update:流式优化暂不稳定，先回滚到早期代码

main/xiaozhi-server/config.yaml

@@ -59,6 +59,10 @@ log:
 delete_audio: true
 # 没有语音输入多久后断开连接(秒)，默认2分钟，即120秒
 close_connection_no_voice_time: 120
+# TTS请求超时时间(秒)
+tts_timeout: 10
+# 开启唤醒词加速
+enable_wakeup_words_response_cache: true
 # 开场是否回复唤醒词
 enable_greeting: true
 # 说完话是否开启提示音

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

@@ -1,13 +1,33 @@
+import time
 import json
+import random
 import asyncio
+from core.utils.dialogue import Message
+from core.utils.util import audio_to_data
+from core.handle.sendAudioHandle import sendAudioMessage, send_stt_message
+from core.utils.util import remove_punctuation_and_length, opus_datas_to_wav_bytes
+from core.providers.tts.dto.dto import ContentType, SentenceType
 from core.providers.tools.device_mcp import (
     MCPClient,
     send_mcp_initialize_message,
     send_mcp_tools_list_request,
 )
+from core.utils.wakeup_word import WakeupWordsConfig
 
 TAG = __name__
 
+WAKEUP_CONFIG = {
+    "refresh_time": 5,
+    "words": ["你好", "你好啊", "嘿，你好", "嗨"],
+}
+
+# 创建全局的唤醒词配置管理器
+wakeup_words_config = WakeupWordsConfig()
+
+# 用于防止并发调用wakeupWordsResponse的锁
+_wakeup_response_lock = asyncio.Lock()
+
+
 async def handleHelloMessage(conn, msg_json):
     """处理hello消息"""
     audio_params = msg_json.get("audio_params")
@@ -28,4 +48,94 @@ async def handleHelloMessage(conn, msg_json):
             # 发送mcp消息，获取tools列表
             asyncio.create_task(send_mcp_tools_list_request(conn))
 
-    await conn.websocket.send(json.dumps(conn.welcome_msg))
+    await conn.websocket.send(json.dumps(conn.welcome_msg))
+
+
+async def checkWakeupWords(conn, text):
+    enable_wakeup_words_response_cache = conn.config[
+        "enable_wakeup_words_response_cache"
+    ]
+
+    if not enable_wakeup_words_response_cache or not conn.tts:
+        return False
+
+    _, filtered_text = remove_punctuation_and_length(text)
+    if filtered_text not in conn.config.get("wakeup_words"):
+        return False
+
+    conn.just_woken_up = True
+    await send_stt_message(conn, text)
+
+    # 获取当前音色
+    voice = getattr(conn.tts, "voice", "default")
+    if not voice:
+        voice = "default"
+
+    # 获取唤醒词回复配置
+    response = wakeup_words_config.get_wakeup_response(voice)
+    if not response or not response.get("file_path"):
+        response = {
+            "voice": "default",
+            "file_path": "config/assets/wakeup_words.wav",
+            "time": 0,
+            "text": "哈啰啊，我是小智啦，声音好听的台湾女孩一枚，超开心认识你耶，最近在忙啥，别忘了给我来点有趣的料哦，我超爱听八卦的啦",
+        }
+
+    # 播放唤醒词回复
+    conn.client_abort = False
+    opus_packets, _ = audio_to_data(response.get("file_path"))
+
+    conn.logger.bind(tag=TAG).info(f"播放唤醒词回复: {response.get('text')}")
+    await sendAudioMessage(conn, SentenceType.FIRST, opus_packets, response.get("text"))
+    await sendAudioMessage(conn, SentenceType.LAST, [], None)
+
+    # 补充对话
+    conn.dialogue.put(Message(role="assistant", content=response.get("text")))
+
+    # 检查是否需要更新唤醒词回复
+    if time.time() - response.get("time", 0) > WAKEUP_CONFIG["refresh_time"]:
+        if not _wakeup_response_lock.locked():
+            asyncio.create_task(wakeupWordsResponse(conn))
+    return True
+
+
+async def wakeupWordsResponse(conn):
+    if not conn.tts or not conn.llm or not conn.llm.response_no_stream:
+        return
+
+    try:
+        # 尝试获取锁，如果获取不到就返回
+        if not await _wakeup_response_lock.acquire():
+            return
+
+        # 生成唤醒词回复
+        wakeup_word = random.choice(WAKEUP_CONFIG["words"])
+        question = (
+            "此刻用户正在和你说```"
+            + wakeup_word
+            + "```。\n请你根据以上用户的内容进行20-30字回复。要符合系统设置的角色情感和态度，不要像机器人一样说话。\n"
+            + "请勿对这条内容本身进行任何解释和回应，请勿返回表情符号，仅返回对用户的内容的回复。"
+        )
+
+        result = conn.llm.response_no_stream(conn.config["prompt"], question)
+        if not result or len(result) == 0:
+            return
+
+        # 生成TTS音频
+        tts_result = await asyncio.to_thread(conn.tts.to_tts, result)
+        if not tts_result:
+            return
+
+        # 获取当前音色
+        voice = getattr(conn.tts, "voice", "default")
+
+        wav_bytes = opus_datas_to_wav_bytes(tts_result, sample_rate=16000)
+        file_path = wakeup_words_config.generate_file_path(voice)
+        with open(file_path, "wb") as f:
+            f.write(wav_bytes)
+        # 更新配置
+        wakeup_words_config.update_wakeup_response(voice, file_path, result)
+    finally:
+        # 确保在任何情况下都释放锁
+        if _wakeup_response_lock.locked():
+            _wakeup_response_lock.release()

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

@@ -1,11 +1,12 @@
 import json
-import uuid
 import asyncio
-from core.utils.dialogue import Message
-from core.providers.tts.dto.dto import ContentType
-from plugins_func.register import Action, ActionResponse
+import uuid
 from core.handle.sendAudioHandle import send_stt_message
+from core.handle.helloHandle import checkWakeupWords
 from core.utils.util import remove_punctuation_and_length
+from core.providers.tts.dto.dto import ContentType
+from core.utils.dialogue import Message
+from plugins_func.register import Action, ActionResponse
 from core.providers.tts.dto.dto import TTSMessageDTO, SentenceType
 
 TAG = __name__
@@ -23,9 +24,12 @@ async def handle_user_intent(conn, text):
         pass
 
     # 检查是否有明确的退出命令
-    _, filtered_text = remove_punctuation_and_length(text)
+    filtered_text = remove_punctuation_and_length(text)[1]
     if await check_direct_exit(conn, filtered_text):
         return True
+    # 检查是否是唤醒词
+    if await checkWakeupWords(conn, filtered_text):
+        return True
 
     if conn.intent_type == "function_call":
         # 使用支持function calling的聊天方法,不再进行意图分析

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

@@ -1,21 +1,28 @@
-import time
-import json
-from core.handle.abortHandle import handleAbortMessage
+from core.handle.sendAudioHandle import send_stt_message
 from core.handle.intentHandler import handle_user_intent
 from core.utils.output_counter import check_device_output_limit
-from core.utils.util import play_audio_frames, play_audio_response
-from core.handle.sendAudioHandle import send_stt_message, SentenceType
+from core.handle.abortHandle import handleAbortMessage
+import time
+import asyncio
+import json
+from core.handle.sendAudioHandle import SentenceType
+from core.utils.util import audio_to_data
 
 TAG = __name__
 
 
 async def handleAudioMessage(conn, audio):
-    # 检查是否在唤醒处理锁定期内
-    if getattr(conn, 'wakeup_processing_lock', 0) > time.monotonic():
-        return
-    
     # 当前片段是否有人说话
     have_voice = conn.vad.is_vad(conn, audio)
+    # 如果设备刚刚被唤醒，短暂忽略VAD检测
+    if have_voice and hasattr(conn, "just_woken_up") and conn.just_woken_up:
+        have_voice = False
+        # 设置一个短暂延迟后恢复VAD检测
+        conn.asr_audio.clear()
+        if not hasattr(conn, "vad_resume_task") or conn.vad_resume_task.done():
+            conn.vad_resume_task = asyncio.create_task(resume_vad_detection(conn))
+        return
+
     if have_voice:
         if conn.client_is_speaking:
             await handleAbortMessage(conn)
@@ -24,11 +31,18 @@ async def handleAudioMessage(conn, audio):
     # 接收音频
     await conn.asr.receive_audio(conn, audio, have_voice)
 
+
+async def resume_vad_detection(conn):
+    # 等待2秒后恢复VAD检测
+    await asyncio.sleep(1)
+    conn.just_woken_up = False
+
+
 async def startToChat(conn, text):
     # 检查输入是否是JSON格式（包含说话人信息）
     speaker_name = None
     actual_text = text
-
+    
     try:
         # 尝试解析JSON格式的输入
         if text.strip().startswith('{') and text.strip().endswith('}'):
@@ -37,13 +51,13 @@ async def startToChat(conn, text):
                 speaker_name = data['speaker']
                 actual_text = data['content']
                 conn.logger.bind(tag=TAG).info(f"解析到说话人信息: {speaker_name}")
-
+                
                 # 直接使用JSON格式的文本，不解析
                 actual_text = text
     except (json.JSONDecodeError, KeyError):
         # 如果解析失败，继续使用原始文本
         pass
-
+    
     # 保存说话人信息到连接对象
     if speaker_name:
         conn.current_speaker = speaker_name
@@ -107,7 +121,8 @@ async def max_out_size(conn):
     text = "不好意思，我现在有点事情要忙，明天这个时候我们再聊，约好了哦！明天不见不散，拜拜！"
     await send_stt_message(conn, text)
     file_path = "config/assets/max_output_size.wav"
-    play_audio_response(conn, {"text": text, "file_path": file_path})
+    opus_packets, _ = audio_to_data(file_path)
+    conn.tts.tts_audio_queue.put((SentenceType.LAST, opus_packets, text))
     conn.close_after_chat = True
 
 
@@ -125,15 +140,16 @@ async def check_bind_device(conn):
 
         # 播放提示音
         music_path = "config/assets/bind_code.wav"
-        conn.tts.tts_audio_queue.put((SentenceType.FIRST, [], text))
-        play_audio_frames(conn, music_path)
+        opus_packets, _ = audio_to_data(music_path)
+        conn.tts.tts_audio_queue.put((SentenceType.FIRST, opus_packets, text))
 
         # 逐个播放数字
         for i in range(6):  # 确保只播放6位数字
             try:
                 digit = conn.bind_code[i]
                 num_path = f"config/assets/bind_code/{digit}.wav"
-                play_audio_frames(conn, num_path)
+                num_packets, _ = audio_to_data(num_path)
+                conn.tts.tts_audio_queue.put((SentenceType.MIDDLE, num_packets, None))
             except Exception as e:
                 conn.logger.bind(tag=TAG).error(f"播放数字音频失败: {e}")
                 continue
@@ -142,4 +158,5 @@ async def check_bind_device(conn):
         text = f"没有找到该设备的版本信息，请正确配置 OTA地址，然后重新编译固件。"
         await send_stt_message(conn, text)
         music_path = "config/assets/bind_not_found.wav"
-        play_audio_response(conn, {"text": text, "file_path": music_path})
+        opus_packets, _ = audio_to_data(music_path)
+        conn.tts.tts_audio_queue.put((SentenceType.LAST, opus_packets, text))

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

@@ -1,25 +1,25 @@
 import json
-import time
 import asyncio
-from core.utils import textUtils
+import time
 from core.providers.tts.dto.dto import SentenceType
+from core.utils import textUtils
 
 TAG = __name__
 
 
 async def sendAudioMessage(conn, sentenceType, audios, text):
+    # 发送句子开始消息
+    conn.logger.bind(tag=TAG).info(f"发送音频消息: {sentenceType}, {text}")
+
+    pre_buffer = False
     if conn.tts.tts_audio_first_sentence:
         conn.logger.bind(tag=TAG).info(f"发送第一段语音: {text}")
         conn.tts.tts_audio_first_sentence = False
-        await send_tts_message(conn, "start", None)
+        pre_buffer = True
 
-    if sentenceType == SentenceType.FIRST:
-        await send_tts_message(conn, "sentence_start", text)
+    await send_tts_message(conn, "sentence_start", text)
 
-    await sendAudio(conn, audios)
-    # 发送句子开始消息
-    if sentenceType is not SentenceType.MIDDLE:
-        conn.logger.bind(tag=TAG).info(f"发送音频消息: {sentenceType}, {text}")
+    await sendAudio(conn, audios, pre_buffer)
 
     # 发送结束消息（如果是最后一个文本）
     if conn.llm_finish_task and sentenceType == SentenceType.LAST:
@@ -30,59 +30,45 @@ async def sendAudioMessage(conn, sentenceType, audios, text):
 
 
 # 播放音频
-async def sendAudio(conn, audios, pre_buffer=False):
-    """
-    发送单个opus包，支持流控
-    Args:
-        conn: 连接对象
-        opus_packet: 单个opus数据包
-        pre_buffer: 快速发送音频
-    """
+async def sendAudio(conn, audios, pre_buffer=True):
     if audios is None or len(audios) == 0:
         return
+    # 流控参数优化
+    frame_duration = 60  # 帧时长（毫秒），匹配 Opus 编码
+    start_time = time.perf_counter()
+    play_position = 0
 
-    if isinstance(audios, bytes):
+    # 仅当第一句话时执行预缓冲
+    if pre_buffer:
+        pre_buffer_frames = min(3, len(audios))
+        for i in range(pre_buffer_frames):
+            await conn.websocket.send(audios[i])
+        remaining_audios = audios[pre_buffer_frames:]
+    else:
+        remaining_audios = audios
+
+    # 播放剩余音频帧
+    for opus_packet in remaining_audios:
         if conn.client_abort:
-            return
-
-        # 短音频直接发送（例如：提示音）
-        if pre_buffer:
-            await conn.websocket.send(audios)
-            return
+            break
 
+        # 重置没有声音的状态
         conn.last_activity_time = time.time() * 1000
 
-        # 获取或初始化流控状态
-        if not hasattr(conn, "audio_flow_control"):
-            conn.audio_flow_control = {
-                "last_send_time": 0,
-                "packet_count": 0,
-                "start_time": time.perf_counter(),
-            }
-
-        frame_duration=60
-        flow_control = conn.audio_flow_control
-        current_time = time.perf_counter()
         # 计算预期发送时间
-        expected_time = flow_control["start_time"] + (
-            flow_control["packet_count"] * frame_duration / 1000
-        )
+        expected_time = start_time + (play_position / 1000)
+        current_time = time.perf_counter()
         delay = expected_time - current_time
         if delay > 0:
             await asyncio.sleep(delay)
 
-        # 发送数据包
-        await conn.websocket.send(audios)
+        await conn.websocket.send(opus_packet)
 
-        # 更新流控状态
-        flow_control["packet_count"] += 1
-        flow_control["last_send_time"] = time.perf_counter()
+        play_position += frame_duration
 
 
 async def send_tts_message(conn, state, text=None):
     """发送 TTS 状态消息"""
-    if text is None and state == "sentence_start":
-        return
     message = {"type": "tts", "state": state, "session_id": conn.session_id}
     if text is not None:
         message["text"] = textUtils.check_emoji(text)
@@ -95,12 +81,8 @@ async def send_tts_message(conn, state, text=None):
             stop_tts_notify_voice = conn.config.get(
                 "stop_tts_notify_voice", "config/assets/tts_notify.mp3"
             )
-            conn.tts.audio_to_opus_data_stream(
-                stop_tts_notify_voice,
-                callback=lambda audio_data: asyncio.run_coroutine_threadsafe(
-                    sendAudio(conn, audio_data, True), conn.loop
-                ),
-            )
+            audios, _ = conn.tts.audio_to_opus_data(stop_tts_notify_voice)
+            await sendAudio(conn, audios)
         # 清除服务端讲话状态
         conn.clearSpeakStatus()
 
@@ -109,17 +91,18 @@ async def send_tts_message(conn, state, text=None):
 
 
 async def send_stt_message(conn, text):
-    """发送 STT 状态消息"""
     end_prompt_str = conn.config.get("end_prompt", {}).get("prompt")
     if end_prompt_str and end_prompt_str == text:
         await send_tts_message(conn, "start")
         return
 
+    """发送 STT 状态消息"""
+    
     # 解析JSON格式，提取实际的用户说话内容
     display_text = text
     try:
         # 尝试解析JSON格式
-        if text.strip().startswith("{") and text.strip().endswith("}"):
+        if text.strip().startswith('{') and text.strip().endswith('}'):
             parsed_data = json.loads(text)
             if isinstance(parsed_data, dict) and "content" in parsed_data:
                 # 如果是包含说话人信息的JSON格式，只显示content部分

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

@@ -1,12 +1,14 @@
 import json
-import asyncio
-from core.utils.util import filter_sensitive_info
+import time
 from core.handle.abortHandle import handleAbortMessage
 from core.handle.helloHandle import handleHelloMessage
-from core.handle.receiveAudioHandle import handleAudioMessage
 from core.providers.tools.device_mcp import handle_mcp_message
+from core.utils.util import remove_punctuation_and_length, filter_sensitive_info
+from core.handle.receiveAudioHandle import startToChat, handleAudioMessage
 from core.handle.sendAudioHandle import send_stt_message, send_tts_message
 from core.providers.tools.device_iot import handleIotDescriptors, handleIotStatus
+from core.handle.reportHandle import enqueue_asr_report
+import asyncio
 
 TAG = __name__
 
@@ -44,9 +46,14 @@ async def handleTextMessage(conn, message):
                 conn.client_have_voice = False
                 conn.asr_audio.clear()
                 if "text" in msg_json:
-                    original_text = msg_json["text"]  # 保留设备上传的文本
+                    conn.last_activity_time = time.time() * 1000
+                    original_text = msg_json["text"]  # 保留原始文本
+                    filtered_len, filtered_text = remove_punctuation_and_length(
+                        original_text
+                    )
+
                     # 识别是否是唤醒词
-                    is_wakeup_words = original_text in conn.config.get("wakeup_words")
+                    is_wakeup_words = filtered_text in conn.config.get("wakeup_words")
                     # 是否开启唤醒词回复
                     enable_greeting = conn.config.get("enable_greeting", True)
 
@@ -55,10 +62,16 @@ async def handleTextMessage(conn, message):
                         await send_stt_message(conn, original_text)
                         await send_tts_message(conn, "stop", None)
                         conn.client_is_speaking = False
+                    elif is_wakeup_words:
+                        conn.just_woken_up = True
+                        # 上报纯文字数据（复用ASR上报功能，但不提供音频数据）
+                        enqueue_asr_report(conn, "嘿，你好呀", [])
+                        await startToChat(conn, "嘿，你好呀")
                     else:
-                        # 检测到唤醒词，开始等待后续进行声纹识别
-                        conn.wakeup_mode = True
-                        conn.logger.bind(tag=TAG).info(f"检测到唤醒词~")
+                        # 上报纯文字数据（复用ASR上报功能，但不提供音频数据）
+                        enqueue_asr_report(conn, original_text, [])
+                        # 否则需要LLM对文字内容进行答复
+                        await startToChat(conn, original_text)
         elif msg_json["type"] == "iot":
             conn.logger.bind(tag=TAG).info(f"收到iot消息：{message}")
             if "descriptors" in msg_json:

main/xiaozhi-server/core/providers/asr/base.py

@@ -1,18 +1,18 @@
 import os
-import io
 import wave
 import uuid
-import json
-import time
 import queue
 import asyncio
 import traceback
 import threading
 import opuslib_next
+import json
+import io
+import time
 import concurrent.futures
 from abc import ABC, abstractmethod
 from config.logger import setup_logging
-from typing import Optional, Tuple, List
+from typing import Optional, Tuple, List, Dict, Any
 from core.handle.receiveAudioHandle import startToChat
 from core.handle.reportHandle import enqueue_asr_report
 from core.utils.util import remove_punctuation_and_length
@@ -53,10 +53,6 @@ class ASRProviderBase(ABC):
 
     # 接收音频
     async def receive_audio(self, conn, audio, audio_have_voice):
-        # 检查是否在唤醒处理锁定期内
-        if getattr(conn, 'wakeup_processing_lock', 0) > time.monotonic():
-            return
-            
         if conn.client_listen_mode == "auto" or conn.client_listen_mode == "realtime":
             have_voice = audio_have_voice
         else:
@@ -66,14 +62,6 @@ class ASRProviderBase(ABC):
         if not have_voice and not conn.client_have_voice:
             conn.asr_audio = conn.asr_audio[-10:]
             return
-        
-        # 检查是否处于唤醒模式
-        if getattr(conn, 'wakeup_mode', False) and len(conn.asr_audio) >= 10:
-            asr_audio_task = conn.asr_audio.copy()
-            conn.reset_vad_states()
-            conn.asr_audio.clear()
-                
-            await self.handle_voice_stop(conn, asr_audio_task)
 
         if conn.client_voice_stop:
             asr_audio_task = conn.asr_audio.copy()
@@ -99,13 +87,32 @@ class ASRProviderBase(ABC):
             
             # 预先准备WAV数据
             wav_data = None
+            # 使用连接的声纹识别提供者
             if conn.voiceprint_provider and combined_pcm_data:
                 wav_data = self._pcm_to_wav(combined_pcm_data)
             
-            # 检查是否处于唤醒模式
-            wakeup_mode = getattr(conn, 'wakeup_mode', False)
             
-            # 声纹识别任务（公共逻辑）
+            # 定义ASR任务
+            def run_asr():
+                start_time = time.monotonic()
+                try:
+                    loop = asyncio.new_event_loop()
+                    asyncio.set_event_loop(loop)
+                    try:
+                        result = loop.run_until_complete(
+                            self.speech_to_text(asr_audio_task, conn.session_id, conn.audio_format)
+                        )
+                        end_time = time.monotonic()
+                        logger.bind(tag=TAG).info(f"ASR耗时: {end_time - start_time:.3f}s")
+                        return result
+                    finally:
+                        loop.close()
+                except Exception as e:
+                    end_time = time.monotonic()
+                    logger.bind(tag=TAG).error(f"ASR失败: {e}")
+                    return ("", None)
+            
+            # 定义声纹识别任务
             def run_voiceprint():
                 if not wav_data:
                     return None
@@ -124,83 +131,50 @@ class ASRProviderBase(ABC):
                     logger.bind(tag=TAG).error(f"声纹识别失败: {e}")
                     return None
             
-            if wakeup_mode and conn.voiceprint_provider and wav_data:
-                conn.wakeup_mode = False
-                # 设置处理锁，防止后续音频片段重复处理
-                conn.wakeup_processing_lock = time.monotonic() + 3  # 3秒锁定期
+            # 使用线程池执行器并行运行
+            parallel_start_time = time.monotonic()
+            
+            with concurrent.futures.ThreadPoolExecutor(max_workers=2) as thread_executor:
+                asr_future = thread_executor.submit(run_asr)
                 
-                # 唤醒模式：只执行声纹识别
-                with concurrent.futures.ThreadPoolExecutor(max_workers=1) as thread_executor:
+                if conn.voiceprint_provider and wav_data:
                     voiceprint_future = thread_executor.submit(run_voiceprint)
+                    
+                    # 等待两个线程都完成
+                    asr_result = asr_future.result(timeout=15)
                     voiceprint_result = voiceprint_future.result(timeout=15)
-
-                speaker_name = voiceprint_result
+                    
+                    results = {"asr": asr_result, "voiceprint": voiceprint_result}
+                else:
+                    asr_result = asr_future.result(timeout=15)
+                    results = {"asr": asr_result, "voiceprint": None}
+            
+            
+            # 处理结果
+            raw_text, file_path = results.get("asr", ("", None))
+            speaker_name = results.get("voiceprint", None)
+            
+            # 记录识别结果
+            if raw_text:
+                logger.bind(tag=TAG).info(f"识别文本: {raw_text}")
+            if speaker_name:
                 logger.bind(tag=TAG).info(f"识别说话人: {speaker_name}")
-
-                fixed_text = "嘿，你好啊"
-                enhanced_text = self._build_enhanced_text(fixed_text, speaker_name)
+            
+            # 性能监控
+            total_time = time.monotonic() - total_start_time
+            logger.bind(tag=TAG).info(f"总处理耗时: {total_time:.3f}s")
+            
+            # 检查文本长度
+            text_len, _ = remove_punctuation_and_length(raw_text)
+            self.stop_ws_connection()
+            
+            if text_len > 0:
+                # 构建包含说话人信息的JSON字符串
+                enhanced_text = self._build_enhanced_text(raw_text, speaker_name)
                 
-                # 性能监控
-                total_time = time.monotonic() - total_start_time
-                logger.bind(tag=TAG).info(f"唤醒模式总处理耗时: {total_time:.3f}s")
-
+                # 使用自定义模块进行上报
                 await startToChat(conn, enhanced_text)
                 enqueue_asr_report(conn, enhanced_text, asr_audio_task)
-            else:
-                # 正常模式：执行声纹识别和ASR
-                def run_asr():
-                    start_time = time.monotonic()
-                    try:
-                        loop = asyncio.new_event_loop()
-                        asyncio.set_event_loop(loop)
-                        try:
-                            result = loop.run_until_complete(
-                                self.speech_to_text(asr_audio_task, conn.session_id, conn.audio_format)
-                            )
-                            end_time = time.monotonic()
-                            logger.bind(tag=TAG).info(f"ASR耗时: {end_time - start_time:.3f}s")
-                            return result
-                        finally:
-                            loop.close()
-                    except Exception as e:
-                        end_time = time.monotonic()
-                        logger.bind(tag=TAG).error(f"ASR失败: {e}")
-                        return ("", None)
-                
-                # 使用线程池执行器并行运行
-                with concurrent.futures.ThreadPoolExecutor(max_workers=2) as thread_executor:
-                    asr_future = thread_executor.submit(run_asr)
-                    
-                    if conn.voiceprint_provider and wav_data:
-                        voiceprint_future = thread_executor.submit(run_voiceprint)
-                        asr_result = asr_future.result(timeout=15)
-                        voiceprint_result = voiceprint_future.result(timeout=15)
-                        results = {"asr": asr_result, "voiceprint": voiceprint_result}
-                    else:
-                        asr_result = asr_future.result(timeout=15)
-                        results = {"asr": asr_result, "voiceprint": None}
-                
-                # 处理结果
-                raw_text, _ = results.get("asr", ("", None))
-                speaker_name = results.get("voiceprint", None)
-                
-                if raw_text:
-                    logger.bind(tag=TAG).info(f"识别文本: {raw_text}")
-                if speaker_name:
-                    logger.bind(tag=TAG).info(f"识别说话人: {speaker_name}")
-                
-                # 性能监控
-                total_time = time.monotonic() - total_start_time
-                logger.bind(tag=TAG).info(f"总处理耗时: {total_time:.3f}s")
-                
-                # 检查文本长度
-                text_len, _ = remove_punctuation_and_length(raw_text)
-                self.stop_ws_connection()
-                
-                if text_len > 0:
-                    enhanced_text = self._build_enhanced_text(raw_text, speaker_name)
-                    await startToChat(conn, enhanced_text)
-                    enqueue_asr_report(conn, enhanced_text, asr_audio_task)
                 
         except Exception as e:
             logger.bind(tag=TAG).error(f"处理语音停止失败: {e}")

main/xiaozhi-server/core/providers/tts/aliyun_stream.py

@@ -268,7 +268,11 @@ class TTSProvider(TTSProviderBase):
                     )
                     if message.content_file and os.path.exists(message.content_file):
                         # 先处理文件音频数据
-                        self._process_audio_file_stream(message.content_file, callback=lambda audio_data: self.handle_audio_file(audio_data, message.content_detail))
+                        file_audio = self._process_audio_file(message.content_file)
+                        self.before_stop_play_files.append(
+                            (file_audio, message.content_detail)
+                        )
+
                 if message.sentence_type == SentenceType.LAST:
                     try:
                         logger.bind(tag=TAG).info("开始结束TTS会话...")
@@ -418,6 +422,9 @@ class TTSProvider(TTSProviderBase):
 
     async def _start_monitor_tts_response(self):
         """监听TTS响应"""
+        opus_datas_cache = []
+        is_first_sentence = True
+        first_sentence_segment_count = 0  # 添加计数器
         try:
             session_finished = False  # 标记会话是否正常结束
             while not self.conn.stop_event.is_set():
@@ -438,16 +445,28 @@ class TTSProvider(TTSProviderBase):
                                 self.tts_audio_queue.put(
                                     (SentenceType.FIRST, [], None)
                                 )
+                            elif event_name == "SentenceBegin":
+                                opus_datas_cache = []
                             elif event_name == "SentenceEnd":
-                                # 发送缓存的数据
-                                if self.conn.tts_MessageText:
-                                    logger.bind(tag=TAG).info(
-                                        f"句子语音生成成功： {self.conn.tts_MessageText}"
-                                    )
-                                    self.tts_audio_queue.put(
-                                        (SentenceType.FIRST, [], self.conn.tts_MessageText)
-                                    )
-                                    self.conn.tts_MessageText = None
+                                if (
+                                    not is_first_sentence
+                                    or first_sentence_segment_count > 10
+                                ):
+                                    # 发送缓存的数据
+                                    if self.conn.tts_MessageText:
+                                        logger.bind(tag=TAG).info(
+                                            f"句子语音生成成功： {self.conn.tts_MessageText}"
+                                        )
+                                        self.tts_audio_queue.put(
+                                            (SentenceType.MIDDLE, opus_datas_cache, self.conn.tts_MessageText)
+                                        )
+                                        self.conn.tts_MessageText = None
+                                    else:
+                                        self.tts_audio_queue.put(
+                                            (SentenceType.MIDDLE, opus_datas_cache, None)
+                                        )
+                                # 第一句话结束后，将标志设置为False
+                                is_first_sentence = False
                             elif event_name == "SynthesisCompleted":
                                 logger.bind(tag=TAG).debug(f"会话结束～～")
                                 self._process_before_stop_play_files()
@@ -458,7 +477,22 @@ class TTSProvider(TTSProviderBase):
                     # 二进制消息（音频数据）
                     elif isinstance(msg, (bytes, bytearray)):
                         logger.bind(tag=TAG).debug(f"推送数据到队列里面～～")
-                        self.opus_encoder.encode_pcm_to_opus_stream(msg, False, self.handle_opus)
+                        opus_datas = self.opus_encoder.encode_pcm_to_opus(msg, False)
+                        logger.bind(tag=TAG).debug(
+                            f"推送数据到队列里面帧数～～{len(opus_datas)}"
+                        )
+                        if is_first_sentence:
+                            first_sentence_segment_count += 1
+                            if first_sentence_segment_count <= 6:
+                                self.tts_audio_queue.put(
+                                    (SentenceType.MIDDLE, opus_datas, None)
+                                )
+                            else:
+                                opus_datas_cache.extend(opus_datas)
+                        else:
+                            # 后续句子缓存
+                            opus_datas_cache.extend(opus_datas)
+
                 except websockets.ConnectionClosed:
                     logger.bind(tag=TAG).warning("WebSocket连接已关闭")
                     break
@@ -478,3 +512,142 @@ class TTSProvider(TTSProviderBase):
         finally:
             self._monitor_task = None
 
+    def to_tts(self, text: str) -> list:
+        """非流式TTS处理，用于测试及保存音频文件的场景"""
+        try:
+            # 创建新的事件循环
+            loop = asyncio.new_event_loop()
+            asyncio.set_event_loop(loop)
+
+            # 生成会话ID
+            session_id = uuid.uuid4().hex
+            # 存储音频数据
+            audio_data = []
+
+            async def _generate_audio():
+                # 刷新Token（如果需要）
+                if self._is_token_expired():
+                    self._refresh_token()
+
+                # 建立WebSocket连接
+                ws = await websockets.connect(
+                    self.ws_url,
+                    additional_headers={"X-NLS-Token": self.token},
+                    ping_interval=30,
+                    ping_timeout=10,
+                    close_timeout=10,
+                )
+                try:
+                    # 发送StartSynthesis请求
+                    start_message_id = str(uuid.uuid4().hex)
+                    start_request = {
+                        "header": {
+                            "message_id": start_message_id,
+                            "task_id": session_id,
+                            "namespace": "FlowingSpeechSynthesizer",
+                            "name": "StartSynthesis",
+                            "appkey": self.appkey,
+                        },
+                        "payload": {
+                            "voice": self.voice,
+                            "format": self.format,
+                            "sample_rate": self.sample_rate,
+                            "volume": self.volume,
+                            "speech_rate": self.speech_rate,
+                            "pitch_rate": self.pitch_rate,
+                            "enable_subtitle": True,
+                        },
+                    }
+                    await ws.send(json.dumps(start_request))
+
+                    # 等待SynthesisStarted响应
+                    synthesis_started = False
+                    while not synthesis_started:
+                        msg = await ws.recv()
+                        if isinstance(msg, str):
+                            data = json.loads(msg)
+                            header = data.get("header", {})
+                            if header.get("name") == "SynthesisStarted":
+                                synthesis_started = True
+                                logger.bind(tag=TAG).debug("TTS合成已启动")
+                            elif header.get("name") == "TaskFailed":
+                                error_info = data.get("payload", {}).get(
+                                    "error_info", {}
+                                )
+                                error_code = error_info.get("error_code")
+                                error_message = error_info.get(
+                                    "error_message", "未知错误"
+                                )
+                                raise Exception(
+                                    f"启动合成失败: {error_code} - {error_message}"
+                                )
+
+                    # 发送文本合成请求
+                    filtered_text = MarkdownCleaner.clean_markdown(text)
+                    run_message_id = str(uuid.uuid4().hex)
+                    run_request = {
+                        "header": {
+                            "message_id": run_message_id,
+                            "task_id": session_id,
+                            "namespace": "FlowingSpeechSynthesizer",
+                            "name": "RunSynthesis",
+                            "appkey": self.appkey,
+                        },
+                        "payload": {"text": filtered_text},
+                    }
+                    await ws.send(json.dumps(run_request))
+
+                    # 发送停止合成请求
+                    stop_message_id = str(uuid.uuid4().hex)
+                    stop_request = {
+                        "header": {
+                            "message_id": stop_message_id,
+                            "task_id": session_id,
+                            "namespace": "FlowingSpeechSynthesizer",
+                            "name": "StopSynthesis",
+                            "appkey": self.appkey,
+                        }
+                    }
+                    await ws.send(json.dumps(stop_request))
+
+                    # 接收音频数据
+                    synthesis_completed = False
+                    while not synthesis_completed:
+                        msg = await ws.recv()
+                        if isinstance(msg, (bytes, bytearray)):
+                            # 编码为Opus并收集
+                            opus_frames = self.opus_encoder.encode_pcm_to_opus(
+                                msg, False
+                            )
+                            audio_data.extend(opus_frames)
+                        elif isinstance(msg, str):
+                            data = json.loads(msg)
+                            header = data.get("header", {})
+                            event_name = header.get("name")
+                            if event_name == "SynthesisCompleted":
+                                synthesis_completed = True
+                                logger.bind(tag=TAG).debug("TTS合成完成")
+                            elif event_name == "TaskFailed":
+                                error_info = data.get("payload", {}).get(
+                                    "error_info", {}
+                                )
+                                error_code = error_info.get("error_code")
+                                error_message = error_info.get(
+                                    "error_message", "未知错误"
+                                )
+                                raise Exception(
+                                    f"合成失败: {error_code} - {error_message}"
+                                )
+                finally:
+                    try:
+                        await ws.close()
+                    except:
+                        pass
+
+            loop.run_until_complete(_generate_audio())
+            loop.close()
+
+            return audio_data
+        except Exception as e:
+            logger.bind(tag=TAG).error(f"生成音频数据失败: {str(e)}")
+            return []

main/xiaozhi-server/core/providers/tts/base.py

@@ -4,14 +4,12 @@ import queue
 import uuid
 import asyncio
 import threading
-from typing import Callable, Any
 from core.utils import p3
-import time
 from datetime import datetime
 from core.utils import textUtils
 from abc import ABC, abstractmethod
 from config.logger import setup_logging
-from core.utils.util import audio_bytes_to_data_stream, audio_to_data_stream
+from core.utils.util import audio_to_data, audio_bytes_to_data
 from core.utils.tts import MarkdownCleaner
 from core.utils.output_counter import add_device_output
 from core.handle.reportHandle import enqueue_tts_report
@@ -33,6 +31,7 @@ class TTSProviderBase(ABC):
     def __init__(self, config, delete_audio_file):
         self.interface_type = InterfaceType.NON_STREAM
         self.conn = None
+        self.tts_timeout = 10
         self.delete_audio_file = delete_audio_file
         self.audio_file_type = "wav"
         self.output_file = config.get("output_dir", "tmp/")
@@ -51,9 +50,11 @@ class TTSProviderBase(ABC):
             "；",
             ";",
             "：",
+            "~",
         )
         self.first_sentence_punctuations = (
             "，",
+            "～",
             "~",
             "、",
             ",",
@@ -76,14 +77,7 @@ class TTSProviderBase(ABC):
             f"tts-{datetime.now().date()}@{uuid.uuid4().hex}{extension}",
         )
 
-    def handle_opus(self, opus_data: bytes):
-        logger.bind(tag=TAG).debug(f"推送数据到队列里面帧数～～ {len(opus_data)}")
-        self.tts_audio_queue.put((SentenceType.MIDDLE, opus_data, None))
-
-    def handle_audio_file(self, file_audio: bytes, text):
-        self.before_stop_play_files.append((file_audio, text))
-
-    def to_tts_stream(self, text, opus_handler: Callable[[bytes], None] = None) -> None:
+    def to_tts(self, text):
         text = MarkdownCleaner.clean_markdown(text)
         max_repeat_time = 5
         if self.delete_audio_file:
@@ -92,14 +86,10 @@ class TTSProviderBase(ABC):
                 try:
                     audio_bytes = asyncio.run(self.text_to_speak(text, None))
                     if audio_bytes:
-                        self.tts_audio_queue.put((SentenceType.FIRST, None, text))
-                        audio_bytes_to_data_stream(
-                            audio_bytes,
-                            file_type=self.audio_file_type,
-                            is_opus=True,
-                            callback=opus_handler,
+                        audio_datas, _ = audio_bytes_to_data(
+                            audio_bytes, file_type=self.audio_file_type, is_opus=True
                         )
-                        break
+                        return audio_datas
                     else:
                         max_repeat_time -= 1
                 except Exception as e:
@@ -139,8 +129,8 @@ class TTSProviderBase(ABC):
                     logger.bind(tag=TAG).error(
                         f"语音生成失败: {text}，请检查网络或服务是否正常"
                     )
-                    self.tts_audio_queue.put((SentenceType.FIRST, None, text))
-                self._process_audio_file_stream(tmp_file, callback=opus_handler)
+
+                return tmp_file
             except Exception as e:
                 logger.bind(tag=TAG).error(f"Failed to generate TTS file: {e}")
                 return None
@@ -149,17 +139,13 @@ class TTSProviderBase(ABC):
     async def text_to_speak(self, text, output_file):
         pass
 
-    def audio_to_pcm_data_stream(
-        self, audio_file_path, callback: Callable[[Any], Any] = None
-    ):
+    def audio_to_pcm_data(self, audio_file_path):
         """音频文件转换为PCM编码"""
-        return audio_to_data_stream(audio_file_path, is_opus=False, callback=callback)
+        return audio_to_data(audio_file_path, is_opus=False)
 
-    def audio_to_opus_data_stream(
-        self, audio_file_path, callback: Callable[[Any], Any] = None
-    ):
+    def audio_to_opus_data(self, audio_file_path):
         """音频文件转换为Opus编码"""
-        return audio_to_data_stream(audio_file_path, is_opus=True, callback=callback)
+        return audio_to_data(audio_file_path, is_opus=True)
 
     def tts_one_sentence(
         self,
@@ -191,6 +177,7 @@ class TTSProviderBase(ABC):
 
     async def open_audio_channels(self, conn):
         self.conn = conn
+        self.tts_timeout = conn.config.get("tts_timeout", 10)
         # tts 消化线程
         self.tts_priority_thread = threading.Thread(
             target=self.tts_text_priority_thread, daemon=True
@@ -225,16 +212,30 @@ class TTSProviderBase(ABC):
                     self.tts_text_buff.append(message.content_detail)
                     segment_text = self._get_segment_text()
                     if segment_text:
-                        self.to_tts_stream(segment_text, opus_handler=self.handle_opus)
+                        if self.delete_audio_file:
+                            audio_datas = self.to_tts(segment_text)
+                            if audio_datas:
+                                self.tts_audio_queue.put(
+                                    (message.sentence_type, audio_datas, segment_text)
+                                )
+                        else:
+                            tts_file = self.to_tts(segment_text)
+                            if tts_file:
+                                audio_datas = self._process_audio_file(tts_file)
+                                self.tts_audio_queue.put(
+                                    (message.sentence_type, audio_datas, segment_text)
+                                )
                 elif ContentType.FILE == message.content_type:
-                    self._process_remaining_text_stream(opus_handler=self.handle_opus)
+                    self._process_remaining_text()
                     tts_file = message.content_file
                     if tts_file and os.path.exists(tts_file):
-                        self._process_audio_file_stream(
-                            tts_file, callback=self.handle_opus
+                        audio_datas = self._process_audio_file(tts_file)
+                        self.tts_audio_queue.put(
+                            (message.sentence_type, audio_datas, message.content_detail)
                         )
+
                 if message.sentence_type == SentenceType.LAST:
-                    self._process_remaining_text_stream(opus_handler=self.handle_opus)
+                    self._process_remaining_text()
                     self.tts_audio_queue.put(
                         (message.sentence_type, [], message.content_detail)
                     )
@@ -248,59 +249,29 @@ class TTSProviderBase(ABC):
                 continue
 
     def _audio_play_priority_thread(self):
-        # 需要上报的文本和音频列表
-        enqueue_text = None
-        enqueue_audio = None
         while not self.conn.stop_event.is_set():
             text = None
             try:
                 try:
                     sentence_type, audio_datas, text = self.tts_audio_queue.get(
-                        timeout=0.1
+                        timeout=1
                     )
                 except queue.Empty:
                     if self.conn.stop_event.is_set():
                         break
                     continue
-
-                if self.conn.client_abort:
-                    logger.bind(tag=TAG).debug("收到打断信号，跳过当前音频数据")
-                    enqueue_text, enqueue_audio = None, []
-                    continue
-
-                # 收到下一个文本开始或会话结束时进行上报
-                if sentence_type is not SentenceType.MIDDLE:
-                    # 重置音频流控状态（新句子开始或者结束）
-                    if hasattr(self.conn, 'audio_flow_control'):
-                        self.conn.audio_flow_control = {
-                            'last_send_time': 0,
-                            'packet_count': 0,
-                            'start_time': time.perf_counter()
-                        }
-                    
-                    # 上报TTS数据
-                    if enqueue_text is not None and enqueue_audio is not None:
-                        enqueue_tts_report(self.conn, enqueue_text, enqueue_audio)
-                    enqueue_audio = []
-                    enqueue_text = text
-
-                # 收集上报音频数据
-                if isinstance(audio_datas, bytes) and enqueue_audio is not None:
-                    enqueue_audio.append(audio_datas)
-
-                # 发送音频
                 future = asyncio.run_coroutine_threadsafe(
                     sendAudioMessage(self.conn, sentence_type, audio_datas, text),
                     self.conn.loop,
                 )
                 future.result()
-
-                # 记录输出和报告
                 if self.conn.max_output_size > 0 and text:
                     add_device_output(self.conn.headers.get("device-id"), len(text))
-
+                enqueue_tts_report(self.conn, text, audio_datas)
             except Exception as e:
-                logger.bind(tag=TAG).error(f"audio_play_priority_thread: {text} {e}")
+                logger.bind(tag=TAG).error(
+                    f"audio_play_priority priority_thread: {text} {e}"
+                )
 
     async def start_session(self, session_id):
         pass
@@ -352,21 +323,22 @@ class TTSProviderBase(ABC):
         else:
             return None
 
-    def _process_audio_file_stream(
-        self, tts_file, callback: Callable[[Any], Any]
-    ) -> None:
+    def _process_audio_file(self, tts_file):
         """处理音频文件并转换为指定格式
 
         Args:
             tts_file: 音频文件路径
-            callback: 文件处理函数
+            content_detail: 内容详情
+
+        Returns:
+            tuple: (sentence_type, audio_datas, content_detail)
         """
         if tts_file.endswith(".p3"):
-            p3.decode_opus_from_file_stream(tts_file, callback=callback)
+            audio_datas, _ = p3.decode_opus_from_file(tts_file)
         elif self.conn.audio_format == "pcm":
-            self.audio_to_pcm_data_stream(tts_file, callback=callback)
+            audio_datas, _ = self.audio_to_pcm_data(tts_file)
         else:
-            self.audio_to_opus_data_stream(tts_file, callback=callback)
+            audio_datas, _ = self.audio_to_opus_data(tts_file)
 
         if (
             self.delete_audio_file
@@ -375,6 +347,7 @@ class TTSProviderBase(ABC):
             and tts_file.startswith(self.output_file)
         ):
             os.remove(tts_file)
+        return audio_datas
 
     def _process_before_stop_play_files(self):
         for audio_datas, text in self.before_stop_play_files:
@@ -382,9 +355,7 @@ class TTSProviderBase(ABC):
         self.before_stop_play_files.clear()
         self.tts_audio_queue.put((SentenceType.LAST, [], None))
 
-    def _process_remaining_text_stream(
-        self, opus_handler: Callable[[bytes], None] = None
-    ):
+    def _process_remaining_text(self):
         """处理剩余的文本并生成语音
 
         Returns:
@@ -395,7 +366,18 @@ class TTSProviderBase(ABC):
         if remaining_text:
             segment_text = textUtils.get_string_no_punctuation_or_emoji(remaining_text)
             if segment_text:
-                self.to_tts_stream(segment_text, opus_handler=opus_handler)
+                if self.delete_audio_file:
+                    audio_datas = self.to_tts(segment_text)
+                    if audio_datas:
+                        self.tts_audio_queue.put(
+                            (SentenceType.MIDDLE, audio_datas, segment_text)
+                        )
+                else:
+                    tts_file = self.to_tts(segment_text)
+                    audio_datas = self._process_audio_file(tts_file)
+                    self.tts_audio_queue.put(
+                        (SentenceType.MIDDLE, audio_datas, segment_text)
+                    )
                 self.processed_chars += len(full_text)
                 return True
         return False

main/xiaozhi-server/core/providers/tts/huoshan_double_stream.py

@@ -4,7 +4,6 @@ import json
 import queue
 import asyncio
 import traceback
-from typing import Callable, Any
 import websockets
 from core.utils.tts import MarkdownCleaner
 from config.logger import setup_logging
@@ -267,7 +266,11 @@ class TTSProvider(TTSProviderBase):
                     )
                     if message.content_file and os.path.exists(message.content_file):
                         # 先处理文件音频数据
-                        self._process_audio_file_stream(message.content_file, callback=lambda audio_data: self.handle_audio_file(audio_data, message.content_detail))
+                        file_audio = self._process_audio_file(message.content_file)
+                        self.before_stop_play_files.append(
+                            (file_audio, message.content_detail)
+                        )
+
                 if message.sentence_type == SentenceType.LAST:
                     try:
                         logger.bind(tag=TAG).info("开始结束TTS会话...")
@@ -425,6 +428,9 @@ class TTSProvider(TTSProviderBase):
 
     async def _start_monitor_tts_response(self):
         """监听TTS响应"""
+        opus_datas_cache = []
+        is_first_sentence = True
+        first_sentence_segment_count = 0  # 添加计数器
         try:
             session_finished = False  # 标记会话是否正常结束
             while not self.conn.stop_event.is_set():
@@ -445,14 +451,37 @@ class TTSProvider(TTSProviderBase):
                         self.tts_audio_queue.put(
                             (SentenceType.FIRST, [], self.tts_text)
                         )
+                        opus_datas_cache = []
+                        first_sentence_segment_count = 0  # 重置计数器
                     elif (
                         res.optional.event == EVENT_TTSResponse
                         and res.header.message_type == AUDIO_ONLY_RESPONSE
                     ):
                         logger.bind(tag=TAG).debug(f"推送数据到队列里面～～")
-                        self.wav_to_opus_data_audio_raw_stream(res.payload, callback=self.handle_opus)
+                        opus_datas = self.wav_to_opus_data_audio_raw(res.payload)
+                        logger.bind(tag=TAG).debug(
+                            f"推送数据到队列里面帧数～～{len(opus_datas)}"
+                        )
+                        if is_first_sentence:
+                            first_sentence_segment_count += 1
+                            if first_sentence_segment_count <= 6:
+                                self.tts_audio_queue.put(
+                                    (SentenceType.MIDDLE, opus_datas, None)
+                                )
+                            else:
+                                opus_datas_cache.extend(opus_datas)
+                        else:
+                            # 后续句子缓存
+                            opus_datas_cache.extend(opus_datas)
                     elif res.optional.event == EVENT_TTSSentenceEnd:
                         logger.bind(tag=TAG).info(f"句子语音生成成功：{self.tts_text}")
+                        if not is_first_sentence or first_sentence_segment_count > 10:
+                            # 发送缓存的数据
+                            self.tts_audio_queue.put(
+                                (SentenceType.MIDDLE, opus_datas_cache, None)
+                            )
+                        # 第一句话结束后，将标志设置为False
+                        is_first_sentence = False
                     elif res.optional.event == EVENT_SessionFinished:
                         logger.bind(tag=TAG).debug(f"会话结束～～")
                         self._process_before_stop_play_files()
@@ -626,5 +655,110 @@ class TTSProvider(TTSProviderBase):
             )
         )
 
-    def wav_to_opus_data_audio_raw_stream(self, raw_data_var, is_end=False, callback: Callable[[Any], Any]=None):
-        return self.opus_encoder.encode_pcm_to_opus_stream(raw_data_var, is_end, callback=callback)
+    def wav_to_opus_data_audio_raw(self, raw_data_var, is_end=False):
+        opus_datas = self.opus_encoder.encode_pcm_to_opus(raw_data_var, is_end)
+        return opus_datas
+
+    def to_tts(self, text: str) -> list:
+        """非流式生成音频数据，用于生成音频及测试场景
+
+        Args:
+            text: 要转换的文本
+
+        Returns:
+            list: 音频数据列表
+        """
+        try:
+            # 创建事件循环
+            loop = asyncio.new_event_loop()
+            asyncio.set_event_loop(loop)
+
+            # 生成会话ID
+            session_id = uuid.uuid4().__str__().replace("-", "")
+
+            # 存储音频数据
+            audio_data = []
+
+            async def _generate_audio():
+                # 创建新的WebSocket连接
+                ws_header = {
+                    "X-Api-App-Key": self.appId,
+                    "X-Api-Access-Key": self.access_token,
+                    "X-Api-Resource-Id": self.resource_id,
+                    "X-Api-Connect-Id": uuid.uuid4(),
+                }
+                ws = await websockets.connect(
+                    self.ws_url, additional_headers=ws_header, max_size=1000000000
+                )
+
+                try:
+                    # 启动会话
+                    header = Header(
+                        message_type=FULL_CLIENT_REQUEST,
+                        message_type_specific_flags=MsgTypeFlagWithEvent,
+                        serial_method=JSON,
+                    ).as_bytes()
+                    optional = Optional(
+                        event=EVENT_StartSession, sessionId=session_id
+                    ).as_bytes()
+                    payload = self.get_payload_bytes(
+                        event=EVENT_StartSession, speaker=self.voice
+                    )
+                    await self.send_event(ws, header, optional, payload)
+
+                    # 发送文本
+                    header = Header(
+                        message_type=FULL_CLIENT_REQUEST,
+                        message_type_specific_flags=MsgTypeFlagWithEvent,
+                        serial_method=JSON,
+                    ).as_bytes()
+                    optional = Optional(
+                        event=EVENT_TaskRequest, sessionId=session_id
+                    ).as_bytes()
+                    payload = self.get_payload_bytes(
+                        event=EVENT_TaskRequest, text=text, speaker=self.voice
+                    )
+                    await self.send_event(ws, header, optional, payload)
+
+                    # 发送结束会话请求
+                    header = Header(
+                        message_type=FULL_CLIENT_REQUEST,
+                        message_type_specific_flags=MsgTypeFlagWithEvent,
+                        serial_method=JSON,
+                    ).as_bytes()
+                    optional = Optional(
+                        event=EVENT_FinishSession, sessionId=session_id
+                    ).as_bytes()
+                    payload = str.encode("{}")
+                    await self.send_event(ws, header, optional, payload)
+
+                    # 接收音频数据
+                    while True:
+                        msg = await ws.recv()
+                        res = self.parser_response(msg)
+
+                        if (
+                            res.optional.event == EVENT_TTSResponse
+                            and res.header.message_type == AUDIO_ONLY_RESPONSE
+                        ):
+                            opus_datas = self.wav_to_opus_data_audio_raw(res.payload)
+                            audio_data.extend(opus_datas)
+                        elif res.optional.event == EVENT_SessionFinished:
+                            break
+
+                finally:
+                    # 清理资源
+                    try:
+                        await ws.close()
+                    except:
+                        pass
+
+            # 运行异步任务
+            loop.run_until_complete(_generate_audio())
+            loop.close()
+
+            return audio_data
+
+        except Exception as e:
+            logger.bind(tag=TAG).error(f"生成音频数据失败: {str(e)}")
+            return []

main/xiaozhi-server/core/providers/tts/index_stream.py

@@ -3,6 +3,8 @@ import queue
 import asyncio
 import traceback
 import aiohttp
+import requests
+import time
 from config.logger import setup_logging
 from core.utils.tts import MarkdownCleaner
 from core.providers.tts.base import TTSProviderBase
@@ -25,13 +27,15 @@ class TTSProvider(TTSProviderBase):
         self.api_url = config.get("api_url", "http://8.138.114.124:11996/tts")
         self.audio_format = "pcm"
         self.before_stop_play_files = []
+        self.segment_count = 0
 
         # 创建Opus编码器 需注意接口返回的采样率为24000
         self.opus_encoder = opus_encoder_utils.OpusEncoderUtils(
             sample_rate=24000, channels=1, frame_size_ms=60
         )
 
-        # PCM缓冲区
+        # 文本缓冲区和PCM缓冲区
+        self.text_buffer = ""
         self.pcm_buffer = bytearray()
 
     def tts_text_priority_thread(self):
@@ -44,6 +48,7 @@ class TTSProvider(TTSProviderBase):
                     self.tts_stop_request = False
                     self.processed_chars = 0
                     self.tts_text_buff = []
+                    self.segment_count = 0
                     self.before_stop_play_files.clear()
                 elif ContentType.TEXT == message.content_type:
                     self.tts_text_buff.append(message.content_detail)
@@ -57,11 +62,14 @@ class TTSProvider(TTSProviderBase):
                     )
                     if message.content_file and os.path.exists(message.content_file):
                         # 先处理文件音频数据
-                        self._process_audio_file_stream(message.content_file, callback=lambda audio_data: self.handle_audio_file(audio_data, message.content_detail))
+                        file_audio = self._process_audio_file(message.content_file)
+                        self.before_stop_play_files.append(
+                            (file_audio, message.content_detail)
+                        )
 
                 if message.sentence_type == SentenceType.LAST:
                     # 处理剩余的文本
-                    self._process_remaining_text_stream(True)
+                    self._process_remaining_text(True)
 
             except queue.Empty:
                 continue
@@ -70,7 +78,7 @@ class TTSProvider(TTSProviderBase):
                     f"处理TTS文本失败: {str(e)}, 类型: {type(e).__name__}, 堆栈: {traceback.format_exc()}"
                 )
 
-    def _process_remaining_text_stream(self, is_last=False):
+    def _process_remaining_text(self, is_last=False):
         """处理剩余的文本并生成语音
         Returns:
             bool: 是否成功处理了文本
@@ -135,6 +143,8 @@ class TTSProvider(TTSProviderBase):
                         return
 
                     self.pcm_buffer.clear()
+                    opus_datas_cache = []
+
                     self.tts_audio_queue.put((SentenceType.FIRST, [], text))
 
                     # 处理音频流数据
@@ -148,22 +158,41 @@ class TTSProvider(TTSProviderBase):
                         while len(self.pcm_buffer) >= frame_bytes:
                             frame = bytes(self.pcm_buffer[:frame_bytes])
                             del self.pcm_buffer[:frame_bytes]
-
-                            self.opus_encoder.encode_pcm_to_opus_stream(
-                                frame,
-                                end_of_stream=False,
-                                callback=self.handle_opus
+                            opus = self.opus_encoder.encode_pcm_to_opus(
+                                frame, end_of_stream=False
                             )
+                            if opus:
+                                if self.segment_count < 10:  # 前10个片段直接发送
+                                    self.tts_audio_queue.put(
+                                        (SentenceType.MIDDLE, opus, None)
+                                    )
+                                    self.segment_count += 1
+                                else:
+                                    opus_datas_cache.extend(opus)
 
                     # flush 剩余不足一帧的数据
                     if self.pcm_buffer:
-                        self.opus_encoder.encode_pcm_to_opus_stream(
-                            bytes(self.pcm_buffer),
-                            end_of_stream=True,
-                            callback=self.handle_opus
+                        opus = self.opus_encoder.encode_pcm_to_opus(
+                            bytes(self.pcm_buffer), end_of_stream=True
                         )
+                        if opus:
+                            if self.segment_count < 10:  # 前10个片段直接发送
+                                # 直接发送
+                                self.tts_audio_queue.put(
+                                    (SentenceType.MIDDLE, opus, None)
+                                )
+                                self.segment_count += 1
+                            else:
+                                # 后续片段缓存
+                                opus_datas_cache.extend(opus)
                         self.pcm_buffer.clear()
 
+                    # 如果不是前10个片段，发送缓存的数据
+                    if self.segment_count >= 10 and opus_datas_cache:
+                        self.tts_audio_queue.put(
+                            (SentenceType.MIDDLE, opus_datas_cache, None)
+                        )
+
                     # 如果是最后一段，输出音频获取完毕
                     if is_last:
                         self._process_before_stop_play_files()
@@ -177,3 +206,59 @@ class TTSProvider(TTSProviderBase):
         await super().close()
         if hasattr(self, "opus_encoder"):
             self.opus_encoder.close()
+
+    def to_tts(self, text: str) -> list:
+        """非流式TTS处理，用于测试及保存音频文件的场景
+
+        Args:
+            text: 要转换的文本
+
+        Returns:
+            list: 返回opus编码后的音频数据列表
+        """
+        start_time = time.time()
+        text = MarkdownCleaner.clean_markdown(text)
+
+        payload = {"text": text, "character": self.character}
+
+        try:
+            with requests.post(self.api_url, json=payload, timeout=5) as response:
+                if response.status_code != 200:
+                    logger.bind(tag=TAG).error(
+                        f"TTS请求失败: {response.status_code}, {response.text}"
+                    )
+                    return []
+
+                logger.info(f"TTS请求成功: {text}, 耗时: {time.time() - start_time}秒")
+
+                # 使用opus编码器处理PCM数据
+                opus_datas = []
+                pcm_data = response.content
+
+                # 计算每帧的字节数
+                frame_bytes = int(
+                    self.opus_encoder.sample_rate
+                    * self.opus_encoder.channels
+                    * self.opus_encoder.frame_size_ms
+                    / 1000
+                    * 2
+                )
+
+                # 分帧处理PCM数据
+                for i in range(0, len(pcm_data), frame_bytes):
+                    frame = pcm_data[i : i + frame_bytes]
+                    if len(frame) < frame_bytes:
+                        # 最后一帧可能不足，用0填充
+                        frame = frame + b"\x00" * (frame_bytes - len(frame))
+
+                    opus = self.opus_encoder.encode_pcm_to_opus(
+                        frame, end_of_stream=(i + frame_bytes >= len(pcm_data))
+                    )
+                    if opus:
+                        opus_datas.extend(opus)
+
+                return opus_datas
+
+        except Exception as e:
+            logger.bind(tag=TAG).error(f"TTS请求异常: {e}")
+            return []

main/xiaozhi-server/core/providers/tts/linkerai.py

@@ -1,8 +1,10 @@
 import os
 import queue
-import aiohttp
 import asyncio
 import traceback
+import aiohttp
+import requests
+import time
 from config.logger import setup_logging
 from core.utils.tts import MarkdownCleaner
 from core.providers.tts.base import TTSProviderBase
@@ -22,15 +24,23 @@ class TTSProvider(TTSProviderBase):
         self.api_url = config.get("api_url")
         self.audio_format = "pcm"
         self.before_stop_play_files = []
+        self.segment_count = 0  # 添加片段计数器
 
         # 创建Opus编码器
         self.opus_encoder = opus_encoder_utils.OpusEncoderUtils(
             sample_rate=16000, channels=1, frame_size_ms=60
         )
 
+        # 添加文本缓冲区
+        self.text_buffer = ""
+
         # PCM缓冲区
         self.pcm_buffer = bytearray()
 
+    ###################################################################################
+    # linkerai单流式TTS重写父类的方法--开始
+    ###################################################################################
+
     def tts_text_priority_thread(self):
         """流式文本处理线程"""
         while not self.conn.stop_event.is_set():
@@ -41,6 +51,7 @@ class TTSProvider(TTSProviderBase):
                     self.tts_stop_request = False
                     self.processed_chars = 0
                     self.tts_text_buff = []
+                    self.segment_count = 0
                     self.before_stop_play_files.clear()
                 elif ContentType.TEXT == message.content_type:
                     self.tts_text_buff.append(message.content_detail)
@@ -54,10 +65,14 @@ class TTSProvider(TTSProviderBase):
                     )
                     if message.content_file and os.path.exists(message.content_file):
                         # 先处理文件音频数据
-                        self._process_audio_file_stream(message.content_file, callback=lambda audio_data: self.handle_audio_file(audio_data, message.content_detail))
+                        file_audio = self._process_audio_file(message.content_file)
+                        self.before_stop_play_files.append(
+                            (file_audio, message.content_detail)
+                        )
+
                 if message.sentence_type == SentenceType.LAST:
                     # 处理剩余的文本
-                    self._process_remaining_text_stream(True)
+                    self._process_remaining_text(True)
 
             except queue.Empty:
                 continue
@@ -66,7 +81,7 @@ class TTSProvider(TTSProviderBase):
                     f"处理TTS文本失败: {str(e)}, 类型: {type(e).__name__}, 堆栈: {traceback.format_exc()}"
                 )
 
-    def _process_remaining_text_stream(self, is_last=False):
+    def _process_remaining_text(self, is_last=False):
         """处理剩余的文本并生成语音
 
         Returns:
@@ -109,6 +124,10 @@ class TTSProvider(TTSProviderBase):
         finally:
             return None
 
+    ###################################################################################
+    # linkerai单流式TTS重写父类的方法--结束
+    ###################################################################################
+
     async def text_to_speak(self, text, is_last):
         """流式处理TTS音频，每句只推送一次音频列表"""
         await self._tts_request(text, is_last)
@@ -157,6 +176,8 @@ class TTSProvider(TTSProviderBase):
                         return
 
                     self.pcm_buffer.clear()
+                    opus_datas_cache = []
+
                     self.tts_audio_queue.put((SentenceType.FIRST, [], text))
 
                     # 兼容 iter_chunked / iter_chunks / iter_any
@@ -173,21 +194,41 @@ class TTSProvider(TTSProviderBase):
                             frame = bytes(self.pcm_buffer[:frame_bytes])
                             del self.pcm_buffer[:frame_bytes]
 
-                            self.opus_encoder.encode_pcm_to_opus_stream(
-                                frame,
-                                end_of_stream=False,
-                                callback=self.handle_opus
+                            opus = self.opus_encoder.encode_pcm_to_opus(
+                                frame, end_of_stream=False
                             )
+                            if opus:
+                                if self.segment_count < 10:  # 前10个片段直接发送
+                                    self.tts_audio_queue.put(
+                                        (SentenceType.MIDDLE, opus, None)
+                                    )
+                                    self.segment_count += 1
+                                else:
+                                    opus_datas_cache.extend(opus)
 
                     # flush 剩余不足一帧的数据
                     if self.pcm_buffer:
-                        self.opus_encoder.encode_pcm_to_opus_stream(
-                            bytes(self.pcm_buffer),
-                            end_of_stream=True,
-                            callback=self.handle_opus
+                        opus = self.opus_encoder.encode_pcm_to_opus(
+                            bytes(self.pcm_buffer), end_of_stream=True
                         )
+                        if opus:
+                            if self.segment_count < 10:  # 前10个片段直接发送
+                                # 直接发送
+                                self.tts_audio_queue.put(
+                                    (SentenceType.MIDDLE, opus, None)
+                                )
+                                self.segment_count += 1
+                            else:
+                                # 后续片段缓存
+                                opus_datas_cache.extend(opus)
                         self.pcm_buffer.clear()
 
+                    # 如果不是前10个片段，发送缓存的数据
+                    if self.segment_count >= 10 and opus_datas_cache:
+                        self.tts_audio_queue.put(
+                            (SentenceType.MIDDLE, opus_datas_cache, None)
+                        )
+
                     # 如果是最后一段，输出音频获取完毕
                     if is_last:
                         self._process_before_stop_play_files()
@@ -195,3 +236,73 @@ class TTSProvider(TTSProviderBase):
         except Exception as e:
             logger.bind(tag=TAG).error(f"TTS请求异常: {e}")
             self.tts_audio_queue.put((SentenceType.LAST, [], None))
+
+    def to_tts(self, text: str) -> list:
+        """非流式TTS处理，用于测试及保存音频文件的场景
+
+        Args:
+            text: 要转换的文本
+
+        Returns:
+            list: 返回opus编码后的音频数据列表
+        """
+        start_time = time.time()
+        text = MarkdownCleaner.clean_markdown(text)
+
+        params = {
+            "tts_text": text,
+            "spk_id": self.voice,
+            "frame_duration": 60,
+            "stream": False,
+            "target_sr": 16000,
+            "audio_format": self.audio_format,
+            "instruct_text": "请生成一段自然流畅的语音",
+        }
+        headers = {
+            "Authorization": f"Bearer {self.access_token}",
+            "Content-Type": "application/json",
+        }
+
+        try:
+            with requests.get(
+                self.api_url, params=params, headers=headers, timeout=5
+            ) as response:
+                if response.status_code != 200:
+                    logger.bind(tag=TAG).error(
+                        f"TTS请求失败: {response.status_code}, {response.text}"
+                    )
+                    return []
+
+                logger.info(f"TTS请求成功: {text}, 耗时: {time.time() - start_time}秒")
+
+                # 使用opus编码器处理PCM数据
+                opus_datas = []
+                pcm_data = response.content
+
+                # 计算每帧的字节数
+                frame_bytes = int(
+                    self.opus_encoder.sample_rate
+                    * self.opus_encoder.channels
+                    * self.opus_encoder.frame_size_ms
+                    / 1000
+                    * 2
+                )
+
+                # 分帧处理PCM数据
+                for i in range(0, len(pcm_data), frame_bytes):
+                    frame = pcm_data[i : i + frame_bytes]
+                    if len(frame) < frame_bytes:
+                        # 最后一帧可能不足，用0填充
+                        frame = frame + b"\x00" * (frame_bytes - len(frame))
+
+                    opus = self.opus_encoder.encode_pcm_to_opus(
+                        frame, end_of_stream=(i + frame_bytes >= len(pcm_data))
+                    )
+                    if opus:
+                        opus_datas.extend(opus)
+
+                return opus_datas
+
+        except Exception as e:
+            logger.bind(tag=TAG).error(f"TTS请求异常: {e}")
+            return []

main/xiaozhi-server/core/providers/tts/paddle_speech.py

@@ -8,7 +8,6 @@ import wave
 import websockets
 from core.providers.tts.base import TTSProviderBase
 from config.logger import setup_logging
-from datetime import datetime
 
 TAG = __name__
 logger = setup_logging()
@@ -19,12 +18,11 @@ class TTSProvider(TTSProviderBase):
         super().__init__(config, delete_audio_file)
         self.url = config.get("url", "ws://192.168.1.10:8092/paddlespeech/tts/streaming")
         self.protocol = config.get("protocol", "websocket")
-        
         if config.get("private_voice"):
             self.spk_id = int(config.get("private_voice"))
         else:
-            self.spk_id = int(config.get("spk_id", "0"))
-            
+            self.spk_id = int(config.get("spk_id", "0"))  
+        
         sample_rate = config.get("sample_rate", 24000)
         self.sample_rate = float(sample_rate) if sample_rate else 24000
         
@@ -34,21 +32,7 @@ class TTSProvider(TTSProviderBase):
         volume = config.get("volume", 1.0)
         self.volume = float(volume) if volume else 1.0
         
-        self.delete_audio_file = config.get("delete_audio", True)
-        if not self.delete_audio_file:
-            timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
-            save_path = config.get("save_path")
-            if save_path:
-                if not save_path.endswith('.wav'):
-                    save_path = f"{save_path}_{timestamp}.wav"
-                else:
-                    other_path = save_path[:-4]
-                    save_path = f"{other_path}_{timestamp}.wav"
-                self.save_path = save_path
-            else:
-                self.save_path = f"./streaming_tts_{timestamp}.wav"
-        else:
-            self.save_path = None
+        self.save_path = config.get("save_path", "./streaming_tts.wav")
 
     async def pcm_to_wav(self, pcm_data: bytes, sample_rate: int = 24000, num_channels: int = 1,
                          bits_per_sample: int = 16) -> bytes:
@@ -167,12 +151,6 @@ class TTSProvider(TTSProviderBase):
                 # 接收结束响应避免服务抛出异常
                 await ws.recv()
 
-                # 根据配置决定是否保存文件
-                if not self.delete_audio_file and self.save_path:
-                    with open(self.save_path, "wb") as f:
-                        f.write(wav_data)
-                    logger.bind(tag=TAG).info(f"音频文件已保存到: {self.save_path}")
-                
                 # 返回或保存音频数据
                 if output_file:
                     with open(output_file, "wb") as file_to_save:
@@ -181,4 +159,4 @@ class TTSProvider(TTSProviderBase):
                     return wav_data
 
         except Exception as e:
-            raise Exception(f"Error during TTS WebSocket request: {e} while processing text: {text}")
+            raise Exception(f"Error during TTS WebSocket request: {e} while processing text: {text}")

main/xiaozhi-server/core/providers/vad/silero.py

@@ -33,7 +33,7 @@ class VADProvider(VADProviderBase):
             int(min_silence_duration_ms) if min_silence_duration_ms else 1000
         )
 
-        # 至少要多少帧才算有语音,增加灵敏度
+        # 至少要多少帧才算有语音
         self.frame_window_threshold = 1
 
     def is_vad(self, conn, opus_packet):
@@ -70,7 +70,9 @@ class VADProvider(VADProviderBase):
 
                 # 更新滑动窗口
                 conn.client_voice_window.append(is_voice)
-                client_have_voice = (conn.client_voice_window.count(True) >= self.frame_window_threshold)
+                client_have_voice = (
+                    conn.client_voice_window.count(True) >= self.frame_window_threshold
+                )
 
                 # 如果之前有声音，但本次没有声音，且与上次有声音的时间差已经超过了静默阈值，则认为已经说完一句话
                 if conn.client_have_voice and not client_have_voice:

main/xiaozhi-server/core/utils/opus_encoder_utils.py

@@ -5,8 +5,9 @@ Opus编码工具类
 
 import logging
 import traceback
+
 import numpy as np
-from typing import Optional, Callable, Any
+from typing import List, Optional
 from opuslib_next import Encoder
 from opuslib_next import constants
 
@@ -55,14 +56,13 @@ class OpusEncoderUtils:
         self.encoder.reset_state()
         self.buffer = np.array([], dtype=np.int16)
 
-    def encode_pcm_to_opus_stream(self, pcm_data: bytes, end_of_stream: bool, callback: Callable[[Any], Any]):
+    def encode_pcm_to_opus(self, pcm_data: bytes, end_of_stream: bool) -> List[bytes]:
         """
-        将PCM数据编码为Opus格式，以流式方式进行处理
+        将PCM数据编码为Opus格式
 
         Args:
             pcm_data: PCM字节数据
-            end_of_stream: 是否为流的结束,
-            callback: opus处理方法
+            end_of_stream: 是否为流的结束
 
         Returns:
             Opus数据包列表
@@ -76,6 +76,7 @@ class OpusEncoderUtils:
         # 将新数据追加到缓冲区
         self.buffer = np.append(self.buffer, new_samples)
 
+        opus_packets = []
         offset = 0
 
         # 处理所有完整帧
@@ -83,7 +84,7 @@ class OpusEncoderUtils:
             frame = self.buffer[offset : offset + self.total_frame_size]
             output = self._encode(frame)
             if output:
-                callback(output)
+                opus_packets.append(output)
             offset += self.total_frame_size
 
         # 保留未处理的样本
@@ -97,9 +98,11 @@ class OpusEncoderUtils:
 
             output = self._encode(last_frame)
             if output:
-                callback(output)
+                opus_packets.append(output)
             self.buffer = np.array([], dtype=np.int16)
 
+        return opus_packets
+
     def _encode(self, frame: np.ndarray) -> Optional[bytes]:
         """编码一帧音频数据"""
         try:

main/xiaozhi-server/core/utils/p3.py

@@ -1,12 +1,15 @@
-import io
 import struct
-from typing import Callable, Any
 
+def decode_opus_from_file(input_file):
+    """
+    从p3文件中解码 Opus 数据，并返回一个 Opus 数据包的列表以及总时长。
+    """
+    opus_datas = []
+    total_frames = 0
+    sample_rate = 16000  # 文件采样率
+    frame_duration_ms = 60  # 帧时长
+    frame_size = int(sample_rate * frame_duration_ms / 1000)
 
-def decode_opus_from_file_stream(input_file, callback: Callable[[Any], Any]):
-    """
-    从p3文件中解码 Opus 数据，由 callback 处理 Opus 数据包。
-    """
     with open(input_file, 'rb') as f:
         while True:
             # 读取头部（4字节）：[1字节类型，1字节保留，2字节长度]
@@ -22,13 +25,23 @@ def decode_opus_from_file_stream(input_file, callback: Callable[[Any], Any]):
             if len(opus_data) != data_len:
                 raise ValueError(f"Data length({len(opus_data)}) mismatch({data_len}) in the file.")
 
-            callback(opus_data)
+            opus_datas.append(opus_data)
+            total_frames += 1
 
+    # 计算总时长
+    total_duration = (total_frames * frame_duration_ms) / 1000.0
+    return opus_datas, total_duration
 
-def decode_opus_from_bytes_stream(input_bytes, callback: Callable[[Any], Any]):
+def decode_opus_from_bytes(input_bytes):
     """
-    从p3二进制数据中解码 Opus 数据，由 callback 处理 Opus 数据包。
+    从p3二进制数据中解码 Opus 数据，并返回一个 Opus 数据包的列表以及总时长。
     """
+    import io
+    opus_datas = []
+    total_frames = 0
+    sample_rate = 16000  # 文件采样率
+    frame_duration_ms = 60  # 帧时长
+    frame_size = int(sample_rate * frame_duration_ms / 1000)
 
     f = io.BytesIO(input_bytes)
     while True:
@@ -39,4 +52,8 @@ def decode_opus_from_bytes_stream(input_bytes, callback: Callable[[Any], Any]):
         opus_data = f.read(data_len)
         if len(opus_data) != data_len:
             raise ValueError(f"Data length({len(opus_data)}) mismatch({data_len}) in the bytes.")
-        callback(opus_data)
+        opus_datas.append(opus_data)
+        total_frames += 1
+
+    total_duration = (total_frames * frame_duration_ms) / 1000.0
+    return opus_datas, total_duration

main/xiaozhi-server/core/utils/util.py

@@ -1,17 +1,16 @@
+import json
+import socket
+import subprocess
 import re
 import os
-import json
-import copy
-import socket
-import requests
-import subprocess
-import numpy as np
-import opuslib_next
+import wave
 from io import BytesIO
 from core.utils import p3
+import numpy as np
+import requests
+import opuslib_next
 from pydub import AudioSegment
-from typing import Callable, Any
-from core.providers.tts.dto.dto import SentenceType
+import copy
 
 TAG = __name__
 emoji_map = {
@@ -212,7 +211,7 @@ def extract_json_from_string(input_string):
     return None
 
 
-def audio_to_data_stream(audio_file_path, is_opus=True, callback: Callable[[Any], Any]=None) -> None:
+def audio_to_data(audio_file_path, is_opus=True):
     # 获取文件后缀名
     file_type = os.path.splitext(audio_file_path)[1]
     if file_type:
@@ -225,29 +224,33 @@ def audio_to_data_stream(audio_file_path, is_opus=True, callback: Callable[[Any]
     # 转换为单声道/16kHz采样率/16位小端编码（确保与编码器匹配）
     audio = audio.set_channels(1).set_frame_rate(16000).set_sample_width(2)
 
+    # 音频时长(秒)
+    duration = len(audio) / 1000.0
+
     # 获取原始PCM数据（16位小端）
     raw_data = audio.raw_data
-    pcm_to_data_stream(raw_data, is_opus, callback)
+    return pcm_to_data(raw_data, is_opus), duration
 
 
-def audio_bytes_to_data_stream(audio_bytes, file_type, is_opus, callback: Callable[[Any], Any]) -> None:
+def audio_bytes_to_data(audio_bytes, file_type, is_opus=True):
     """
     直接用音频二进制数据转为opus/pcm数据，支持wav、mp3、p3
     """
     if file_type == "p3":
         # 直接用p3解码
-        return p3.decode_opus_from_bytes_stream(audio_bytes, callback)
+        return p3.decode_opus_from_bytes(audio_bytes)
     else:
         # 其他格式用pydub
         audio = AudioSegment.from_file(
             BytesIO(audio_bytes), format=file_type, parameters=["-nostdin"]
         )
         audio = audio.set_channels(1).set_frame_rate(16000).set_sample_width(2)
+        duration = len(audio) / 1000.0
         raw_data = audio.raw_data
-        pcm_to_data_stream(raw_data, is_opus, callback)
+        return pcm_to_data(raw_data, is_opus), duration
 
 
-def pcm_to_data_stream(raw_data, is_opus=True, callback: Callable[[Any], Any] = None):
+def pcm_to_data(raw_data, is_opus=True):
     # 初始化Opus编码器
     encoder = opuslib_next.Encoder(16000, 1, opuslib_next.APPLICATION_AUDIO)
 
@@ -255,6 +258,7 @@ def pcm_to_data_stream(raw_data, is_opus=True, callback: Callable[[Any], Any] =
     frame_duration = 60  # 60ms per frame
     frame_size = int(16000 * frame_duration / 1000)  # 960 samples/frame
 
+    datas = []
     # 按帧处理所有音频数据（包括最后一帧可能补零）
     for i in range(0, len(raw_data), frame_size * 2):  # 16bit=2bytes/sample
         # 获取当前帧的二进制数据
@@ -269,28 +273,40 @@ def pcm_to_data_stream(raw_data, is_opus=True, callback: Callable[[Any], Any] =
             np_frame = np.frombuffer(chunk, dtype=np.int16)
             # 编码Opus数据
             frame_data = encoder.encode(np_frame.tobytes(), frame_size)
-            callback(frame_data)
         else:
             frame_data = chunk if isinstance(chunk, bytes) else bytes(chunk)
-            callback(frame_data)
 
-def play_audio_response(conn, response):
-    """音频响应处理"""
-    conn.tts.tts_audio_queue.put((SentenceType.FIRST, [], response.get("text")))
-    play_audio_frames(conn, response.get("file_path"))
-    conn.tts.tts_audio_queue.put((SentenceType.LAST, [], None))
+        datas.append(frame_data)
+
+    return datas
 
 
-def play_audio_frames(conn, file_path):
-    """播放音频文件并处理发送帧数据"""
-    def handle_audio_frame(frame_data):
-        conn.tts.tts_audio_queue.put((SentenceType.MIDDLE, frame_data, None))
+def opus_datas_to_wav_bytes(opus_datas, sample_rate=16000, channels=1):
+    """
+    将opus帧列表解码为wav字节流
+    """
+    decoder = opuslib_next.Decoder(sample_rate, channels)
+    pcm_datas = []
+
+    frame_duration = 60  # ms
+    frame_size = int(sample_rate * frame_duration / 1000)  # 960
+
+    for opus_frame in opus_datas:
+        # 解码为PCM（返回bytes，2字节/采样点）
+        pcm = decoder.decode(opus_frame, frame_size)
+        pcm_datas.append(pcm)
+
+    pcm_bytes = b"".join(pcm_datas)
+
+    # 写入wav字节流
+    wav_buffer = BytesIO()
+    with wave.open(wav_buffer, "wb") as wf:
+        wf.setnchannels(channels)
+        wf.setsampwidth(2)  # 16bit
+        wf.setframerate(sample_rate)
+        wf.writeframes(pcm_bytes)
+    return wav_buffer.getvalue()
 
-    audio_to_data_stream(
-        file_path,
-        is_opus=True,
-        callback=handle_audio_frame
-    )
 
 def check_vad_update(before_config, new_config):
     if (

main/xiaozhi-server/core/utils/wakeup_word.py

@@ -0,0 +1,140 @@
+import os
+import re
+import yaml
+import time
+import hashlib
+import portalocker
+from typing import Dict
+
+
+class FileLock:
+    def __init__(self, file, timeout=5):
+        self.file = file
+        self.timeout = timeout
+        self.start_time = None
+
+    def __enter__(self):
+        self.start_time = time.time()
+        while True:
+            try:
+                portalocker.lock(self.file, portalocker.LOCK_EX | portalocker.LOCK_NB)
+                return self.file
+            except portalocker.LockException:
+                if time.time() - self.start_time > self.timeout:
+                    raise TimeoutError("获取文件锁超时")
+                time.sleep(0.1)
+
+    def __exit__(self, exc_type, exc_val, exc_tb):
+        portalocker.unlock(self.file)
+
+
+class WakeupWordsConfig:
+    def __init__(self):
+        self.config_file = "data/.wakeup_words.yaml"
+        self.assets_dir = "config/assets/wakeup_words"
+        self._ensure_directories()
+        self._config_cache = None
+        self._last_load_time = 0
+        self._cache_ttl = 1  # 缓存有效期（秒）
+        self._lock_timeout = 5  # 文件锁超时时间（秒）
+
+    def _ensure_directories(self):
+        """确保必要的目录存在"""
+        os.makedirs(os.path.dirname(self.config_file), exist_ok=True)
+        os.makedirs(self.assets_dir, exist_ok=True)
+
+    def _load_config(self) -> Dict:
+        """加载配置文件，使用缓存机制"""
+        current_time = time.time()
+
+        # 如果缓存有效，直接返回缓存
+        if (
+            self._config_cache is not None
+            and current_time - self._last_load_time < self._cache_ttl
+        ):
+            return self._config_cache
+
+        try:
+            with open(self.config_file, "a+") as f:
+                with FileLock(f, timeout=self._lock_timeout):
+                    f.seek(0)
+                    content = f.read()
+                    config = yaml.safe_load(content) if content else {}
+                    self._config_cache = config
+                    self._last_load_time = current_time
+                    return config
+        except (TimeoutError, IOError) as e:
+            print(f"加载配置文件失败: {e}")
+            return {}
+        except Exception as e:
+            print(f"加载配置文件时发生未知错误: {e}")
+            return {}
+
+    def _save_config(self, config: Dict):
+        """保存配置到文件，使用文件锁保护"""
+        try:
+            with open(self.config_file, "w") as f:
+                with FileLock(f, timeout=self._lock_timeout):
+                    yaml.dump(config, f, allow_unicode=True)
+                    self._config_cache = config
+                    self._last_load_time = time.time()
+        except (TimeoutError, IOError) as e:
+            print(f"保存配置文件失败: {e}")
+            raise
+        except Exception as e:
+            print(f"保存配置文件时发生未知错误: {e}")
+            raise
+
+    def get_wakeup_response(self, voice: str) -> Dict:
+        voice = hashlib.md5(voice.encode()).hexdigest()
+        """获取唤醒词回复配置"""
+        config = self._load_config()
+
+        if not config or voice not in config:
+            return None
+
+        # 检查文件大小
+        file_path = config[voice]["file_path"]
+        if not os.path.exists(file_path) or os.stat(file_path).st_size < (15 * 1024):
+            return None
+
+        return config[voice]
+
+    def update_wakeup_response(self, voice: str, file_path: str, text: str):
+        """更新唤醒词回复配置"""
+        try:
+            # 过滤表情符号
+            filtered_text = re.sub(r'[\U0001F600-\U0001F64F\U0001F900-\U0001F9FF]', '', text)
+            
+            config = self._load_config()
+            voice_hash = hashlib.md5(voice.encode()).hexdigest()
+            config[voice_hash] = {
+                "voice": voice,
+                "file_path": file_path,
+                "time": time.time(),
+                "text": filtered_text,
+            }
+            self._save_config(config)
+        except Exception as e:
+            print(f"更新唤醒词回复配置失败: {e}")
+            raise
+
+    def generate_file_path(self, voice: str) -> str:
+        """生成音频文件路径，使用voice的哈希值作为文件名"""
+        try:
+            # 生成voice的哈希值
+            voice_hash = hashlib.md5(voice.encode()).hexdigest()
+            file_path = os.path.join(self.assets_dir, f"{voice_hash}.wav")
+
+            # 如果文件已存在，先删除
+            if os.path.exists(file_path):
+                try:
+                    os.remove(file_path)
+                except Exception as e:
+                    print(f"删除已存在的音频文件失败: {e}")
+                    raise
+
+            return file_path
+        except Exception as e:
+            print(f"生成音频文件路径失败: {e}")
+            raise

main/xiaozhi-server/plugins_func/functions/play_music.py

@@ -212,6 +212,7 @@ async def play_local_music(conn, specific_file=None):
             conn.logger.bind(tag=TAG).error(f"选定的音乐文件不存在: {music_path}")
             return
         text = _get_random_play_prompt(selected_music)
+        await send_stt_message(conn, text)
         conn.dialogue.put(Message(role="assistant", content=text))
 
         if conn.intent_type == "intent_llm":