Update app.py

app.py

@@ -96,6 +96,8 @@ from langchain_core.runnables import RunnableLambda
 from langchain.chains import LLMChain
 from langchain.chains.combine_documents.stuff import StuffDocumentsChain
 from sentence_transformers import SentenceTransformer
+# === AutoGen for multi-intent collaboration ===
+from autogen import AssistantAgent, UserProxyAgent, GroupChat, GroupChatManager
 
 
 
@@ -1252,6 +1254,28 @@ INTENT_LABELS = {
     "General": ["who are you", "tell me something", "what can you do", "fun fact"],
 }
 
+# === AutoGen 多代理人協作邏輯 ===
+def autogen_multi_intent_agent(query: str, docs: list) -> str:
+    try:
+        context = "\n\n".join(d.page_content for d in docs[:10])
+        system_prompt = f"""You are a helpful assistant. Your task is to answer the following user question using two strategies:
+1. Use context-based question answering based on the document below.
+2. Also generate a short summary of the document, in case that helps interpret the question.
+
+Document Context:
+{context}
+"""
+        user_proxy = UserProxyAgent(name="User", is_termination_msg=lambda x: True, human_input_mode="NEVER")
+        qa_agent = AssistantAgent(name="QA_Agent", system_message="You are great at document-based QA.")
+        sum_agent = AssistantAgent(name="Summary_Agent", system_message="You are great at summarising text.")
+        group_chat = GroupChat(agents=[user_proxy, qa_agent, sum_agent], messages=[], max_round=3)
+        manager = GroupChatManager(groupchat=group_chat, llm_config={"config_list": [{"model": "gpt-4o", "api_key": openai_api_key}]})
+
+        user_proxy.initiate_chat(manager, message=query)
+        return user_proxy.last_message()["content"]
+    except Exception as e:
+        return f"[AutoGen Error] {e}"
+        
 def detect_intent_embedding(query, file_names=[]):
     query_emb = embedding_model.encode(query, normalize_embeddings=True)
     best_label = None
@@ -1470,6 +1494,10 @@ def langgraph_tab6_main(query: str, file=None):
         state = {"query": query, "file_names": file_names}
         if retriever is not None:
             state["retriever"] = retriever
+        # 判斷是否為模糊查詢（summarise + QA）
+        if any(kw in query.lower() for kw in ["same project", "same paper", "are they the same", "is this document", "summarise and answer", "is it related to"]):
+            autogen_response = autogen_multi_intent_agent(query, all_docs)
+            return autogen_response
         result = graph.invoke(state)
         if "answer" in result:
             return result["answer"]