Update main.py

main.py

@@ -572,6 +572,49 @@ async def followup_agent(query: FollowupQueryModel, background_tasks: Background
 
     return StreamingResponse(process_response(), media_type="text/event-stream")
 
+@app.post("/v3/followup-agent")
+async def followup_agent(query: FollowupQueryModel, background_tasks: BackgroundTasks, api_key: str = Depends(verify_api_key)):
+    """
+    Followup agent endpoint that provides helpful responses or generates clarifying questions based on user queries.
+    Requires API Key authentication via X-API-Key header.
+    """
+    logger.info(f"Received followup agent query: {query.query}")
+
+    if query.conversation_id not in conversations:
+        conversations[query.conversation_id] = [
+            {"role": "system", "content": FOLLOWUP_AGENT_PROMPT}
+        ]
+    
+    conversations[query.conversation_id].append({"role": "user", "content": query.query})
+    last_activity[query.conversation_id] = time.time()
+    
+    # Limit tokens in the conversation history
+    limited_conversation = conversations[query.conversation_id]
+
+    def process_response():
+        full_response = ""
+        for content in chat_with_llama_stream(limited_conversation, model=query.model_id):
+            full_response += content
+            yield content
+
+        logger.info(f"LLM RAW response for query: {query.query}: {full_response}")
+        response_content, interact = parse_followup_response(full_response)
+        
+        result = {
+            "clarification": interact
+        }
+        
+        yield "<json>" +"[[["+ json.dumps(result)+"]]]"+"</json>"
+        
+        # Add the assistant's response to the conversation history
+        conversations[query.conversation_id].append({"role": "assistant", "content": full_response})
+        
+        background_tasks.add_task(update_db, query.user_id, query.conversation_id, query.query, full_response)
+        logger.info(f"Completed followup agent response for query: {query.query}, send result: {result}")
+
+    return StreamingResponse(process_response(), media_type="text/event-stream")
+
+
 if __name__ == "__main__":
     import uvicorn
     logger.info("Starting the application")