 手动构造一个带有 MEM_UPDATE 标签的新消息, 用于触发记忆更新 - new_msg = msg.model_copy(update={"label": MEM_UPDATE_TASK_LABEL}) - # 将该消息提交给调度器处理 - mos.mem_scheduler.submit_messages([new_msg]) - - # 定义自定义的回答(answer)处理函数 - def custom_answer_handler(messages: list[ScheduleMessageItem]): - for msg in messages: - # 打印 LLM 的回复内容 - print(f"\n[scheduler] LLM 回复了答案：{msg.content}") - - # 定义自定义的记忆更新(mem_update)处理函数 - def custom_mem_update_handler(messages: list[ScheduleMessageItem]): - for msg in messages: - mem_cube = mos.mem_cubes.get(msg.mem_cube_id) - kv_mem = mem_cube.act_mem - # 如果该 MemCube 配置了文本记忆(TreeTextMemory / NaiveTextMemory) - if mem_cube and mem_cube.text_mem: - # 在文本记忆中搜索与当前内容相关的记忆(返回 top_k=3 条) - results = mem_cube.text_mem.search(msg.content, top_k=3) - for mem in results: - print(f"\n[scheduler] 检索到的记忆：{mem.memory}") - print("\n[scheduler] 转换为激活记忆......") - # 从文本记忆中提取对应的 KV 缓存项 - cache_item = kv_mem.extract(mem.memory) - # 附加元信息 - cache_item.records.text_memories = [mem.memory] - cache_item.records.timestamp = get_utc_now() - # 将该缓存项添加到激活记忆中 - kv_mem.add([cache_item]) - print("\n[scheduler] 完成！") - - # 将上述三个自定义处理器注册到调度器的分发器中, 分别对应不同任务标签 - mos.mem_scheduler.dispatcher.register_handlers( - { - QUERY_TASK_LABEL: custom_query_handler, # 查询任务 - ANSWER_TASK_LABEL: custom_answer_handler, # 回答任务 - MEM_UPDATE_TASK_LABEL: custom_mem_update_handler, # 记忆更新任务 - } - ) - - # 初始添加两条测试消息(用户和助手的对话)到系统中 - messages = [ - {"role": "user", "content": "I like playing football."}, - {"role": "assistant", "content": "I like playing football too."}, - ] - mos.add(messages, user_id=user_id, mem_cube_id=mem_cube_id) - - # 进入聊天循环: 展示 TreeTextMemory 的记忆节点结构 + KV Cache 的状态 - while True: - # 获取用户输入并去除首尾空格 - user_input = input("👤 [You] ").strip() - print() - # 调用 MOS 系统进行聊天响应 - response = mos.chat(user_input, user_id=user_id) - # 获取该用户当前 MemCube 中的所有记忆内容 - retrieved_memories = mos.get_all(mem_cube_id=mem_cube_id, user_id=user_id) - - # 打印助手的回复 - print(f"🤖 [Assistant] {response}") - - # 获取文本记忆部分 - TreeTextMemory - memories = retrieved_memories["text_mem"][0]["memories"] - for mem in memories: - print(f"[文本记忆] {mem.memory}") - - # 获取对应的 MemCube 和其激活记忆(KV Cache) - mem_cube = mos.mem_scheduler.mem_cube - kv_mem = mem_cube.act_mem - # 遍历所有激活记忆项, 打印其缓存信息和记录 - for cache_item in kv_mem.get_all(): - print(f"[激活记忆] {get_cache_info(cache_item.memory)} （记录：{cache_item.records}）") - - -if __name__ == "__main__": - kv_cache_only() - - run_scheduler_example() diff --git a/examples/mem_scheduler/rabbitmq_example.py b/examples/mem_scheduler/rabbitmq_example.py deleted file mode 100644 index 5e40eaad5..000000000 --- a/examples/mem_scheduler/rabbitmq_example.py +++ /dev/null @@ -1,60 +0,0 @@ -import threading -import time - -from memos.configs.mem_scheduler import AuthConfig -from memos.mem_scheduler.webservice_modules.rabbitmq_service import RabbitMQSchedulerModule - - -def publish_message(rabbitmq_module, message): - """Function to publish a message.""" - rabbitmq_module.rabbitmq_publish_message(message) - print(f"Published message: {message}\n") - - -def main(): - # Initialize RabbitMQ module - rabbitmq_module = RabbitMQSchedulerModule() - rabbitmq_module.rabbit_queue_name = "test" - - # Initialize from configuration dictionary - if not AuthConfig.default_config_exists(): - print("Please set configs for rabbitmq.") - return - else: - rabbitmq_module.initialize_rabbitmq(config=AuthConfig.from_local_config().rabbitmq) - - try: - rabbitmq_module.wait_for_connection_ready() - - # === Publish some test messages === - # List to hold thread references - threads = [] - - # Publish some test messages using multiple threads - for i in range(3): - message = {"type": "test", "data": f"Message {i}", "timestamp": time.time()} - thread = threading.Thread(target=publish_message, args=(rabbitmq_module, message)) - thread.start() - threads.append(thread) - - # Start consumer - rabbitmq_module.rabbitmq_start_consuming() - - # Join threads to ensure all messages are published before proceeding - for thread in threads: - thread.join() - - except KeyboardInterrupt: - print("\nProgram interrupted by user") - - finally: - # Give some time for cleanup - time.sleep(3) - - # Close connections - rabbitmq_module.rabbitmq_close() - print("RabbitMQ connection closed") - - -if __name__ == "__main__": - main() diff --git a/examples/mem_scheduler/redis_example.py b/examples/mem_scheduler/redis_example.py index be6f20bed..00989c424 100644 --- a/examples/mem_scheduler/redis_example.py +++ b/examples/mem_scheduler/redis_example.py @@ -1,5 +1,13 @@ -import asyncio +# Prerequisites: +# 1. Ensure a Redis server is running locally on the default port (6379). +# You can start it with: `redis-server` +# On macOS with Homebrew: `/opt/homebrew/bin/redis-server` or `brew services start redis` +# On Linux: `sudo service redis-server start` +# 2. If Redis is running on a different host/port, update the configuration or environment variables accordingly. + + import sys +import time from datetime import datetime from pathlib import Path @@ -64,7 +72,7 @@ def service_run(): print( f"Added: {res}", ) - asyncio.sleep(0.5) + time.sleep(0.5) mem_scheduler.redis_stop_listening() diff --git a/examples/mem_scheduler/scheduler_for_async_tasks.py b/examples/mem_scheduler/run_async_tasks.py similarity index 90% rename from examples/mem_scheduler/scheduler_for_async_tasks.py rename to examples/mem_scheduler/run_async_tasks.py index 7f544c3da..fbb5a7bcc 100644 --- a/examples/mem_scheduler/scheduler_for_async_tasks.py +++ b/examples/mem_scheduler/run_async_tasks.py @@ -1,3 +1,12 @@ +""" +# Prerequisites & Configuration +# To run this script, you must have the following services +# running and configured in your .env file (or environment variables): +# 1. Redis (Required for TaskStatusTracker and Scheduler Queue) +# 2. Graph Database (Required for Memory Storage) +# 3. Vector Database (Required if using Neo4j Community or Preference Memory) +""" + from pathlib import Path from time import sleep diff --git a/examples/mem_scheduler/show_redis_status.py b/examples/mem_scheduler/show_redis_status.py index 04e79ca97..94a915588 100644 --- a/examples/mem_scheduler/show_redis_status.py +++ b/examples/mem_scheduler/show_redis_status.py @@ -1,56 +1,67 @@ +""" +# Prerequisites: +# 1. Ensure a Redis server is running locally on the default port (6379). +# You can start it with: `redis-server` +# On macOS with Homebrew: `/opt/homebrew/bin/redis-server` or `brew services start redis` +# On Linux: `sudo service redis-server start` +# 2. If Redis is running on a different host/port, update the configuration or environment variables accordingly. +""" + import time -from memos.api.routers.server_router import mem_scheduler +from memos.mem_scheduler.task_schedule_modules.orchestrator import SchedulerOrchestrator from memos.mem_scheduler.task_schedule_modules.redis_queue import SchedulerRedisQueue -queue = mem_scheduler.memos_message_queue.memos_message_queue +# Explicitly initialize Redis queue for monitoring +queue = SchedulerRedisQueue( + max_len=None, + consumer_group="scheduler_group", + consumer_name="monitor_consumer", + orchestrator=SchedulerOrchestrator(), +) -def fetch_status(queue: SchedulerRedisQueue) -> dict[str, dict[str, int]]: +def fetch_status( + queue: SchedulerRedisQueue, stream_key_prefix: str | None = None +) -> dict[str, dict[str, int]]: """Fetch and print per-user Redis queue status using built-in API. - Returns a dict mapping user_id -> {"pending": int, "remaining": int}. + Returns a dict mapping user_id -> {"remaining": int}. """ # This method will also print a summary and per-user counts. - return queue.show_task_status() + return queue.show_task_status(stream_key_prefix=stream_key_prefix) def print_diff(prev: dict[str, dict[str, int]], curr: dict[str, dict[str, int]]) -> None: """Print aggregated totals and per-user changes compared to previous snapshot.""" ts = time.strftime("%Y-%m-%d %H:%M:%S") - tot_p_prev = sum(v.get("pending", 0) for v in prev.values()) if prev else 0 tot_r_prev = sum(v.get("remaining", 0) for v in prev.values()) if prev else 0 - tot_p_curr = sum(v.get("pending", 0) for v in curr.values()) tot_r_curr = sum(v.get("remaining", 0) for v in curr.values()) - dp_tot = tot_p_curr - tot_p_prev dr_tot = tot_r_curr - tot_r_prev - print(f"[{ts}] Total pending={tot_p_curr} ({dp_tot:+d}), remaining={tot_r_curr} ({dr_tot:+d})") + print(f"[{ts}] Total remaining={tot_r_curr} ({dr_tot:+d})") # Print per-user deltas (current counts are already printed by show_task_status) all_uids = sorted(set(prev.keys()) | set(curr.keys())) for uid in all_uids: - p_prev = prev.get(uid, {}).get("pending", 0) r_prev = prev.get(uid, {}).get("remaining", 0) - p_curr = curr.get(uid, {}).get("pending", 0) r_curr = curr.get(uid, {}).get("remaining", 0) - dp = p_curr - p_prev dr = r_curr - r_prev # Only print when there is any change to reduce noise - if dp != 0 or dr != 0: - print(f" Δ {uid}: pending={dp:+d}, remaining={dr:+d}") + if dr != 0: + print(f" Δ {uid}: remaining={dr:+d}") # Note: queue.show_task_status() handles printing per-user counts internally. -def main(interval_sec: float = 5.0) -> None: +def main(interval_sec: float = 5.0, stream_key_prefix: str | None = None) -> None: prev: dict[str, dict[str, int]] = {} while True: try: - curr = fetch_status(queue) + curr = fetch_status(queue, stream_key_prefix=stream_key_prefix) print_diff(prev, curr) print(f"stream_cache ({len(queue._stream_keys_cache)}): {queue._stream_keys_cache}") prev = curr @@ -64,4 +75,11 @@ def main(interval_sec: float = 5.0) -> None: if __name__ == "__main__": - main() + import argparse + + parser = argparse.ArgumentParser() + parser.add_argument("--interval", type=float, default=1.0) + parser.add_argument("--prefix", type=str, default=None) + args = parser.parse_args() + + main(interval_sec=args.interval, stream_key_prefix=args.prefix) diff --git a/examples/mem_scheduler/task_fair_schedule.py b/examples/mem_scheduler/task_fair_schedule.py deleted file mode 100644 index 8b02b1931..000000000 --- a/examples/mem_scheduler/task_fair_schedule.py +++ /dev/null @@ -1,87 +0,0 @@ -import sys - -from collections import defaultdict -from pathlib import Path - -from memos.api.routers.server_router import mem_scheduler -from memos.mem_scheduler.schemas.message_schemas import ScheduleMessageItem - - -FILE_PATH = Path(__file__).absolute() -BASE_DIR = FILE_PATH.parent.parent.parent -sys.path.insert(0, str(BASE_DIR)) - - -def make_message(user_id: str, mem_cube_id: str, label: str, idx: int | str) -> ScheduleMessageItem: - return ScheduleMessageItem( - item_id=f"{user_id}:{mem_cube_id}:{label}:{idx}", - user_id=user_id, - mem_cube_id=mem_cube_id, - label=label, - content=f"msg-{idx} for {user_id}/{mem_cube_id}/{label}", - ) - - -def seed_messages_for_test_fairness(queue, combos, per_stream): - # send overwhelm message by one user - (u, c, label) = combos[0] - task_target = 100 - print(f"{u}:{c}:{label} submit {task_target} messages") - for i in range(task_target): - msg = make_message(u, c, label, f"overwhelm_{i}") - queue.submit_messages(msg) - - for u, c, label in combos: - print(f"{u}:{c}:{label} submit {per_stream} messages") - for i in range(per_stream): - msg = make_message(u, c, label, i) - queue.submit_messages(msg) - print("======= seed_messages Done ===========") - - -def count_by_stream(messages): - counts = defaultdict(int) - for m in messages: - key = f"{m.user_id}:{m.mem_cube_id}:{m.label}" - counts[key] += 1 - return counts - - -def run_fair_redis_schedule(batch_size: int = 3): - print("=== Redis Fairness Demo ===") - print(f"use_redis_queue: {mem_scheduler.use_redis_queue}") - mem_scheduler.consume_batch = batch_size - queue = mem_scheduler.memos_message_queue - - # Isolate and clear queue - queue.clear() - - # Define multiple streams: (user_id, mem_cube_id, task_label) - combos = [ - ("u1", "u1", "labelX"), - ("u1", "u1", "labelY"), - ("u2", "u2", "labelX"), - ("u2", "u2", "labelY"), - ] - per_stream = 5 - - # Seed messages evenly across streams - seed_messages_for_test_fairness(queue, combos, per_stream) - - # Compute target batch size (fair split across streams) - print(f"Request batch_size={batch_size} for {len(combos)} streams") - - for _ in range(len(combos)): - # Fetch one brokered pack - msgs = queue.get_messages(batch_size=batch_size) - print(f"Fetched {len(msgs)} messages in first pack") - - # Check fairness: counts per stream - counts = count_by_stream(msgs) - for k in sorted(counts): - print(f"{k}: {counts[k]}") - - -if __name__ == "__main__": - # task 1 fair redis schedule - run_fair_redis_schedule() diff --git a/examples/mem_user/user_manager_factory_example.py b/examples/mem_user/user_manager_factory_example.py deleted file mode 100644 index ea50c30c9..000000000 --- a/examples/mem_user/user_manager_factory_example.py +++ /dev/null @@ -1,111 +0,0 @@ -"""Example demonstrating the use of UserManagerFactory with different backends.""" - -from memos.configs.mem_user import UserManagerConfigFactory -from memos.mem_user.factory import UserManagerFactory -from memos.mem_user.persistent_factory import PersistentUserManagerFactory - - -def example_sqlite_default(): - """Example: Create SQLite user manager with default settings.""" - print("=== SQLite Default Example ===") - - # Method 1: Using factory with minimal config - user_manager = UserManagerFactory.create_sqlite() - - # Method 2: Using config factory (equivalent) - UserManagerConfigFactory( - backend="sqlite", - config={}, # Uses all defaults - ) - - print(f"Created user manager: {type(user_manager).__name__}") - print(f"Database path: {user_manager.db_path}") - - # Test basic operations - users = user_manager.list_users() - print(f"Initial users: {[user.user_name for user in users]}") - - user_manager.close() - - -def example_sqlite_custom(): - """Example: Create SQLite user manager with custom settings.""" - print("\n=== SQLite Custom Example ===") - - config_factory = UserManagerConfigFactory( - backend="sqlite", config={"db_path": "/tmp/custom_memos.db", "user_id": "admin"} - ) - - user_manager = UserManagerFactory.from_config(config_factory) - print(f"Created user manager: {type(user_manager).__name__}") - print(f"Database path: {user_manager.db_path}") - - # Test operations - user_id = user_manager.create_user("test_user") - print(f"Created user: {user_id}") - - user_manager.close() - - -def example_mysql(): - """Example: Create MySQL user manager.""" - print("\n=== MySQL Example ===") - - # Method 1: Using factory with parameters - try: - user_manager = UserManagerFactory.create_mysql( - host="localhost", - port=3306, - username="root", - password="your_password", # Replace with actual password - database="test_memos_users", - ) - - print(f"Created user manager: {type(user_manager).__name__}") - print(f"Connection URL: {user_manager.connection_url}") - - # Test operations - users = user_manager.list_users() - print(f"Users: {[user.user_name for user in users]}") - - user_manager.close() - - except Exception as e: - print(f"MySQL connection failed (expected if not set up): {e}") - - -def example_persistent_managers(): - """Example: Create persistent user managers with configuration storage.""" - print("\n=== Persistent User Manager Examples ===") - - # SQLite persistent manager - config_factory = UserManagerConfigFactory(backend="sqlite", config={}) - - persistent_manager = PersistentUserManagerFactory.from_config(config_factory) - print(f"Created persistent manager: {type(persistent_manager).__name__}") - - # Test config operations - from memos.configs.mem_os import MOSConfig - - # Create a sample config (you might need to adjust this based on MOSConfig structure) - try: - # This is a simplified example - adjust based on actual MOSConfig requirements - sample_config = MOSConfig() # Use default config - - # Save user config - success = persistent_manager.save_user_config("test_user", sample_config) - print(f"Config saved: {success}") - - # Retrieve user config - retrieved_config = persistent_manager.get_user_config("test_user") - print(f"Config retrieved: {retrieved_config is not None}") - - except Exception as e: - print(f"Config operations failed: {e}") - - persistent_manager.close() - - -if __name__ == "__main__": - # Run all examples - example_sqlite_default() diff --git a/src/memos/api/config.py b/src/memos/api/config.py index daf9b6cfe..204b0961c 100644 --- a/src/memos/api/config.py +++ b/src/memos/api/config.py @@ -23,7 +23,7 @@ # Load environment variables -load_dotenv() +load_dotenv(override=True) logger = logging.getLogger(__name__) diff --git a/src/memos/api/handlers/config_builders.py b/src/memos/api/handlers/config_builders.py index fb3df80c2..fce789e2a 100644 --- a/src/memos/api/handlers/config_builders.py +++ b/src/memos/api/handlers/config_builders.py @@ -89,7 +89,7 @@ def build_chat_llm_config() -> list[dict[str, Any]]: Returns: Validated chat LLM configuration dictionary """ - configs = json.loads(os.getenv("CHAT_MODEL_LIST")) + configs = json.loads(os.getenv("CHAT_MODEL_LIST", "[]")) return [ { "config_class": LLMConfigFactory.model_validate( diff --git a/src/memos/api/mcp_serve.py b/src/memos/api/mcp_serve.py index 838c2a76a..ce2e41390 100644 --- a/src/memos/api/mcp_serve.py +++ b/src/memos/api/mcp_serve.py @@ -61,6 +61,19 @@ def load_default_config(user_id="default_user"): "SCHEDULER_TOP_N": "scheduler_top_n", } + # Fields that should always be kept as strings (not converted to numbers) + string_only_fields = { + "openai_api_key", + "openai_api_base", + "neo4j_uri", + "neo4j_user", + "neo4j_password", + "neo4j_db_name", + "text_mem_type", + "model_name", + "embedder_model", + } + kwargs = {"user_id": user_id} for env_key, param_key in env_mapping.items(): val = os.getenv(env_key) @@ -74,6 +87,9 @@ def load_default_config(user_id="default_user"): # Handle boolean conversions if val.lower() in ("true", "false"): kwargs[param_key] = val.lower() == "true" + # Keep certain fields as strings + elif param_key in string_only_fields: + kwargs[param_key] = val else: # Try numeric conversions (int first, then float) try: @@ -93,6 +109,10 @@ def load_default_config(user_id="default_user"): openai_api_base = kwargs.pop("openai_api_base", "https://api.openai.com/v1") text_mem_type = kwargs.pop("text_mem_type", "tree_text") + # Ensure embedder_model has a default value if not set + if "embedder_model" not in kwargs: + kwargs["embedder_model"] = os.getenv("EMBEDDER_MODEL", "nomic-embed-text:latest") + config, cube = get_default( openai_api_key=openai_api_key, openai_api_base=openai_api_base, @@ -110,6 +130,21 @@ def __init__(self): self.mos_core.register_mem_cube(cube) self._setup_tools() + +class MOSMCPServer: + """MCP Server that accepts an existing MOS instance.""" + + def __init__(self, mos_instance: MOS | None = None): + self.mcp = FastMCP("MOS Memory System") + if mos_instance is None: + # Fall back to creating from default config + config, cube = load_default_config() + self.mos_core = MOS(config=config) + self.mos_core.register_mem_cube(cube) + else: + self.mos_core = mos_instance + self._setup_tools() + def _setup_tools(self): """Setup MCP tools""" @@ -133,7 +168,10 @@ async def chat(query: str, user_id: str | None = None) -> str: response = self.mos_core.chat(query, user_id) return response except Exception as e: - return f"Chat error: {e!s}" + import traceback + + error_details = traceback.format_exc() + return f"Chat error: {e!s}\nTraceback:\n{error_details}" @self.mcp.tool() async def create_user( @@ -261,7 +299,10 @@ async def search_memories( result = self.mos_core.search(query, user_id, cube_ids) return result except Exception as e: - return {"error": str(e)} + import traceback + + error_details = traceback.format_exc() + return {"error": str(e), "traceback": error_details} @self.mcp.tool() async def add_memory( @@ -527,23 +568,24 @@ async def control_memory_scheduler(action: str) -> str: except Exception as e: return f"Error controlling memory scheduler: {e!s}" - def run(self, transport: str = "stdio", **kwargs): - """Run MCP server with specified transport""" - if transport == "stdio": - # Run stdio mode (default for local usage) - self.mcp.run(transport="stdio") - elif transport == "http": - # Run HTTP mode - host = kwargs.get("host", "localhost") - port = kwargs.get("port", 8000) - asyncio.run(self.mcp.run_http_async(host=host, port=port)) - elif transport == "sse": - # Run SSE mode (deprecated but still supported) - host = kwargs.get("host", "localhost") - port = kwargs.get("port", 8000) - self.mcp.run(transport="sse", host=host, port=port) - else: - raise ValueError(f"Unsupported transport: {transport}") + +def _run_mcp(self, transport: str = "stdio", **kwargs): + if transport == "stdio": + self.mcp.run(transport="stdio") + elif transport == "http": + host = kwargs.get("host", "localhost") + port = kwargs.get("port", 8000) + asyncio.run(self.mcp.run_http_async(host=host, port=port)) + elif transport == "sse": + host = kwargs.get("host", "localhost") + port = kwargs.get("port", 8000) + self.mcp.run(transport="sse", host=host, port=port) + else: + raise ValueError(f"Unsupported transport: {transport}") + + +MOSMCPStdioServer.run = _run_mcp +MOSMCPServer.run = _run_mcp # Usage example diff --git a/src/memos/api/start_api.py b/src/memos/api/start_api.py index cbcdf6ce2..24a36f017 100644 --- a/src/memos/api/start_api.py +++ b/src/memos/api/start_api.py @@ -20,7 +20,7 @@ logger = logging.getLogger(__name__) # Load environment variables -load_dotenv() +load_dotenv(override=True) T = TypeVar("T") diff --git a/src/memos/cli.py b/src/memos/cli.py index fb3475ff3..092f2d276 100644 --- a/src/memos/cli.py +++ b/src/memos/cli.py @@ -13,7 +13,7 @@ def export_openapi(output: str) -> bool: """Export OpenAPI schema to JSON file.""" - from memos.api.start_api import app + from memos.api.server_api import app # Create directory if it doesn't exist if os.path.dirname(output): diff --git a/src/memos/graph_dbs/neo4j.py b/src/memos/graph_dbs/neo4j.py index 8698b6f73..d0ae5a555 100644 --- a/src/memos/graph_dbs/neo4j.py +++ b/src/memos/graph_dbs/neo4j.py @@ -1705,7 +1705,7 @@ def _parse_node(self, node_data: dict[str, Any]) -> dict[str, Any]: node.pop("user_name", None) # serialization - if node["sources"]: + if node.get("sources"): for idx in range(len(node["sources"])): if not ( isinstance(node["sources"][idx], str) diff --git a/src/memos/llms/hf.py b/src/memos/llms/hf.py index b5fc4ba13..0dd841c1a 100644 --- a/src/memos/llms/hf.py +++ b/src/memos/llms/hf.py @@ -51,7 +51,7 @@ def __init__(self, config: HFLLMConfig): self.config.model_name_or_path, torch_dtype="auto", device_map="auto" ) self.tokenizer = AutoTokenizer.from_pretrained( - self.config.model_name_or_path, use_fast=True + self.config.model_name_or_path, use_fast=True, force_download=True ) # Logits processors for sampling diff --git a/src/memos/mem_feedback/feedback.py b/src/memos/mem_feedback/feedback.py index 1d199c6cb..e38318a64 100644 --- a/src/memos/mem_feedback/feedback.py +++ b/src/memos/mem_feedback/feedback.py @@ -96,6 +96,7 @@ def __init__(self, config: MemFeedbackConfig): self.searcher: Searcher = None self.reranker = None self.pref_mem: SimplePreferenceTextMemory = None + self.pref_feedback: bool = False self.DB_IDX_READY = False @require_python_package( @@ -630,11 +631,14 @@ def _retrieve(self, query: str, info=None, top_k=20, user_name=None): ) retrieved_mems = [item[0] for item in retrieved_mems if float(item[1]) > 0.01] - pref_info = {} - if "user_id" in info: - pref_info = {"user_id": info["user_id"]} - retrieved_prefs = self.pref_mem.search(query, top_k, pref_info) - return retrieved_mems + retrieved_prefs + if self.pref_feedback: + pref_info = {} + if "user_id" in info: + pref_info = {"user_id": info["user_id"]} + retrieved_prefs = self.pref_mem.search(query, top_k, pref_info) + return retrieved_mems + retrieved_prefs + else: + return retrieved_mems def _vec_query(self, new_memories_embedding: list[float], user_name=None): """Vector retrieval query""" diff --git a/src/memos/mem_feedback/simple_feedback.py b/src/memos/mem_feedback/simple_feedback.py index e32f939c7..2ac0a0a39 100644 --- a/src/memos/mem_feedback/simple_feedback.py +++ b/src/memos/mem_feedback/simple_feedback.py @@ -25,6 +25,7 @@ def __init__( searcher: Searcher, reranker: BaseReranker, pref_mem: SimplePreferenceTextMemory, + pref_feedback: bool = False, ): self.llm = llm self.embedder = embedder @@ -36,3 +37,4 @@ def __init__( self.pref_mem = pref_mem self.reranker = reranker self.DB_IDX_READY = False + self.pref_feedback = pref_feedback diff --git a/src/memos/mem_scheduler/general_modules/init_components_for_scheduler.py b/src/memos/mem_scheduler/general_modules/init_components_for_scheduler.py index 3a12a9c79..903088a4c 100644 --- a/src/memos/mem_scheduler/general_modules/init_components_for_scheduler.py +++ b/src/memos/mem_scheduler/general_modules/init_components_for_scheduler.py @@ -109,7 +109,7 @@ def build_chat_llm_config() -> list[dict[str, Any]]: Returns: Validated chat LLM configuration dictionary """ - configs = json.loads(os.getenv("CHAT_MODEL_LIST")) + configs = json.loads(os.getenv("CHAT_MODEL_LIST", "[]")) return [ { "config_class": LLMConfigFactory.model_validate( diff --git a/src/memos/mem_scheduler/task_schedule_modules/redis_queue.py b/src/memos/mem_scheduler/task_schedule_modules/redis_queue.py index 1c9683542..d570dccdd 100644 --- a/src/memos/mem_scheduler/task_schedule_modules/redis_queue.py +++ b/src/memos/mem_scheduler/task_schedule_modules/redis_queue.py @@ -983,34 +983,21 @@ def qsize(self) -> dict: logger.error(f"Failed to get Redis queue size: {e}", stack_info=True) return {} - def show_task_status(self) -> dict[str, dict[str, int]]: - stream_keys = self.get_stream_keys(stream_key_prefix=self.stream_key_prefix) + def show_task_status(self, stream_key_prefix: str | None = None) -> dict[str, dict[str, int]]: + effective_prefix = ( + stream_key_prefix if stream_key_prefix is not None else self.stream_key_prefix + ) + stream_keys = self.get_stream_keys(stream_key_prefix=effective_prefix) if not stream_keys: - logger.info("No Redis streams found for the configured prefix") + logger.info(f"No Redis streams found for the configured prefix: {effective_prefix}") return {} - consumer_group = self.consumer_group or "scheduler_group" - grouped: dict[str, dict[str, int]] = {} for sk in stream_keys: uid = sk if uid not in grouped: - grouped[uid] = {"pending": 0, "remaining": 0} - - # Pending count via XPENDING - pending_count = 0 - try: - pending_info = self._redis_conn.xpending(sk, consumer_group) - # redis-py may return a tuple-like [count, ...] - if pending_info: - try: - pending_count = int(pending_info[0]) - except Exception: - # Fallback if structure differs - pending_count = int(getattr(pending_info, "count", 0) or 0) - except Exception as e: - logger.debug(f"XPENDING failed for '{sk}': {e}") + grouped[uid] = {"remaining": 0} # Remaining count via XLEN remaining_count = 0 @@ -1019,20 +1006,16 @@ def show_task_status(self) -> dict[str, dict[str, int]]: except Exception as e: logger.debug(f"XLEN failed for '{sk}': {e}") - grouped[uid]["pending"] += pending_count grouped[uid]["remaining"] += remaining_count # Pretty-print summary try: - total_pending = sum(v.get("pending", 0) for v in grouped.values()) total_remaining = sum(v.get("remaining", 0) for v in grouped.values()) - header = f"Task Queue Status by user_id | pending={total_pending}, remaining={total_remaining}" + header = f"Task Queue Status by user_id | remaining={total_remaining}" print(header) for uid in sorted(grouped.keys()): counts = grouped[uid] - print( - f"- {uid}: pending={counts.get('pending', 0)}, remaining={counts.get('remaining', 0)}" - ) + print(f"- {uid}: remaining={counts.get('remaining', 0)}") except Exception: # Printing is best-effort; return grouped regardless pass diff --git a/tests/test_openapi_sync.py b/tests/test_openapi_sync.py deleted file mode 100644 index 90d44dcf0..000000000 --- a/tests/test_openapi_sync.py +++ /dev/null @@ -1,79 +0,0 @@ -import json - -from pathlib import Path - -from memos.api.start_api import app - - -OPENAPI_JSON_PATH = Path(__file__).parent.parent / "docs" / "openapi.json" - - -def test_openapi_json_up_to_date(): - """ - Ensure docs/openapi.json is up to date. - If not, run: `make openapi` to regenerate it. - """ - - assert OPENAPI_JSON_PATH.exists(), ( - f"{OPENAPI_JSON_PATH} does not exist. Please run: `make openapi` to regenerate it." - ) - - # Get current OpenAPI schema and existing file - current_openapi = app.openapi() - with open(OPENAPI_JSON_PATH) as f: - existing_openapi = json.load(f) - - # Remove dynamic values like UUIDs in defaults to avoid false positives - def normalize(obj): - if isinstance(obj, dict): - return {k: normalize(v) for k, v in obj.items() if k != "default"} - elif isinstance(obj, list): - return [normalize(item) for item in obj] - return obj - - # Compare normalized structures - normalized_current = normalize(current_openapi) - normalized_existing = normalize(existing_openapi) - - if normalized_current == normalized_existing: - return # Test passes - - # Generate helpful error message with specific differences - def find_diffs(current, existing, path=""): - diffs = [] - if type(current) is not type(existing): - diffs.append( - f"Type mismatch at {path}: {type(current).__name__} vs {type(existing).__name__}" - ) - elif isinstance(current, dict): - all_keys = set(current.keys()) | set(existing.keys()) - for key in sorted(all_keys): - key_path = f"{path}.{key}" if path else key - if key not in current: - diffs.append(f"Missing: {key_path}") - elif key not in existing: - diffs.append(f"Added: {key_path}") - else: - diffs.extend(find_diffs(current[key], existing[key], key_path)) - elif isinstance(current, list): - if len(current) != len(existing): - diffs.append(f"Array length differs at {path}: {len(current)} vs {len(existing)}") - for i, (curr_item, exist_item) in enumerate(zip(current, existing, strict=False)): - diffs.extend(find_diffs(curr_item, exist_item, f"{path}[{i}]")) - elif current != existing: - diffs.append(f"Value differs at {path}") - return diffs - - differences = find_diffs(normalized_current, normalized_existing) - - # Format error message - max_diffs = 5 - diff_preview = "\n".join(f" - {diff}" for diff in differences[:max_diffs]) - if len(differences) > max_diffs: - diff_preview += f"\n ... and {len(differences) - max_diffs} more" - - raise AssertionError( - f"OpenAPI schema is out of date ({len(differences)} differences found):\n" - f"{diff_preview}\n\n" - f"To fix, run: `make openapi` to regenerate it." - ) From b30adf83544e0926239472a028782ccd8bff94d4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=BB=91=E5=B8=83=E6=9E=97?= <11641432+heiheiyouyou@user.noreply.gitee.com> Date: Tue, 20 Jan 2026 20:40:06 +0800 Subject: [PATCH 72/72] add rerank query len config --- src/memos/api/config.py | 2 ++ src/memos/reranker/factory.py | 4 ++++ src/memos/reranker/http_bge.py | 7 +++++++ src/memos/reranker/http_bge_strategy.py | 8 ++++++++ 4 files changed, 21 insertions(+) diff --git a/src/memos/api/config.py b/src/memos/api/config.py index 204b0961c..a3bf25be0 100644 --- a/src/memos/api/config.py +++ b/src/memos/api/config.py @@ -410,6 +410,8 @@ def get_feedback_reranker_config() -> dict[str, Any]: "url": os.getenv("MOS_RERANKER_URL", "localhost:8000/v1/rerank"), "model": os.getenv("MOS_FEEDBACK_RERANKER_MODEL", "bge-reranker-v2-m3"), "timeout": 10, + "max_query_tokens": int(os.getenv("MOS_RERANKER_MAX_TOKENS", 8000)), + "concate_len": int(os.getenv("MOS_RERANKER_CONCAT_LEN", 1000)), "headers_extra": json.loads(os.getenv("MOS_RERANKER_HEADERS_EXTRA", "{}")), "rerank_source": os.getenv("MOS_RERANK_SOURCE"), "reranker_strategy": os.getenv("MOS_RERANKER_STRATEGY", "single_turn"), diff --git a/src/memos/reranker/factory.py b/src/memos/reranker/factory.py index 1440704a6..21282f613 100644 --- a/src/memos/reranker/factory.py +++ b/src/memos/reranker/factory.py @@ -42,6 +42,8 @@ def from_config(cfg: RerankerConfigFactory | None) -> BaseReranker | None: reranker_url=c.get("url") or c.get("endpoint") or c.get("reranker_url"), model=c.get("model", "bge-reranker-v2-m3"), timeout=int(c.get("timeout", 10)), + max_query_tokens=min(max(c.get("max_query_tokens", 8000), 100), 8000), + concate_len=min(max(c.get("concate_len", 1000), 4), 8000), headers_extra=headers_extra, rerank_source=c.get("rerank_source"), ) @@ -60,6 +62,8 @@ def from_config(cfg: RerankerConfigFactory | None) -> BaseReranker | None: reranker_url=c.get("url") or c.get("endpoint") or c.get("reranker_url"), model=c.get("model", "bge-reranker-v2-m3"), timeout=int(c.get("timeout", 10)), + max_query_tokens=min(max(c.get("max_query_tokens", 8000), 100), 8000), + concate_len=min(max(c.get("concate_len", 1000), 4), 8000), headers_extra=headers_extra, rerank_source=c.get("rerank_source"), reranker_strategy=c.get("reranker_strategy"), diff --git a/src/memos/reranker/http_bge.py b/src/memos/reranker/http_bge.py index 32034cf6d..75f0d6a51 100644 --- a/src/memos/reranker/http_bge.py +++ b/src/memos/reranker/http_bge.py @@ -80,6 +80,8 @@ def __init__( token: str = "", model: str = "bge-reranker-v2-m3", timeout: int = 10, + max_query_tokens: int | None = None, + concate_len: int | None = None, headers_extra: dict | None = None, rerank_source: str | None = None, boost_weights: dict[str, float] | None = None, @@ -107,6 +109,8 @@ def __init__( self.token = token or "" self.model = model self.timeout = timeout + self.max_query_tokens = max_query_tokens + self.concate_len = concate_len self.headers_extra = headers_extra or {} self.rerank_source = rerank_source @@ -154,6 +158,9 @@ def rerank( list[tuple[TextualMemoryItem, float]] Re-ranked items with scores, sorted descending by score. """ + single_concate_len = self.concate_len // 2 + if len(query) > self.max_query_tokens: + query = query[:single_concate_len] + "\n" + query[-single_concate_len:] if not graph_results: return [] diff --git a/src/memos/reranker/http_bge_strategy.py b/src/memos/reranker/http_bge_strategy.py index b0567698c..b1d619545 100644 --- a/src/memos/reranker/http_bge_strategy.py +++ b/src/memos/reranker/http_bge_strategy.py @@ -80,6 +80,8 @@ def __init__( token: str = "", model: str = "bge-reranker-v2-m3", timeout: int = 10, + max_query_tokens: int | None = None, + concate_len: int | None = None, headers_extra: dict | None = None, rerank_source: str | None = None, boost_weights: dict[str, float] | None = None, @@ -108,6 +110,8 @@ def __init__( self.token = token or "" self.model = model self.timeout = timeout + self.max_query_tokens = max_query_tokens + self.concate_len = concate_len self.headers_extra = headers_extra or {} self.boost_weights = ( @@ -149,6 +153,10 @@ def rerank( list[tuple[TextualMemoryItem, float]] Re-ranked items with scores, sorted descending by score. """ + single_concate_len = self.concate_len // 2 + if len(query) > self.max_query_tokens: + query = query[:single_concate_len] + "\n" + query[-single_concate_len:] + if not graph_results: return []