Update main.py

main.py

@@ -577,8 +577,8 @@ def followup_agent(query: FollowupQueryModel, background_tasks: BackgroundTasks,
     def process_response():
         full_response = ""
         for content in chat_with_llama_stream(limited_conversation, model=query.model_id):
-            full_response += content
             yield content
+            full_response += content
         
         logger.info(f"LLM RAW response for query: {query.query}: {full_response}")
         response_content, interact, tools = parse_followup_and_tools(full_response)
@@ -603,6 +603,7 @@ def followup_agent(query: FollowupQueryModel, background_tasks: BackgroundTasks,
                 yield "<report>"
                 for content in search_response():
                     yield content
+                    full_response += content
                 yield "</report>"
                 
         background_tasks.add_task(update_db, query.user_id, query.conversation_id, query.query, full_response)