Method handle_chat_stream

rag_system/api_server.py:304–499

Stream internal phases and final answer using SSE (text/event-stream).

(self)

            self.send_json_response({"error": f"Server error: {str(e)}"}, status_code=500)

    def handle_chat_stream(self):
        """Stream internal phases and final answer using SSE (text/event-stream)."""
        try:
            content_length = int(self.headers['Content-Length'])
            post_data = self.rfile.read(content_length)
            data = json.loads(post_data.decode('utf-8'))

            query = data.get('query')
            session_id = data.get('session_id')
            compose_flag = data.get('compose_sub_answers')
            decomp_flag = data.get('query_decompose')
            ai_rerank_flag = data.get('ai_rerank')
            ctx_expand_flag = data.get('context_expand')
            verify_flag = data.get('verify')
            
            # ✨ NEW RETRIEVAL PARAMETERS
            retrieval_k = data.get('retrieval_k', 20)
            context_window_size = data.get('context_window_size', 1)
            reranker_top_k = data.get('reranker_top_k', 10)
            search_type = data.get('search_type', 'hybrid')
            dense_weight = data.get('dense_weight', 0.7)

            # 🚩 NEW: Force RAG override from frontend
            force_rag = bool(data.get('force_rag', False))

            # 🌿 Provence sentence pruning
            provence_prune = data.get('provence_prune')
            provence_threshold = data.get('provence_threshold')

            # User-selected generation model
            requested_model = data.get('model')
            if isinstance(requested_model,str) and requested_model:
                RAG_AGENT.ollama_config['generation_model']=requested_model

            if not query:
                self.send_json_response({"error": "Query is required"}, status_code=400)
                return

            # 🔄 UPDATE SESSION TITLE: If this is the first message in the session, update the title
            if session_id:
                try:
                    # Check if this is the first message by calling the backend server
                    backend_url = f"http://localhost:8000/sessions/{session_id}"
                    session_resp = requests.get(backend_url)
                    if session_resp.status_code == 200:
                        session_data = session_resp.json()
                        session = session_data.get('session', {})
                        # If message_count is 0, this is the first message
                        if session.get('message_count', 0) == 0:
                            # Generate a title from the first message
                            title = generate_session_title(query)
                            # Update the session title via backend API
                            # We'll need to add this endpoint to the backend, for now let's make a direct database call
                            # This is a temporary solution until we add a proper API endpoint
                            db.update_session_title(session_id, title)
                            print(f"📝 Updated session title to: {title}")
                            
                            # 💾 STORE USER MESSAGE: Add the user message to the database