[FEAT] streamlit error

requirements.txt

@@ -7,4 +7,4 @@ tiktoken
 faiss-cpu
 langchain==0.1.0
 python-dotenv
-chromadb==0.3.29
+chromadb==0.4.22

src/app.py

@@ -76,6 +76,7 @@ def search_page_sunwoo():
     uploaded_files = st.file_uploader("Upload a CSV file", type=["csv"], accept_multiple_files=True)
 
     search_sunwoo = Search_sunwoo()
+    retriever_tool = None
     pipeline_compression_retriever = None
 
     st.subheader("웹툰을 검색해보세요!")
@@ -94,7 +95,7 @@ def search_page_sunwoo():
             datas.append(df)
         if len(datas) > 0:
             # print("TEST!!!!!!!" + str(len(datas)))
-            pipeline_compression_retriever = search_sunwoo.make_retriever(datas)
+            retriever_tool, pipeline_compression_retriever = search_sunwoo.make_retriever(datas)
 
     chat_input_key = "search_chat_input_sunwoo"
     # 사용자 인풋 받기  
@@ -110,7 +111,7 @@ def search_page_sunwoo():
             message_placeholder = st.empty()
             full_response = ""
             # assistant_response = search.receive_chat(prompt)
-            assistant_response = search_sunwoo.run(prompt, pipeline_compression_retriever)
+            assistant_response = search_sunwoo.run(prompt, retriever_tool, pipeline_compression_retriever)
             print(assistant_response)
 
             message_placeholder.markdown(assistant_response)

src/model/search_sunwoo/search.py

@@ -50,7 +50,7 @@ class Response(BaseModel):
 class Search():
 
     def __init__(self):
-        self.Response = Response
+        pass
 
     def get_data_from_csv(self, file_path):
         """ Get data from csv file """
@@ -108,7 +108,8 @@ def get_embeddings(self, documents, cached_embedder, collection_name="webtoon"):
         vectorstore = Chroma.from_documents(
             documents, 
             cached_embedder,
-            collection_name=collection_name)
+            collection_name=collection_name
+        )
         return vectorstore
 
     def get_retriever(self, vectorstore):
@@ -118,23 +119,6 @@ def get_retriever(self, vectorstore):
         )
         return retriever
 
-    def get_bm25_retriever(self, documents):
-        retriver = BM25Retriever.from_documents(documents)
-        return retriver
-
-    def get_elastic_vector(self, cached_embedder, documents):
-        elasticsearch_url = "https://50da3596960c471fb7fa70548b0a71d1.us-central1.gcp.cloud.es.io:443"
-        elastic_vectorstore = ElasticsearchStore.from_documents(
-            documents,
-            cached_embedder,
-            es_url=elasticsearch_url,
-            es_api_key=ELASTIC_API_KEY,
-            index_name="webtoon",
-            strategy=ElasticsearchStore.ExactRetrievalStrategy()
-        )
-        elastic_vectorstore.client.indices.refresh(index="webtoon")
-        return elastic_vectorstore
-
     def get_pipeline_compression_retriever(self, retriever, embeddings):
         """ Create a pipeline of document transformers and a retriever """
         ## filters
@@ -177,7 +161,7 @@ def parse(self, output):
         else:
             return AgentActionMessageLog(
                 tool=name, tool_input=inputs, log="", message_log=[output]
-        )
+            )
 
     def get_agent(self, retriever_tool):
         system_message = """
@@ -209,9 +193,7 @@ def get_agent(self, retriever_tool):
             openai_api_key=OPENAI_API_KEY,
             max_tokens=2000
         )
-
-        llm_with_tools = llm.bind_functions([retriever_tool, self.Response])
-
+        llm_with_tools = llm.bind_functions([retriever_tool, Response])
         agent = (
             {
                 "title": lambda x: x["title"],
@@ -252,16 +234,16 @@ def make_retriever(self, datas):
         retriever = self.get_retriever(vectorstore)
 
         pipeline_compression_retriever = self.get_pipeline_compression_retriever(retriever, cached_embedder)
-        return pipeline_compression_retriever
-
-    def run(self, query, pipeline_compression_retriever):
         retriever_tool = self.get_retriever_tool(pipeline_compression_retriever)
+        return retriever_tool, pipeline_compression_retriever
+
+    def run(self, query, retriever_tool, pipeline_compression_retriever):
         result = self.get_all_relevant_documents(query, pipeline_compression_retriever)
 
         if len(result) == 0:
             return "검색 결과가 없습니다."
-        agent_executor = self.get_agent(retriever_tool)
 
+        agent_executor = self.get_agent(retriever_tool)
         response = agent_executor(
             {   
                 "title": result[0].metadata["title"],
@@ -281,7 +263,7 @@ def run(self, query, pipeline_compression_retriever):
 
     file_name = "info.csv"
     for title_id in title_id_list:
-        file_path = os.path.join("..", "streamlit/src/model/data/webtoon", title_id, file_name)
+        file_path = os.path.join("..", "streamlit/src/model/data/webtoon/info", title_id+".csv")
         absolute_file_path = os.path.abspath(file_path)
 
         df = pd.read_csv(absolute_file_path)
@@ -292,9 +274,9 @@ def run(self, query, pipeline_compression_retriever):
     ## 화산귀환, 신의탑, 전지적 독자 시점, 가비지타임, 선천적 얼간이들, 내가 키운 S급들, 대학원 탈출일지
     now = time.time()
     # pipeline_compression_retriever, vectorstore = search.make_retriever_from_url(namu_list)
-    pipeline_compression_retriever = search.make_retriever(datas)
+    retriever_tool, pipeline_compression_retriever = search.make_retriever(datas)
     print("Make Retriever: " + str(time.time()- now))
     now2 = time.time()
-    result = search.run(query, pipeline_compression_retriever)
+    result = search.run(query, retriever_tool, pipeline_compression_retriever)
     print("Find result: " + str(time.time()- now2))
     print(result)