debug: 修复了多个bug

1.完成了端侧发送杂音识别
2.修复minimax返回帧解析错误bug
3.修复了返回过慢的bug

main.py

@@ -1,20 +1,26 @@
 from fastapi import FastAPI, Depends, WebSocket, HTTPException
 from fastapi.middleware.cors import CORSMiddleware
 from config import Config
-from app.concrete import Agent, AsrResultNoneError
+from app.concrete import Agent
 from app.model import Assistant, User, get_db
 from app.schemas import *
+from app.dependency import get_logger
+from app.exception import AsrResultNoneError, AbnormalLLMFrame, SideNoiseError
 import uvicorn
 import uuid
 import json
 
 # 公共函数 ------------------------------------------------------------------------------------------------------------------------------------------------------------------------
-def update_messages(messages, llm_text):
+def update_messages(messages, kid_text,llm_text):
     messages = json.loads(messages)
+    messages.append({"role":"user","content":kid_text})
     messages.append({"role":"assistant","content":llm_text})
     return json.dumps(messages,ensure_ascii=False)
 # --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
 
+# 引入logger对象
+logger = get_logger()
+
 # 创建FastAPI实例
 app = FastAPI()
 
@@ -89,6 +95,7 @@ async def update_assistant_system_prompt(id: str,request: update_assistant_syste
     assistant = db.query(Assistant).filter(Assistant.id == id).first()
     if assistant:
         assistant.system_prompt = request.system_prompt
+        assistant.messages = json.dumps([{"role":"system","content":assistant.system_prompt}],ensure_ascii=False)
         db.commit()
         return {"code":200,"msg":"success","data":{}}
     else:
@@ -163,11 +170,13 @@ async def update_user(id: str,request: update_user_request,db=Depends(get_db)):
 @app.websocket("/api/chat/streaming/temporary")
 async def streaming_chat(ws: WebSocket,db=Depends(get_db)):
     await ws.accept()
-    agent = Agent(asr_type=Config.ASR, llm_type=Config.LLM, tts_type=Config.TTS)
-    assistant = None
-    asr_results = []
-    llm_text = ""
+    logger.debug("WebSocket连接成功")
     try:
+        agent = Agent(asr_type=Config.ASR, llm_type=Config.LLM, tts_type=Config.TTS)
+        assistant = None
+        asr_results = []
+        llm_text = ""
+        logger.debug("开始进行ASR识别")
         while len(asr_results)==0:
             chunk = json.loads(await ws.receive_text())
             if assistant is None:
@@ -175,24 +184,36 @@ async def streaming_chat(ws: WebSocket,db=Depends(get_db)):
                 agent.init_recorder(assistant.user_id)
             chunk["audio"] = agent.user_audio_process(chunk["audio"], db)
             asr_results = await agent.stream_recognize(chunk, db)
+        kid_text = asr_results[0]['text'] #asr结果的[0]默认为孩子(主要用户)的asr结果
+        logger.debug(f"ASR识别成功，识别结果为：{kid_text}")
+        prompt = agent.prompt_process(asr_results, db)
+        agent.recorder.input_text = prompt
+        logger.debug("开始调用大模型")
+        llm_frames = await agent.chat(assistant, prompt, db)
+        async for llm_frame in llm_frames:
+            resp_msgs = agent.llm_msg_process(llm_frame, db)
+            for resp_msg in resp_msgs: 
+                llm_text += resp_msg
+                tts_audio = agent.synthetize(assistant, resp_msg, db)
+                agent.tts_audio_process(tts_audio, db)
+                await ws.send_bytes(agent.encode(resp_msg, tts_audio))
+                logger.debug(f'websocket返回：{resp_msg}')
+        logger.debug(f"大模型返回结束，返回结果为：{llm_text}")
+        await ws.send_text(json.dumps({"type": "close", "code": 200, "msg": ""}, ensure_ascii=False))
+        logger.debug("结束帧发送完毕")
+        assistant.messages = update_messages(assistant.messages, kid_text ,llm_text)
+        db.commit()
+        logger.debug("聊天更新成功")
+        agent.recorder.output_text = llm_text
+        agent.save()
+        logger.debug("音频保存成功")
     except AsrResultNoneError:
-        await ws.send_text(json.dumps({"type":"close","code":201,"msg":""}, ensure_ascii=False))
-        return
-    prompt = agent.prompt_process(asr_results, db)
-    agent.recorder.input_text = prompt
-    llm_frames = await agent.chat(assistant, prompt, db)
-    async for llm_frame in llm_frames:
-        resp_msgs = agent.llm_msg_process(llm_frame, db)
-        for resp_msg in resp_msgs: 
-            llm_text += resp_msg
-            tts_audio = agent.synthetize(assistant, resp_msg, db)
-            agent.tts_audio_process(tts_audio, db)
-            await ws.send_bytes(agent.encode(resp_msg, tts_audio))
-    await ws.send_text(json.dumps({"type": "close", "code": 200, "msg": ""}, ensure_ascii=False))
-    assistant.messages = update_messages(assistant.messages, llm_text)
-    db.commit()
-    agent.recorder.output_text = llm_text
-    agent.save()
+        await ws.send_text(json.dumps({"type":"close","code":201,"msg":"asr结果为空"}, ensure_ascii=False))
+    except AbnormalLLMFrame as e:
+        await ws.send_text(json.dumps({"type":"close","code":202,"msg":str(e)}, ensure_ascii=False))
+    except SideNoiseError as e:
+        await ws.send_text(json.dumps({"type":"close","code":203,"msg":str(e)}, ensure_ascii=False))
+    logger.debug("WebSocket连接断开")
     await ws.close()
 # --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------