Update app.py

app.py

@@ -1466,21 +1466,19 @@ def docqa_run(state):
         return {"answer": result.output}
 
 def general_run(state):
-    result = general_agent.execute_task(general_task, {"query": state["query"]})
-    if isinstance(result, str):
-        output = result.lower()
-    else:
-        output = result.output.lower()
-    if any(x in output for x in ["i don't know", "no idea", "not sure", "can't answer"]):
-        result = search_agent.execute_task(search_task, {"query": state["query"]})
-        if isinstance(result, str):
-            return {"answer": result}
-        else:
-            return {"answer": result.output}
-    if isinstance(result, str):
-        return {"answer": result}
-    else:
-        return {"answer": result.output}
+    """改用直接 LLM 回答取代 General Agent"""
+    try:
+        prompt = f"""You are a helpful AI assistant. Please answer the following question:
+        {state["query"]}
+        
+        Provide a clear and informative answer."""
+        
+        response = llm_gpt4.invoke(prompt)
+        answer = response.content if hasattr(response, 'content') else str(response)
+        return {"answer": answer}
+    except Exception as e:
+        print(f"ERROR in general_run: {str(e)}")
+        return {"answer": "I apologize, but I'm having trouble processing your request."}
 
 def summariser_run(state):
     result = summarizer_agent.execute_task(summariser_task, {"query": state["query"]})
@@ -1509,13 +1507,6 @@ def build_langgraph_pipeline():
 
 from tempfile import mkdtemp
 
-
-from tempfile import mkdtemp
-import os
-
-from tempfile import mkdtemp
-import os
-
 def get_file_path_tab6(file):
     if isinstance(file, str):
         print("DEBUG: File is a string:", file)
@@ -1573,60 +1564,46 @@ def get_file_path_tab6(file):
 def langgraph_tab6_main(query: str, file=None):
     try:
         print(f"DEBUG: Starting processing with query: {query}")
-        print(f"DEBUG: File input type: {type(file)}")
+        
+        # 如果沒有文件，直接使用 general_run
+        if not file:
+            return general_run({"query": query})["answer"]
         
         # 處理文件列表
-        files = file if isinstance(file, list) else [file] if file else []
-        all_docs = []        # 用於建立檢索器
-        file_names = []      # 檔案名稱列表
-        docs_by_file = []    # 每份文件的完整內文
+        files = file if isinstance(file, list) else [file]
+        all_docs = []
+        file_names = []
+        docs_by_file = []
         
-        # 處理上傳的文件
         for f in files:
             try:
                 path = get_file_path_tab6(f)
                 if not path:
-                    print(f"WARNING: Could not process file {f}, skipping")
                     continue
-                    
-                print(f"DEBUG: Successfully got file path: {path}")
+                
                 file_names.append(os.path.basename(path))
                 
-                # 根據文件類型選擇適當的加載器
+                # 根據文件類型選擇加載器
                 if path.lower().endswith('.pdf'):
                     loader = PyPDFLoader(path)
                 elif path.lower().endswith('.docx'):
                     loader = UnstructuredWordDocumentLoader(path)
                 else:
                     loader = TextLoader(path)
-                    
-                print(f"DEBUG: Using loader: {type(loader)}")
                 
-                try:
-                    docs = loader.load()
-                    print(f"DEBUG: Successfully loaded document, got {len(docs)} pages/chunks")
-                    
-                    # 提取文件內容
-                    if docs and hasattr(docs[0], "page_content"):
-                        text = "\n".join([doc.page_content for doc in docs])
-                    else:
-                        text = "\n".join(docs)
-                    
+                docs = loader.load()
+                if docs:
+                    text = "\n".join(doc.page_content for doc in docs if hasattr(doc, 'page_content'))
                     docs_by_file.append(text)
                     all_docs.extend(docs)
-                    
-                except Exception as e:
-                    print(f"ERROR loading document: {str(e)}")
-                    continue
-                    
             except Exception as e:
                 print(f"ERROR processing file: {str(e)}")
                 continue
 
-        if not all_docs:
-            return "No valid documents were processed. Please check your file upload."
+        if not docs_by_file:
+            return general_run({"query": query})["answer"]
 
-        # 建立檢索器和問答鏈
+        # 建立檢索器
         try:
             chunks = RecursiveCharacterTextSplitter(
                 chunk_size=500,
@@ -1634,10 +1611,7 @@ def langgraph_tab6_main(query: str, file=None):
             ).split_documents(all_docs)
             
             db = FAISS.from_documents(chunks, embeddings)
-            retriever = db.as_retriever(
-                search_type="similarity",
-                search_kwargs={"k": 5}
-            )
+            retriever = db.as_retriever(search_kwargs={"k": 5})
             
             global session_retriever, session_qa_chain
             session_retriever = retriever
@@ -1655,16 +1629,12 @@ def langgraph_tab6_main(query: str, file=None):
 
         # 解析查詢意圖
         parsed = parse_query(query)
-        print(f"DEBUG: Parsed query: {parsed}")
-
-        # 檢測是否需要使用 AutoGen 多代理協作
+        
+        # 如果是複雜查詢（比較、關聯分析等），使用 execute_multi_agent
         if needs_multi_agent_processing(query, parsed, docs_by_file):
-            print("DEBUG: Using AutoGen for complex query")
-            return handle_complex_query_with_autogen(query, docs_by_file, file_names)
+            return execute_multi_agent(parsed, docs_by_file, file_names)
         
         # 使用 LangGraph 處理一般查詢
-        print("DEBUG: Using LangGraph for standard query")
-        graph = build_langgraph_pipeline()
         state = {
             "query": query,
             "file_names": file_names,
@@ -1672,12 +1642,26 @@ def langgraph_tab6_main(query: str, file=None):
             "retriever": retriever
         }
         
+        graph = build_langgraph_pipeline()
         result = graph.invoke(state)
-        return process_result(result, query)
+        
+        # 處理結果
+        if isinstance(result, dict):
+            if "answer" in result:
+                return result["answer"]
+            elif "summary" in result:
+                return result["summary"]
+            elif session_qa_chain:
+                try:
+                    return session_qa_chain.run(query)
+                except Exception as e:
+                    print(f"ERROR in QA chain: {str(e)}")
+        
+        return "I apologize, but I couldn't process your query properly."
         
     except Exception as e:
         print(f"ERROR in main function: {str(e)}")
-        return f"[Tab6 Error] {str(e)}"
+        return f"I apologize, but I encountered an error: {str(e)}"
 
 def needs_multi_agent_processing(query: str, parsed: dict, docs: list) -> bool:
     return any([
@@ -1694,88 +1678,17 @@ def needs_multi_agent_processing(query: str, parsed: dict, docs: list) -> bool:
         ])
     ])
 
-def handle_complex_query_with_autogen(query: str, docs: list, file_names: list) -> str:
-    """使用 AutoGen 處理複雜查詢"""
-    try:
-        # 準備文件上下文
-        context = "\n\n".join(
-            f"Document {name}:\n{doc[:2000]}..." 
-            for name, doc in zip(file_names, docs)
-        )
-
-        # 定義專門的代理人
-        qa_agent = AssistantAgent(
-            name="QA_Expert",
-            system_message="You are an expert at analyzing documents and answering questions.",
-            llm_config={"config_list": [{"model": "gpt-4", "api_key": openai_api_key}]}
-        )
-
-        summarizer_agent = AssistantAgent(
-            name="Summarizer",
-            system_message="You are an expert at summarizing documents and identifying key points.",
-            llm_config={"config_list": [{"model": "gpt-4", "api_key": openai_api_key}]}
-        )
-
-        comparison_agent = AssistantAgent(
-            name="Comparator",
-            system_message="You are an expert at comparing documents and finding relationships.",
-            llm_config={"config_list": [{"model": "gpt-4", "api_key": openai_api_key}]}
-        )
-
-        user_proxy = UserProxyAgent(
-            name="User",
-            system_message="A user seeking information from documents.",
-            human_input_mode="NEVER"
-        )
-
-        # 創建群組聊天
-        groupchat = GroupChat(
-            agents=[user_proxy, qa_agent, summarizer_agent, comparison_agent],
-            messages=[],
-            max_round=5
-        )
-
-        manager = GroupChatManager(groupchat=groupchat)
-
-        # 準備任務提示
-        task_prompt = f"""Analyze the following documents and answer the query:
-        
-        Query: {query}
-
-        Context:
-        {context}
-
-        Please provide:
-        1. Document analysis and relevant information
-        2. Direct answer to the query
-        3. Any important relationships or connections found
-        """
-
-        # 執行群組討論
-        user_proxy.initiate_chat(
-            manager,
-            message=task_prompt
-        )
-
-        # 獲取最終結果
-        final_answer = user_proxy.last_message()["content"]
-        
-        # 使用結果融合代理整理最終答案
-        fusion_prompt = f"""Based on the analysis provided, create a clear and concise response that:
-        1. Directly answers the user's query: "{query}"
-        2. Includes relevant supporting information
-        3. Maintains a natural, conversational tone
-        
-        Analysis to summarize:
-        {final_answer}
-        """
-        
-        final_response = llm_gpt4.invoke(fusion_prompt)
-        return final_response.content
-
-    except Exception as e:
-        print(f"ERROR in AutoGen processing: {str(e)}")
-        return f"Error during multi-agent processing: {str(e)}"
+def needs_multi_agent_processing(query: str, parsed: dict, docs: list) -> bool:
+    """判斷是否需要多代理處理"""
+    return any([
+        parsed.get("summarize_files"),
+        parsed.get("compare_files"),
+        parsed.get("find_relations"),
+        len(docs) > 1 and any(x in query.lower() for x in [
+            "compare", "between", "both", "relation",
+            "project", "similar", "different"
+        ])
+    ])
 
 def process_result(result: dict, query: str) -> str:
     """處理查詢結果"""