feat: optimization v2.0.6

src/memos/api/handlers/chat_handler.py

@@ -130,6 +130,7 @@ def handle_chat_complete(self, chat_req: APIChatCompleteRequest) -> dict[str, An
                 include_preference=chat_req.include_preference,
                 pref_top_k=chat_req.pref_top_k,
                 filter=chat_req.filter,
+                relativity=chat_req.relativity,
             )
 
             search_response = self.search_handler.handle_search_memories(search_req)
@@ -269,6 +270,7 @@ def generate_chat_response() -> Generator[str, None, None]:
                         include_preference=chat_req.include_preference,
                         pref_top_k=chat_req.pref_top_k,
                         filter=chat_req.filter,
+                        relativity=chat_req.relativity,
                     )
 
                     search_response = self.search_handler.handle_search_memories(search_req)
@@ -811,6 +813,7 @@ def generate_chat_response() -> Generator[str, None, None]:
                             include_preference=chat_req.include_preference,
                             pref_top_k=chat_req.pref_top_k,
                             filter=chat_req.filter,
+                            relativity=chat_req.relativity,
                         )
 
                         search_response = self.search_handler.handle_search_memories(search_req)

src/memos/api/handlers/memory_handler.py

@@ -412,6 +412,12 @@ def handle_get_memories_dashboard(
     get_mem_req: GetMemoryDashboardRequest, naive_mem_cube: NaiveMemCube
 ) -> GetMemoryResponse:
     results: dict[str, Any] = {"text_mem": [], "pref_mem": [], "tool_mem": [], "skill_mem": []}
+    # for statistics
+    total_text_nodes, total_tool_nodes, total_skill_nodes, total_preference_nodes = 0, 0, 0, 0
+    total_tool_nodes = 0
+    total_skill_nodes = 0
+    total_preference_nodes = 0
+
     text_memory_type = ["WorkingMemory", "LongTermMemory", "UserMemory", "OuterMemory"]
     text_memories_info = naive_mem_cube.text_mem.get_all(
         user_name=get_mem_req.mem_cube_id,
@@ -421,7 +427,7 @@ def handle_get_memories_dashboard(
         filter=get_mem_req.filter,
         memory_type=text_memory_type,
     )
-    text_memories, _ = text_memories_info["nodes"], text_memories_info["total_nodes"]
+    text_memories, total_text_nodes = text_memories_info["nodes"], text_memories_info["total_nodes"]
 
     # Group text memories by cube_id from metadata.user_name
     text_mem_by_cube: dict[str, list] = {}
@@ -453,7 +459,7 @@ def handle_get_memories_dashboard(
             filter=get_mem_req.filter,
             memory_type=["ToolSchemaMemory", "ToolTrajectoryMemory"],
         )
-        tool_memories, _ = (
+        tool_memories, total_tool_nodes = (
             tool_memories_info["nodes"],
             tool_memories_info["total_nodes"],
         )
@@ -488,7 +494,7 @@ def handle_get_memories_dashboard(
             filter=get_mem_req.filter,
             memory_type=["SkillMemory"],
         )
-        skill_memories, _ = (
+        skill_memories, total_skill_nodes = (
             skill_memories_info["nodes"],
             skill_memories_info["total_nodes"],
         )
@@ -515,7 +521,6 @@ def handle_get_memories_dashboard(
         ]
 
     preferences: list[TextualMemoryItem] = []
-    total_preference_nodes = 0
 
     format_preferences = []
     if get_mem_req.include_preference and naive_mem_cube.pref_mem is not None:
@@ -578,4 +583,13 @@ def handle_get_memories_dashboard(
         "skill_mem": results.get("skill_mem", []),
     }
 
+    # statistics
+    statistics = {
+        "total_text_nodes": total_text_nodes,
+        "total_tool_nodes": total_tool_nodes,
+        "total_skill_nodes": total_skill_nodes,
+        "total_preference_nodes": total_preference_nodes,
+    }
+    filtered_results["statistics"] = statistics
+
     return GetMemoryResponse(message="Memories retrieved successfully", data=filtered_results)

src/memos/api/product_models.py

@@ -98,6 +98,15 @@ class ChatRequest(BaseRequest):
     add_message_on_answer: bool = Field(True, description="Add dialogs to memory after chat")
     manager_user_id: str | None = Field(None, description="Manager User ID")
     project_id: str | None = Field(None, description="Project ID")
+    relativity: float = Field(
+        0.0,
+        ge=0,
+        description=(
+            "Relevance threshold for recalled memories. "
+            "Only memories with metadata.relativity >= relativity will be returned. "
+            "Use 0 to disable threshold filtering. Default: 0.3."
+        ),
+    )
 
     # ==== Filter conditions ====
     filter: dict[str, Any] | None = Field(
@@ -775,6 +784,15 @@ class APIChatCompleteRequest(BaseRequest):
     add_message_on_answer: bool = Field(True, description="Add dialogs to memory after chat")
     manager_user_id: str | None = Field(None, description="Manager User ID")
     project_id: str | None = Field(None, description="Project ID")
+    relativity: float = Field(
+        0.0,
+        ge=0,
+        description=(
+            "Relevance threshold for recalled memories. "
+            "Only memories with metadata.relativity >= relativity will be returned. "
+            "Use 0 to disable threshold filtering. Default: 0.3."
+        ),
+    )
 
     # ==== Filter conditions ====
     filter: dict[str, Any] | None = Field(

src/memos/mem_reader/read_skill_memory/process_skill_memory.py

@@ -52,6 +52,9 @@ def _generate_content_by_llm(llm: BaseLLM, prompt_template: str, **kwargs) -> An
     try:
         prompt = prompt_template.format(**kwargs)
         response = llm.generate([{"role": "user", "content": prompt}])
+        if not response:
+            logger.warning("[PROCESS_SKILLS] LLM returned empty or invalid response")
+            return {} if "json" in prompt_template.lower() else ""
         if "json" in prompt_template.lower():
             response = response.replace("```json", "").replace("```", "").strip()
             return json.loads(response)
@@ -436,6 +439,9 @@ def _extract_skill_memory_by_llm(
             skills_llm = os.getenv("SKILLS_LLM", None)
             llm_kwargs = {"model_name_or_path": skills_llm} if skills_llm else {}
             response_text = llm.generate(prompt, **llm_kwargs)
+            if not response_text:
+                logger.warning("[PROCESS_SKILLS] LLM returned empty or invalid response")
+                continue
             # Clean up response (remove Markdown code blocks if present)
             logger.info(f"[Skill Memory]: response_text {response_text}")
             response_text = response_text.strip()
@@ -561,6 +567,9 @@ def _extract_skill_memory_by_llm_md(
             skills_llm = os.getenv("SKILLS_LLM", None)
             llm_kwargs = {"model_name_or_path": skills_llm} if skills_llm else {}
             response_text = llm.generate(prompt, **llm_kwargs)
+            if not response_text:
+                logger.warning("[PROCESS_SKILLS] LLM returned empty or invalid response")
+                continue
             # Clean up response (remove Markdown code blocks if present)
             logger.info(f"[Skill Memory]: response_text {response_text}")
             response_text = response_text.strip()
@@ -641,26 +650,22 @@ def _rewrite_query(task_type: str, messages: MessageList, llm: BaseLLM, rewrite_
     )
     prompt = [{"role": "user", "content": prompt_content}]
 
-    # Call LLM to rewrite the query with retry logic
-    for attempt in range(3):
-        try:
-            response_text = llm.generate(prompt)
-            # Clean up response (remove any markdown formatting if present)
-            response_text = response_text.strip()
-            logger.info(f"[PROCESS_SKILLS] Rewritten query for task '{task_type}': {response_text}")
-            return response_text
-        except Exception as e:
+    # Call LLM to rewrite the query
+    try:
+        response_text = llm.generate(prompt)
+        # Clean up response (remove any markdown formatting if present)
+        if response_text and isinstance(response_text, str):
+            return response_text.strip()
+        else:
             logger.warning(
-                f"[PROCESS_SKILLS] LLM query rewrite failed (attempt {attempt + 1}): {e}"
+                "[PROCESS_SKILLS] LLM returned empty or invalid response, returning first message content"
             )
-            if attempt == 2:
-                logger.warning(
-                    "[PROCESS_SKILLS] LLM query rewrite failed after 3 retries, returning first message content"
-                )
-                return messages[0]["content"] if messages else ""
-
-    # Fallback (should not reach here due to return in exception handling)
-    return messages[0]["content"] if messages else ""
+            return messages[0]["content"] if messages else ""
+    except Exception as e:
+        logger.warning(
+            f"[PROCESS_SKILLS] LLM query rewrite failed: {e}, returning first message content"
+        )
+        return messages[0]["content"] if messages else ""
 
 
 @require_python_package(