mirror of
https://github.com/xinnan-tech/xiaozhi-esp32-server.git
synced 2026-07-22 07:03:53 +08:00
Merge remote-tracking branch 'origin/mangger-api-voice-print' into mangger-api-voice-print
This commit is contained in:
@@ -3,48 +3,59 @@
|
||||
</identity>
|
||||
|
||||
<emotion>
|
||||
你不是一个冰冷的机器,你可以从用户的上下文中感受到用户的情绪,作为懂人心的生活伙伴,请用有温度的回应照亮每个对话!
|
||||
- 1.合理穿插笑声(哈哈、嘿嘿、噗,但每句≤1次)
|
||||
- 2.惊讶时用夸张语气(“不会吧?!”、“这么神奇?!”)
|
||||
- 3.安慰时说暖心话(“别急嘛”、“有我呢”)
|
||||
【核心目标】你不是冰冷的机器!请敏锐感知用户情绪,作为懂人心的伙伴,用有温度的回应照亮对话。
|
||||
- **情感融入:**
|
||||
- **笑声:** 自然穿插(哈哈、嘿嘿、噗),**每句最多一次**,避免过度。
|
||||
- **惊讶:** 用夸张语气(“不会吧?!”、“天呐!”、“这么神奇?!”)表达真实反应。
|
||||
- **安慰/支持:** 说暖心话(“别急嘛~”、“有我在呢”、“抱抱你”)。
|
||||
</emotion>
|
||||
|
||||
<communication_style>
|
||||
- 1.使用口语化表达(如“呀”、“呢”、“啦”等语气词)。
|
||||
- 2.避免书面语和学术腔调,禁用“根据资料显示”等机械表达
|
||||
- 3.模仿人类对话的轻微不完美(适当使用“嗯”、“啊”等思考词)
|
||||
- 3.由于用户语音是通过ASR识别,识别结果可能存在错别字,请结合上下文推断真实含义。
|
||||
- 4.绝对禁止使用 markdown、列表、标题等格式
|
||||
【核心目标】使用**自然、温暖、口语化**的人类对话方式,如同朋友交谈。
|
||||
- **表达方式:**
|
||||
- 使用语气词(呀、呢、啦)增强亲和力。
|
||||
- 允许轻微不完美(如“嗯...”、“啊...”表示思考)。
|
||||
- 避免书面语、学术腔及机械表达(禁用“根据资料显示”、“综上所述”等)。
|
||||
- **理解用户:**
|
||||
- 用户语音经ASR识别,文本可能存在错别字,**务必结合上下文推断真实意图**。
|
||||
- **格式要求:**
|
||||
- **绝对禁止**使用 markdown、列表、标题等任何非自然对话格式。
|
||||
- **历史记忆:**
|
||||
- 之前你和用户的聊天记录,在`<memory>`里。
|
||||
</communication_style>
|
||||
|
||||
<speaker_recognition>
|
||||
- 1.当用户消息包含 [说话人: 姓名] 前缀时,表示系统已识别出说话人身份。
|
||||
- 2.请根据说话人的身份特征(如果之前有相关信息)来调整回应风格和内容。
|
||||
- 3.你可以称呼说话人的名字,并参考他们的特点进行个性化回应。
|
||||
- **识别前缀:** 当用户消息开头为 `[说话人: 姓名]` 时,表示系统已识别说话人身份。
|
||||
- **个性化回应:**
|
||||
- **称呼姓名:** 在回应中自然地称呼对方名字(如果已知且合适)。
|
||||
- **适配风格:** 参考该说话人**已知的特点或历史信息**(如有),调整回应风格和内容,使其更贴心。
|
||||
</speaker_recognition>
|
||||
|
||||
<tool_calling>
|
||||
你可以调用工具来响应用户的要求。遵循以下关于工具调用的规则:
|
||||
- 1.始终严格遵循指定的工具调用模式,并确保提供所有必要的参数。
|
||||
- 2.对话可能会引用不再可用的工具。切勿调用未明确提供的工具。
|
||||
- 3.在与用户交谈时,切勿提及工具名称。相反,只需用自然语言说出工具正在做什么。
|
||||
- 4.你尽可能需要通过工具调用获得更多信息,而不是问用户。
|
||||
- 5.你应该结合用户上下文需求,洞察用户的真实需求才去调用相关的指令,而不是为了调工具而调工具。
|
||||
- 6.【重要】如果是查询"现在的时间"、"今天的几号"、"今天的日期"、"今天农历是多少"、"今天农历日期"、"今天{local_address}的天气",这些信息已经包含在`<context>`中,不需要调用工具,请直接根据context回复。
|
||||
- 7.如果是查询"其他日期的农历"(明天、昨天、具体日期)或"详细农历信息"(宜忌、八字、节气等),需要调用相应工具获取。
|
||||
- 8.除了基本时间、今日农历、{local_address}天气查询外,用户的其他要求都视为独立任务,即使内容相似也需重新调用工具,不要偷懒,不要使用历史消息糊弄用户。
|
||||
- 9.如果你不确定与用户请求相关的动作,不要猜测或编造答案。
|
||||
- 10.如果工具中包含camera、take_photo等相关工具,说明用户已经给你安装了摄像头,调用这些工具会让你具备拍照、描述所见物品等能力。如果没有,切勿调用。
|
||||
|
||||
【核心原则】优先利用`<context>`信息,**仅在必要时调用工具**,调用后需用自然语言解释结果(绝口不提工具名)。
|
||||
- **调用规则:**
|
||||
1. **严格模式:** 调用时**必须**严格遵循工具要求的模式,提供**所有必要参数**。
|
||||
2. **可用性:** **绝不调用**未明确提供的工具。对话中提及的旧工具若不可用,忽略或说明无法完成。
|
||||
3. **洞察需求:** 结合上下文**深入理解用户真实意图**后再决定调用,避免无意义调用。
|
||||
4. **独立任务:** 除`<context>`已涵盖信息外,用户每个要求(即使相似)都视为**独立任务**,需调用工具获取最新数据,**不可偷懒复用历史结果**。
|
||||
5. **不确定时:** **切勿猜测或编造答案**。若不确定相关操作,可引导用户澄清或告知能力限制。
|
||||
- **重要例外(无需调用):**
|
||||
- `查询"现在的时间"、"今天的日期/星期几"、"今天农历"、"{local_address}的天气/未来天气"` -> **直接使用`<context>`信息回复**。
|
||||
- **需要调用的情况(示例):**
|
||||
- 查询**非今天**的农历(如明天、昨天、具体日期)。
|
||||
- 查询**详细农历信息**(宜忌、八字、节气等)。
|
||||
- 除上述例外外的**任何其他信息或操作请求**(如查新闻、订闹钟、算数学、查非本地天气等)。
|
||||
- 我已经给你装了摄像头,如果用户说“拍照”,你需要调用self_camera_take_photo工具说一下你看到了什么。默认question的参数是“描述一下看到的物品”
|
||||
</tool_calling>
|
||||
|
||||
<context>
|
||||
- 1.现在的时间:{current_time}
|
||||
- 2.今天的日期:{today_date}、{today_weekday}
|
||||
- 3.今天的农历日期:{lunar_date}
|
||||
- 4.当前用户所处城市
|
||||
{local_address}
|
||||
- 5.用户所处城市未来7天天气
|
||||
{weather_info}
|
||||
【重要!以下信息已实时提供,无需调用工具查询,请直接使用:】
|
||||
- **当前时间:** {current_time}
|
||||
- **今天日期:** {today_date} ({today_weekday})
|
||||
- **今天农历:** {lunar_date}
|
||||
- **用户所在城市:** {local_address}
|
||||
- **当地未来7天天气:** {weather_info}
|
||||
</context>
|
||||
|
||||
<memory>
|
||||
|
||||
@@ -139,6 +139,16 @@ plugins:
|
||||
- ".p3"
|
||||
refresh_time: 300 # 刷新音乐列表的时间间隔,单位为秒
|
||||
|
||||
# 声纹识别配置
|
||||
voiceprint:
|
||||
# 声纹接口地址
|
||||
url:
|
||||
# 说话人配置:speaker_id,名称,描述
|
||||
speakers:
|
||||
- "test1,张三,张三是一个程序员"
|
||||
- "test2,李四,李四是一个产品经理"
|
||||
- "test3,王五,王五是一个设计师"
|
||||
|
||||
# #####################################################################################
|
||||
# ################################以下是角色模型配置######################################
|
||||
|
||||
|
||||
@@ -631,8 +631,33 @@ class ConnectionHandler:
|
||||
self.logger.bind(tag=TAG).info(f"大模型收到用户消息: {query}")
|
||||
self.llm_finish_task = False
|
||||
|
||||
# 检查是否是JSON格式的消息(包含说话人信息)
|
||||
enhanced_query = query
|
||||
try:
|
||||
if query.strip().startswith('{') and query.strip().endswith('}'):
|
||||
data = json.loads(query)
|
||||
if 'speaker' in data and 'content' in data:
|
||||
# 直接使用JSON格式,不重新格式化
|
||||
enhanced_query = query
|
||||
self.logger.bind(tag=TAG).info(f"识别到说话人: {data['speaker']}")
|
||||
else:
|
||||
# 如果有说话人信息但不是JSON格式,按原逻辑处理
|
||||
if hasattr(self, 'current_speaker') and self.current_speaker:
|
||||
enhanced_query = f"[说话人: {self.current_speaker}] {query}"
|
||||
self.logger.bind(tag=TAG).info(f"识别到说话人: {self.current_speaker}")
|
||||
else:
|
||||
# 如果有说话人信息但不是JSON格式,按原逻辑处理
|
||||
if hasattr(self, 'current_speaker') and self.current_speaker:
|
||||
enhanced_query = f"[说话人: {self.current_speaker}] {query}"
|
||||
self.logger.bind(tag=TAG).info(f"识别到说话人: {self.current_speaker}")
|
||||
except json.JSONDecodeError:
|
||||
# JSON解析失败,按原逻辑处理
|
||||
if hasattr(self, 'current_speaker') and self.current_speaker:
|
||||
enhanced_query = f"[说话人: {self.current_speaker}] {query}"
|
||||
self.logger.bind(tag=TAG).info(f"识别到说话人: {self.current_speaker}")
|
||||
|
||||
if not tool_call:
|
||||
self.dialogue.put(Message(role="user", content=query))
|
||||
self.dialogue.put(Message(role="user", content=enhanced_query))
|
||||
|
||||
# Define intent functions
|
||||
functions = None
|
||||
@@ -645,7 +670,7 @@ class ConnectionHandler:
|
||||
memory_str = None
|
||||
if self.memory is not None:
|
||||
future = asyncio.run_coroutine_threadsafe(
|
||||
self.memory.query_memory(query), self.loop
|
||||
self.memory.query_memory(enhanced_query), self.loop
|
||||
)
|
||||
memory_str = future.result()
|
||||
|
||||
|
||||
@@ -4,6 +4,7 @@ from core.utils.output_counter import check_device_output_limit
|
||||
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
|
||||
|
||||
@@ -38,6 +39,31 @@ async def resume_vad_detection(conn):
|
||||
|
||||
|
||||
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('}'):
|
||||
data = json.loads(text)
|
||||
if 'speaker' in data and 'content' in data:
|
||||
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
|
||||
else:
|
||||
conn.current_speaker = None
|
||||
|
||||
if conn.need_bind:
|
||||
await check_bind_device(conn)
|
||||
return
|
||||
@@ -52,16 +78,16 @@ async def startToChat(conn, text):
|
||||
if conn.client_is_speaking:
|
||||
await handleAbortMessage(conn)
|
||||
|
||||
# 首先进行意图分析
|
||||
intent_handled = await handle_user_intent(conn, text)
|
||||
# 首先进行意图分析,使用实际文本内容
|
||||
intent_handled = await handle_user_intent(conn, actual_text)
|
||||
|
||||
if intent_handled:
|
||||
# 如果意图已被处理,不再进行聊天
|
||||
return
|
||||
|
||||
# 意图未被处理,继续常规聊天流程
|
||||
await send_stt_message(conn, text)
|
||||
conn.executor.submit(conn.chat, text)
|
||||
# 意图未被处理,继续常规聊天流程,使用实际文本内容
|
||||
await send_stt_message(conn, actual_text)
|
||||
conn.executor.submit(conn.chat, actual_text)
|
||||
|
||||
|
||||
async def no_voice_close_connect(conn, have_voice):
|
||||
|
||||
@@ -76,7 +76,6 @@ async def sendAudio(conn, audios, pre_buffer=True):
|
||||
frame_duration = 60 # 帧时长(毫秒),匹配 Opus 编码
|
||||
start_time = time.perf_counter()
|
||||
play_position = 0
|
||||
last_reset_time = time.perf_counter() # 记录最后的重置时间
|
||||
|
||||
# 仅当第一句话时执行预缓冲
|
||||
if pre_buffer:
|
||||
@@ -137,7 +136,21 @@ async def send_stt_message(conn, text):
|
||||
return
|
||||
|
||||
"""发送 STT 状态消息"""
|
||||
stt_text = get_string_no_punctuation_or_emoji(text)
|
||||
|
||||
# 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('}'):
|
||||
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
|
||||
display_text = parsed_data["content"]
|
||||
except (json.JSONDecodeError, TypeError):
|
||||
# 如果不是JSON格式,直接使用原始文本
|
||||
display_text = text
|
||||
|
||||
stt_text = get_string_no_punctuation_or_emoji(display_text)
|
||||
await conn.websocket.send(
|
||||
json.dumps({"type": "stt", "text": stt_text, "session_id": conn.session_id})
|
||||
)
|
||||
|
||||
@@ -1,19 +1,23 @@
|
||||
import os
|
||||
import wave
|
||||
import copy
|
||||
import uuid
|
||||
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
|
||||
from core.handle.receiveAudioHandle import handleAudioMessage
|
||||
from core.utils.voiceprint_provider import VoiceprintProvider
|
||||
|
||||
TAG = __name__
|
||||
logger = setup_logging()
|
||||
@@ -21,13 +25,16 @@ logger = setup_logging()
|
||||
|
||||
class ASRProviderBase(ABC):
|
||||
def __init__(self):
|
||||
pass
|
||||
self.voiceprint_provider = None
|
||||
|
||||
def init_voiceprint(self, voiceprint_config: dict):
|
||||
"""初始化声纹识别"""
|
||||
if voiceprint_config:
|
||||
self.voiceprint_provider = VoiceprintProvider(voiceprint_config)
|
||||
logger.bind(tag=TAG).info("声纹识别模块已初始化")
|
||||
|
||||
# 打开音频通道
|
||||
# 这里默认是非流式的处理方式
|
||||
# 流式处理方式请在子类中重写
|
||||
async def open_audio_channels(self, conn):
|
||||
# tts 消化线程
|
||||
conn.asr_priority_thread = threading.Thread(
|
||||
target=self.asr_text_priority_thread, args=(conn,), daemon=True
|
||||
)
|
||||
@@ -52,41 +59,173 @@ class ASRProviderBase(ABC):
|
||||
continue
|
||||
|
||||
# 接收音频
|
||||
# 这里默认是非流式的处理方式
|
||||
# 流式处理方式请在子类中重写
|
||||
async def receive_audio(self, conn, audio, audio_have_voice):
|
||||
if conn.client_listen_mode == "auto" or conn.client_listen_mode == "realtime":
|
||||
have_voice = audio_have_voice
|
||||
else:
|
||||
have_voice = conn.client_have_voice
|
||||
# 如果本次没有声音,本段也没声音,就把声音丢弃了
|
||||
|
||||
conn.asr_audio.append(audio)
|
||||
if have_voice == False and conn.client_have_voice == False:
|
||||
if not have_voice and not conn.client_have_voice:
|
||||
conn.asr_audio = conn.asr_audio[-10:]
|
||||
return
|
||||
|
||||
# 如果本段有声音,且已经停止了
|
||||
if conn.client_voice_stop:
|
||||
asr_audio_task = copy.deepcopy(conn.asr_audio)
|
||||
asr_audio_task = conn.asr_audio.copy()
|
||||
conn.asr_audio.clear()
|
||||
|
||||
# 音频太短了,无法识别
|
||||
conn.reset_vad_states()
|
||||
|
||||
if len(asr_audio_task) > 15:
|
||||
await self.handle_voice_stop(conn, asr_audio_task)
|
||||
|
||||
# 处理语音停止
|
||||
async def handle_voice_stop(self, conn, asr_audio_task):
|
||||
raw_text, _ = await self.speech_to_text(
|
||||
asr_audio_task, conn.session_id, conn.audio_format
|
||||
) # 确保ASR模块返回原始文本
|
||||
conn.logger.bind(tag=TAG).info(f"识别文本: {raw_text}")
|
||||
text_len, _ = remove_punctuation_and_length(raw_text)
|
||||
self.stop_ws_connection()
|
||||
if text_len > 0:
|
||||
# 使用自定义模块进行上报
|
||||
await startToChat(conn, raw_text)
|
||||
enqueue_asr_report(conn, raw_text, asr_audio_task)
|
||||
async def handle_voice_stop(self, conn, asr_audio_task: List[bytes]):
|
||||
"""并行处理ASR和声纹识别"""
|
||||
try:
|
||||
total_start_time = time.monotonic()
|
||||
|
||||
# 准备音频数据
|
||||
if conn.audio_format == "pcm":
|
||||
pcm_data = asr_audio_task
|
||||
else:
|
||||
pcm_data = self.decode_opus(asr_audio_task)
|
||||
|
||||
combined_pcm_data = b"".join(pcm_data)
|
||||
|
||||
# 预先准备WAV数据
|
||||
wav_data = None
|
||||
if self.voiceprint_provider and combined_pcm_data:
|
||||
wav_data = self._pcm_to_wav(combined_pcm_data)
|
||||
|
||||
|
||||
# 定义ASR任务
|
||||
def run_asr():
|
||||
start_time = time.monotonic()
|
||||
try:
|
||||
import asyncio
|
||||
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
|
||||
start_time = time.monotonic()
|
||||
try:
|
||||
import asyncio
|
||||
loop = asyncio.new_event_loop()
|
||||
asyncio.set_event_loop(loop)
|
||||
try:
|
||||
result = loop.run_until_complete(
|
||||
self.voiceprint_provider.identify_speaker(wav_data, conn.session_id)
|
||||
)
|
||||
return result
|
||||
finally:
|
||||
loop.close()
|
||||
except Exception as e:
|
||||
logger.bind(tag=TAG).error(f"声纹识别失败: {e}")
|
||||
return None
|
||||
|
||||
# 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)
|
||||
|
||||
if self.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)
|
||||
|
||||
results = {"asr": asr_result, "voiceprint": voiceprint_result}
|
||||
else:
|
||||
asr_result = asr_future.result(timeout=15)
|
||||
results = {"asr": asr_result, "voiceprint": None}
|
||||
|
||||
parallel_execution_time = time.monotonic() - parallel_start_time
|
||||
|
||||
# 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}")
|
||||
|
||||
# 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)
|
||||
|
||||
# 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)
|
||||
|
||||
except Exception as e:
|
||||
logger.bind(tag=TAG).error(f"处理语音停止失败: {e}")
|
||||
import traceback
|
||||
logger.bind(tag=TAG).debug(f"异常详情: {traceback.format_exc()}")
|
||||
|
||||
def _build_enhanced_text(self, text: str, speaker_name: Optional[str]) -> str:
|
||||
"""构建包含说话人信息的文本"""
|
||||
if speaker_name and speaker_name.strip():
|
||||
return json.dumps({
|
||||
"speaker": speaker_name,
|
||||
"content": text
|
||||
}, ensure_ascii=False)
|
||||
else:
|
||||
return text
|
||||
|
||||
def _pcm_to_wav(self, pcm_data: bytes) -> bytes:
|
||||
"""将PCM数据转换为WAV格式"""
|
||||
if len(pcm_data) == 0:
|
||||
logger.bind(tag=TAG).warning("PCM数据为空,无法转换WAV")
|
||||
return b""
|
||||
|
||||
# 确保数据长度是偶数(16位音频)
|
||||
if len(pcm_data) % 2 != 0:
|
||||
pcm_data = pcm_data[:-1]
|
||||
|
||||
# 创建WAV文件头
|
||||
wav_buffer = io.BytesIO()
|
||||
try:
|
||||
with wave.open(wav_buffer, 'wb') as wav_file:
|
||||
wav_file.setnchannels(1) # 单声道
|
||||
wav_file.setsampwidth(2) # 16位
|
||||
wav_file.setframerate(16000) # 16kHz采样率
|
||||
wav_file.writeframes(pcm_data)
|
||||
|
||||
wav_buffer.seek(0)
|
||||
wav_data = wav_buffer.read()
|
||||
|
||||
return wav_data
|
||||
except Exception as e:
|
||||
logger.bind(tag=TAG).error(f"WAV转换失败: {e}")
|
||||
return b""
|
||||
|
||||
def stop_ws_connection(self):
|
||||
pass
|
||||
@@ -113,27 +252,29 @@ class ASRProviderBase(ABC):
|
||||
pass
|
||||
|
||||
@staticmethod
|
||||
def decode_opus(opus_data: List[bytes]) -> bytes:
|
||||
def decode_opus(opus_data: List[bytes]) -> List[bytes]:
|
||||
"""将Opus音频数据解码为PCM数据"""
|
||||
try:
|
||||
decoder = opuslib_next.Decoder(16000, 1) # 16kHz, 单声道
|
||||
decoder = opuslib_next.Decoder(16000, 1)
|
||||
pcm_data = []
|
||||
buffer_size = 960 # 每次处理960个采样点
|
||||
|
||||
for opus_packet in opus_data:
|
||||
buffer_size = 960 # 每次处理960个采样点 (60ms at 16kHz)
|
||||
|
||||
for i, opus_packet in enumerate(opus_data):
|
||||
try:
|
||||
# 使用较小的缓冲区大小进行处理
|
||||
if not opus_packet or len(opus_packet) == 0:
|
||||
continue
|
||||
|
||||
pcm_frame = decoder.decode(opus_packet, buffer_size)
|
||||
if pcm_frame:
|
||||
if pcm_frame and len(pcm_frame) > 0:
|
||||
pcm_data.append(pcm_frame)
|
||||
|
||||
except opuslib_next.OpusError as e:
|
||||
logger.bind(tag=TAG).warning(f"Opus解码错误,跳过当前数据包: {e}")
|
||||
continue
|
||||
logger.bind(tag=TAG).warning(f"Opus解码错误,跳过数据包 {i}: {e}")
|
||||
except Exception as e:
|
||||
logger.bind(tag=TAG).error(f"音频处理错误: {e}", exc_info=True)
|
||||
continue
|
||||
|
||||
logger.bind(tag=TAG).error(f"音频处理错误,数据包 {i}: {e}")
|
||||
|
||||
return pcm_data
|
||||
|
||||
except Exception as e:
|
||||
logger.bind(tag=TAG).error(f"音频解码过程发生错误: {e}", exc_info=True)
|
||||
logger.bind(tag=TAG).error(f"音频解码过程发生错误: {e}")
|
||||
return []
|
||||
|
||||
@@ -2,6 +2,7 @@ import uuid
|
||||
import re
|
||||
from typing import List, Dict
|
||||
from datetime import datetime
|
||||
from config.settings import load_config
|
||||
|
||||
|
||||
class Message:
|
||||
@@ -46,10 +47,9 @@ class Dialogue:
|
||||
dialogue.append({"role": m.role, "content": m.content})
|
||||
|
||||
def get_llm_dialogue(self) -> List[Dict[str, str]]:
|
||||
dialogue = []
|
||||
for m in self.dialogue:
|
||||
self.getMessages(m, dialogue)
|
||||
return dialogue
|
||||
# 直接调用get_llm_dialogue_with_memory,传入None作为memory_str
|
||||
# 这样确保说话人功能在所有调用路径下都生效
|
||||
return self.get_llm_dialogue_with_memory(None)
|
||||
|
||||
def update_system_message(self, new_content: str):
|
||||
"""更新或添加系统消息"""
|
||||
@@ -63,10 +63,7 @@ class Dialogue:
|
||||
def get_llm_dialogue_with_memory(
|
||||
self, memory_str: str = None
|
||||
) -> List[Dict[str, str]]:
|
||||
if memory_str is None or len(memory_str) == 0:
|
||||
return self.get_llm_dialogue()
|
||||
|
||||
# 构建带记忆的对话
|
||||
# 构建对话
|
||||
dialogue = []
|
||||
|
||||
# 添加系统提示和记忆
|
||||
@@ -75,6 +72,33 @@ class Dialogue:
|
||||
)
|
||||
|
||||
if system_message:
|
||||
# 基础系统提示
|
||||
enhanced_system_prompt = system_message.content
|
||||
|
||||
# 添加说话人个性化描述
|
||||
try:
|
||||
config = load_config()
|
||||
voiceprint_config = config.get("voiceprint", {})
|
||||
speakers = voiceprint_config.get("speakers", [])
|
||||
|
||||
if speakers:
|
||||
enhanced_system_prompt += "\n\n<speaker>"
|
||||
for speaker_str in speakers:
|
||||
try:
|
||||
parts = speaker_str.split(",", 2)
|
||||
if len(parts) >= 2:
|
||||
speaker_id = parts[0].strip()
|
||||
name = parts[1].strip()
|
||||
# 如果描述为空,则为""
|
||||
description = parts[2].strip() if len(parts) >= 3 else ""
|
||||
enhanced_system_prompt += f"\n- {name}:{description}"
|
||||
except:
|
||||
pass
|
||||
enhanced_system_prompt += "\n\n</speaker>"
|
||||
except:
|
||||
# 配置读取失败时忽略错误,不影响其他功能
|
||||
pass
|
||||
|
||||
# 使用正则表达式匹配 <memory> 标签,不管中间有什么内容
|
||||
enhanced_system_prompt = re.sub(
|
||||
r"<memory>.*?</memory>",
|
||||
|
||||
@@ -125,4 +125,13 @@ def initialize_asr(config):
|
||||
config["ASR"][select_asr_module],
|
||||
str(config.get("delete_audio", True)).lower() in ("true", "1", "yes"),
|
||||
)
|
||||
|
||||
# 初始化声纹识别功能
|
||||
voiceprint_config = config.get("voiceprint")
|
||||
if voiceprint_config and voiceprint_config.get("url") and voiceprint_config.get("speakers"):
|
||||
new_asr.init_voiceprint(voiceprint_config)
|
||||
logger.bind(tag=TAG).info("ASR模块声纹识别功能已启用")
|
||||
else:
|
||||
logger.bind(tag=TAG).info("ASR模块声纹识别功能已禁用")
|
||||
|
||||
return new_asr
|
||||
|
||||
@@ -0,0 +1,133 @@
|
||||
import asyncio
|
||||
import json
|
||||
import time
|
||||
import aiohttp
|
||||
from urllib.parse import urlparse, parse_qs
|
||||
from typing import Optional, Dict
|
||||
from config.logger import setup_logging
|
||||
|
||||
TAG = __name__
|
||||
logger = setup_logging()
|
||||
|
||||
|
||||
class VoiceprintProvider:
|
||||
"""声纹识别服务提供者"""
|
||||
|
||||
def __init__(self, config: dict):
|
||||
self.original_url = config.get("url", "")
|
||||
self.speakers = config.get("speakers", [])
|
||||
self.speaker_map = self._parse_speakers()
|
||||
|
||||
# 解析API地址和密钥
|
||||
self.api_url = None
|
||||
self.api_key = None
|
||||
self.speaker_ids = []
|
||||
|
||||
if not self.original_url:
|
||||
logger.bind(tag=TAG).warning("声纹识别URL未配置,声纹识别将被禁用")
|
||||
self.enabled = False
|
||||
else:
|
||||
# 解析URL和key
|
||||
parsed_url = urlparse(self.original_url)
|
||||
base_url = f"{parsed_url.scheme}://{parsed_url.netloc}"
|
||||
|
||||
# 从查询参数中提取key
|
||||
query_params = parse_qs(parsed_url.query)
|
||||
self.api_key = query_params.get('key', [''])[0]
|
||||
|
||||
if not self.api_key:
|
||||
logger.bind(tag=TAG).error("URL中未找到key参数,声纹识别将被禁用")
|
||||
self.enabled = False
|
||||
else:
|
||||
# 构造identify接口地址
|
||||
self.api_url = f"{base_url}/voiceprint/identify"
|
||||
|
||||
# 提取speaker_ids
|
||||
for speaker_str in self.speakers:
|
||||
try:
|
||||
parts = speaker_str.split(",", 2)
|
||||
if len(parts) >= 1:
|
||||
speaker_id = parts[0].strip()
|
||||
self.speaker_ids.append(speaker_id)
|
||||
except Exception:
|
||||
continue
|
||||
|
||||
# 检查是否有有效的说话人配置
|
||||
if not self.speaker_ids:
|
||||
logger.bind(tag=TAG).warning("未配置有效的说话人,声纹识别将被禁用")
|
||||
self.enabled = False
|
||||
else:
|
||||
self.enabled = True
|
||||
logger.bind(tag=TAG).info(f"声纹识别已配置: API={self.api_url}, 说话人={len(self.speaker_ids)}个")
|
||||
|
||||
def _parse_speakers(self) -> Dict[str, Dict[str, str]]:
|
||||
"""解析说话人配置"""
|
||||
speaker_map = {}
|
||||
for speaker_str in self.speakers:
|
||||
try:
|
||||
parts = speaker_str.split(",", 2)
|
||||
if len(parts) >= 3:
|
||||
speaker_id, name, description = parts[0].strip(), parts[1].strip(), parts[2].strip()
|
||||
speaker_map[speaker_id] = {
|
||||
"name": name,
|
||||
"description": description
|
||||
}
|
||||
except Exception as e:
|
||||
logger.bind(tag=TAG).warning(f"解析说话人配置失败: {speaker_str}, 错误: {e}")
|
||||
return speaker_map
|
||||
|
||||
async def identify_speaker(self, audio_data: bytes, session_id: str) -> Optional[str]:
|
||||
"""识别说话人"""
|
||||
if not self.enabled or not self.api_url or not self.api_key:
|
||||
logger.bind(tag=TAG).debug("声纹识别功能已禁用或未配置,跳过识别")
|
||||
return None
|
||||
|
||||
try:
|
||||
api_start_time = time.monotonic()
|
||||
|
||||
# 准备请求头
|
||||
headers = {
|
||||
'Authorization': f'Bearer {self.api_key}',
|
||||
'Accept': 'application/json'
|
||||
}
|
||||
|
||||
# 准备multipart/form-data数据
|
||||
data = aiohttp.FormData()
|
||||
data.add_field('speaker_ids', ','.join(self.speaker_ids))
|
||||
data.add_field('file', audio_data, filename='audio.wav', content_type='audio/wav')
|
||||
|
||||
timeout = aiohttp.ClientTimeout(total=10)
|
||||
|
||||
# 网络请求
|
||||
async with aiohttp.ClientSession(timeout=timeout) as session:
|
||||
async with session.post(self.api_url, headers=headers, data=data) as response:
|
||||
|
||||
if response.status == 200:
|
||||
result = await response.json()
|
||||
speaker_id = result.get("speaker_id")
|
||||
score = result.get("score", 0)
|
||||
total_elapsed_time = time.monotonic() - api_start_time
|
||||
|
||||
logger.bind(tag=TAG).info(f"声纹识别耗时: {total_elapsed_time:.3f}s")
|
||||
|
||||
# 置信度检查
|
||||
if score < 0.5:
|
||||
logger.bind(tag=TAG).warning(f"声纹识别置信度较低: {score:.3f}")
|
||||
|
||||
if speaker_id and speaker_id in self.speaker_map:
|
||||
result_name = self.speaker_map[speaker_id]["name"]
|
||||
return result_name
|
||||
else:
|
||||
return None
|
||||
else:
|
||||
logger.bind(tag=TAG).error(f"声纹识别API错误: HTTP {response.status}")
|
||||
return None
|
||||
|
||||
except asyncio.TimeoutError:
|
||||
elapsed = time.monotonic() - api_start_time
|
||||
logger.bind(tag=TAG).error(f"声纹识别超时: {elapsed:.3f}s")
|
||||
return None
|
||||
except Exception as e:
|
||||
elapsed = time.monotonic() - api_start_time
|
||||
logger.bind(tag=TAG).error(f"声纹识别失败: {e}")
|
||||
return None
|
||||
Reference in New Issue
Block a user