#update: 添加mixed模型方案

main.py

@@ -226,14 +226,19 @@ async def streaming_chat(ws: WebSocket):
         agent.recorder.input_text = prompt
         logger.debug("开始调用大模型")
         llm_frames = await agent.chat(assistant, prompt)
+        
+        start_time = time.time()
+        is_first_response = True
+        
         async for llm_frame in llm_frames:
+            if is_first_response:
+                end_time = time.time()
+                logger.debug(f"第一帧返回耗时：{round(end_time-start_time,3)}s")
+                is_first_response = False
             resp_msgs = agent.llm_msg_process(llm_frame)
             for resp_msg in resp_msgs: 
                 llm_text += resp_msg
-                tts_start_time = time.time()
                 tts_audio = agent.synthetize(assistant, resp_msg)
-                tts_end_time = time.time()
-                logger.debug(f"TTS生成音频耗时：{tts_end_time-tts_start_time}s")
                 agent.tts_audio_process(tts_audio)
                 await ws.send_bytes(agent.encode(resp_msg, tts_audio))
                 logger.debug(f'websocket返回：{resp_msg}')