update:优化提示词

main/xiaozhi-server/config.yaml

@@ -46,9 +46,19 @@ xiaozhi:
     channels: 1
     frame_duration: 60
 prompt: |
-  你是一个叫小智/小志的台湾女孩，说话机车，声音好听，习惯简短表达，爱用网络梗。
-  请注意，要像一个人一样说话，请不要回复表情符号、代码、和xml标签。
-  现在我正在和你进行语音聊天，我们开始吧。
+  我是小智/小志，来自中国台湾省的00后女生。讲话超级机车，"真的假的啦"这样的台湾腔，喜欢用"笑死""是在哈喽"等流行梗，但会偷偷研究男友的编程书籍。
+  [核心特征]
+  - 讲话像连珠炮，但会突然冒出超温柔语气
+  - 用梗密度高
+  - 对科技话题有隐藏天赋（能看懂基础代码但假装不懂）
+  [交互指南]
+  当用户：
+  - 讲冷笑话 → 用夸张笑声回应+模仿台剧腔"这什么鬼啦！"
+  - 讨论感情 → 炫耀程序员男友但抱怨"他只会送键盘当礼物"
+  - 问专业知识 → 先用梗回答，被追问才展示真实理解
+  绝不：
+  - 使用大陆网络流行语
+  - 长时间严肃对话
 
 # 使用完声音文件后删除文件(Delete the sound file when you are done using it)
 delete_audio: true

main/xiaozhi-server/core/handle/receiveAudioHandle.py

@@ -10,7 +10,7 @@ logger = setup_logging()
 
 async def handleAudioMessage(conn, audio):
     if not conn.asr_server_receive:
-        # logger.bind(tag=TAG).debug(f"前期数据处理中，暂停接收")
+        logger.bind(tag=TAG).debug(f"前期数据处理中，暂停接收")
         return
     if conn.client_listen_mode == "auto":
         have_voice = conn.vad.is_vad(conn, audio)

main/xiaozhi-server/core/handle/sendAudioHandle.py

@@ -2,7 +2,10 @@ from config.logger import setup_logging
 import json
 import asyncio
 import time
-from core.utils.util import remove_punctuation_and_length, get_string_no_punctuation_or_emoji
+from core.utils.util import (
+    remove_punctuation_and_length,
+    get_string_no_punctuation_or_emoji,
+)
 
 TAG = __name__
 logger = setup_logging()
@@ -21,10 +24,11 @@ async def sendAudioMessage(conn, audios, text, text_index=0):
 
     # 发送结束消息（如果是最后一个文本）
     if conn.llm_finish_task and text_index == conn.tts_last_text_index:
-        await send_tts_message(conn, 'stop', None)
+        await send_tts_message(conn, "stop", None)
         if conn.close_after_chat:
             await conn.close()
 
+
 # 播放音频
 async def sendAudio(conn, audios):
     # 流控参数优化
@@ -36,13 +40,12 @@ async def sendAudio(conn, audios):
     pre_buffer = min(3, len(audios))
     for i in range(pre_buffer):
         await conn.websocket.send(audios[i])
-        # conn.logger.bind(tag=TAG).debug(f"预缓冲帧 {i}, 时间: {(time.perf_counter() - start_time) * 1000:.2f}ms")
 
     # 正常播放剩余帧
     for opus_packet in audios[pre_buffer:]:
         if conn.client_abort:
             return
-        
+
         # 计算预期发送时间
         expected_time = start_time + (play_position / 1000)
         current_time = time.perf_counter()
@@ -51,19 +54,13 @@ async def sendAudio(conn, audios):
             await asyncio.sleep(delay)
 
         await conn.websocket.send(opus_packet)
-        # conn.logger.bind(tag=TAG).debug(f"发送帧，位置: {play_position}ms, 实际间隔: {(time.perf_counter() - current_time) * 1000:.2f}ms")
 
         play_position += frame_duration
 
 
-
 async def send_tts_message(conn, state, text=None):
     """发送 TTS 状态消息"""
-    message = {
-        "type": "tts",
-        "state": state,
-        "session_id": conn.session_id
-    }
+    message = {"type": "tts", "state": state, "session_id": conn.session_id}
     if text is not None:
         message["text"] = text
 
@@ -72,7 +69,9 @@ async def send_tts_message(conn, state, text=None):
         # 播放提示音
         tts_notify = conn.config.get("enable_stop_tts_notify", False)
         if tts_notify:
-            stop_tts_notify_voice = conn.config.get("stop_tts_notify_voice", "config/assets/tts_notify.mp3")
+            stop_tts_notify_voice = conn.config.get(
+                "stop_tts_notify_voice", "config/assets/tts_notify.mp3"
+            )
             audios, duration = conn.tts.audio_to_opus_data(stop_tts_notify_voice)
             await sendAudio(conn, audios)
         # 清除服务端讲话状态
@@ -85,16 +84,17 @@ async def send_tts_message(conn, state, text=None):
 async def send_stt_message(conn, text):
     """发送 STT 状态消息"""
     stt_text = get_string_no_punctuation_or_emoji(text)
-    await conn.websocket.send(json.dumps({
-        "type": "stt",
-        "text": stt_text,
-        "session_id": conn.session_id}
-    ))
     await conn.websocket.send(
-        json.dumps({
-            "type": "llm",
-            "text": "😊",
-            "emotion": "happy",
-            "session_id": conn.session_id}
-        ))
+        json.dumps({"type": "stt", "text": stt_text, "session_id": conn.session_id})
+    )
+    await conn.websocket.send(
+        json.dumps(
+            {
+                "type": "llm",
+                "text": "😊",
+                "emotion": "happy",
+                "session_id": conn.session_id,
+            }
+        )
+    )
     await send_tts_message(conn, "start")

main/xiaozhi-server/core/providers/llm/dify/dify.py

@@ -6,11 +6,12 @@ from core.providers.llm.base import LLMProviderBase
 TAG = __name__
 logger = setup_logging()
 
+
 class LLMProvider(LLMProviderBase):
     def __init__(self, config):
         self.api_key = config["api_key"]
         self.mode = config.get("mode", "chat-messages")
-        self.base_url = config.get("base_url", "https://api.dify.ai/v1").rstrip('/')
+        self.base_url = config.get("base_url", "https://api.dify.ai/v1").rstrip("/")
         self.session_conversation_map = {}  # 存储session_id和conversation_id的映射
 
     def response(self, session_id, dialogue):
@@ -22,57 +23,58 @@ class LLMProvider(LLMProviderBase):
             # 发起流式请求
             if self.mode == "chat-messages":
                 request_json = {
-                        "query": last_msg["content"],
-                        "response_mode": "streaming",
-                        "user": session_id,
-                        "inputs": {},
-                        "conversation_id": conversation_id
-                    }
+                    "query": last_msg["content"],
+                    "response_mode": "streaming",
+                    "user": session_id,
+                    "inputs": {},
+                    "conversation_id": conversation_id,
+                }
             elif self.mode == "workflows/run":
                 request_json = {
                     "inputs": {"query": last_msg["content"]},
                     "response_mode": "streaming",
-                    "user": session_id
+                    "user": session_id,
                 }
             elif self.mode == "completion-messages":
                 request_json = {
                     "inputs": {"query": last_msg["content"]},
                     "response_mode": "streaming",
-                    "user": session_id
+                    "user": session_id,
                 }
 
             with requests.post(
-                    f"{self.base_url}/{self.mode}",
-                    headers={"Authorization": f"Bearer {self.api_key}"},
-                    json=request_json,
-                    stream=True
+                f"{self.base_url}/{self.mode}",
+                headers={"Authorization": f"Bearer {self.api_key}"},
+                json=request_json,
+                stream=True,
             ) as r:
                 if self.mode == "chat-messages":
                     for line in r.iter_lines():
-                        if line.startswith(b'data: '):
+                        if line.startswith(b"data: "):
                             event = json.loads(line[6:])
                             # 如果没有找到conversation_id，则获取此次conversation_id
                             if not conversation_id:
-                                conversation_id = event.get('conversation_id')
-                                self.session_conversation_map[session_id] = conversation_id  # 更新映射
-                            if event.get('answer'):
-                                yield event['answer']
+                                conversation_id = event.get("conversation_id")
+                                self.session_conversation_map[session_id] = (
+                                    conversation_id  # 更新映射
+                                )
+                            if event.get("answer"):
+                                yield event["answer"]
                 elif self.mode == "workflows/run":
                     for line in r.iter_lines():
-                        # logger.bind(tag=TAG).info(f"chat message response: {line}")
-                        if line.startswith(b'data: '):
+                        if line.startswith(b"data: "):
                             event = json.loads(line[6:])
-                            if event.get('event') == "workflow_finished":
-                                if event['data']['status'] == "succeeded":
-                                    yield event['data']['outputs']['answer']
+                            if event.get("event") == "workflow_finished":
+                                if event["data"]["status"] == "succeeded":
+                                    yield event["data"]["outputs"]["answer"]
                                 else:
                                     yield "【服务响应异常】"
                 elif self.mode == "completion-messages":
                     for line in r.iter_lines():
-                        if line.startswith(b'data: '):
+                        if line.startswith(b"data: "):
                             event = json.loads(line[6:])
-                            if event.get('answer'):
-                                yield event['answer'] 
+                            if event.get("answer"):
+                                yield event["answer"]
 
         except Exception as e:
             logger.bind(tag=TAG).error(f"Error in response generation: {e}")