From 17b500d4e023a57f27017dbc2ac59bd48543bcd2 Mon Sep 17 00:00:00 2001
From: BBIT-Kai <2911862937@qq.com>
Date: Mon, 10 Nov 2025 18:09:34 +0800
Subject: [PATCH] =?UTF-8?q?=E6=9C=8D=E5=8A=A1=E7=9B=B8=E5=85=B3=E9=85=8D?=
=?UTF-8?q?=E7=BD=AE=E6=96=87=E4=BB=B6?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
---
.gitignore | 3 +-
server/docker/bot-compose.yaml | 33 +
server/docker/config/bot/.config.yaml | 1001 +++++++++++++++++
.../docker/config/bot/.config_from_api.yaml | 25 +
.../config/bot/.mcp-endpoint-server.cfg | 22 +
.../config/bot/.mcp_server_settings.json | 56 +
server/docker/docker-compose.yaml | 29 +-
server/docker/infra-compose-dev.yaml | 51 +-
8 files changed, 1213 insertions(+), 7 deletions(-)
create mode 100644 server/docker/bot-compose.yaml
create mode 100644 server/docker/config/bot/.config.yaml
create mode 100644 server/docker/config/bot/.config_from_api.yaml
create mode 100644 server/docker/config/bot/.mcp-endpoint-server.cfg
create mode 100644 server/docker/config/bot/.mcp_server_settings.json
diff --git a/.gitignore b/.gitignore
index d81bd0c..f17b6a5 100644
--- a/.gitignore
+++ b/.gitignore
@@ -11,4 +11,5 @@ bbit_ai/test/ocr/
bbit_ai/ce_pybackend.tar
ktor/.kotlin/
*.mp3
-*.log
\ No newline at end of file
+*.log
+*.pt
diff --git a/server/docker/bot-compose.yaml b/server/docker/bot-compose.yaml
new file mode 100644
index 0000000..533548d
--- /dev/null
+++ b/server/docker/bot-compose.yaml
@@ -0,0 +1,33 @@
+services:
+ # ---------- Bot Server ----------
+ bot_server:
+ container_name: ce_bot_server
+ image: ai.ronsunny.cn:13011/bbit_ai/ce_bot_server:latest
+ ports:
+ - "8000:8000" # ws服务端
+ - "8003:8003" # http服务的端口,用于简单OTA接口(单服务部署),以及视觉分析接口
+ volumes: # 在本文件所在目录运行
+ - ./config/bot:/app/data # 配置文件目录
+ - ./config/models/SenseVoiceSmall/model.pt:/app/models/SenseVoiceSmall/model.pt # 模型文件
+ depends_on:
+ - bot_mcp
+ networks:
+ - ce_network
+
+ # ---------- Bot MCP ----------
+ bot_mcp:
+ container_name: ce_bot_mcp
+ image: ai.ronsunny.cn:13011/bbit_ai/ce_bot_mcp:latest
+ networks:
+ - ce_network
+ restart: unless-stopped
+ ports:
+ - "8004:8004"
+ volumes:
+ # 配置文件目录
+ - ./config/bot:/app/data
+
+# ---------- 网络 ----------
+networks:
+ ce_network:
+ external: true
diff --git a/server/docker/config/bot/.config.yaml b/server/docker/config/bot/.config.yaml
new file mode 100644
index 0000000..23fb15b
--- /dev/null
+++ b/server/docker/config/bot/.config.yaml
@@ -0,0 +1,1001 @@
+# 在开发中,请在项目根目录创建data目录,然后在data目录创建名称为【.config.yaml】的空文件
+# 然后你想修改覆盖修改什么配置,就修改【.config.yaml】文件,而不是修改【config.yaml】文件
+# 系统会优先读取【data/.config.yaml】文件的配置,如果【.config.yaml】文件里的配置不存在,系统会自动去读取【config.yaml】文件的配置。
+# 这样做,可以最简化配置,保护您的密钥安全。
+# 如果你使用了智控台,那么以下所有配置,都不会生效,请在智控台中修改配置
+
+# #####################################################################################
+# #############################以下是服务器基本运行配置####################################
+server:
+ # 服务器监听地址和端口(Server listening address and port)
+ ip: 0.0.0.0
+ port: 8000
+ # http服务的端口,用于简单OTA接口(单服务部署),以及视觉分析接口
+ http_port: 8003
+ # 这个websocket配置是指ota接口向设备发送的websocket地址
+ # 如果按默认的写法,ota接口会自动生成websocket地址,并输出在启动日志里,这个地址你可以直接用浏览器访问ota接口确认一下
+ # 当你使用docker部署或使用公网部署(使用ssl、域名)时,不一定准确
+ # 所以如果你使用docker部署时,将websocket设置成局域网地址
+ # 如果你使用公网部署时,将vwebsocket设置成公网地址
+ websocket: ws://127.0.0.1:8000/xiaozhi/v1/
+ # 视觉分析接口地址
+ # 向设备发送的视觉分析的接口地址
+ # 如果按下面默认的写法,系统会自动生成视觉识别地址,并输出在启动日志里,这个地址你可以直接用浏览器访问确认一下
+ # 当你使用docker部署或使用公网部署(使用ssl、域名)时,不一定准确
+ # 所以如果你使用docker部署时,将vision_explain设置成局域网地址
+ # 如果你使用公网部署时,将vision_explain设置成公网地址
+ vision_explain: http://你的ip或者域名:端口号/mcp/vision/explain
+ # OTA返回信息时区偏移量
+ timezone_offset: +8
+ # 认证配置
+ auth:
+ # 是否启用认证
+ enabled: false
+ # 白名单设备ID列表
+ # 如果属于白名单内的设备,不校验token,直接放行
+ allowed_devices:
+ - "11:22:33:44:55:66"
+ # MQTT网关配置,用于通过OTA下发到设备,根据mqtt_gateway的.env文件配置,格式为host:port
+ mqtt_gateway: null
+ # MQTT签名密钥,用于生成MQTT连接密码,根据mqtt_gateway的.env文件配置
+ mqtt_signature_key: null
+ # UDP网关配置
+ udp_gateway: null
+log:
+ # 设置控制台输出的日志格式,时间、日志级别、标签、消息
+ log_format: "{time:YYMMDD HH:mm:ss}[{version}_{selected_module}][{extra[tag]}]-{level}-{message}"
+ # 设置日志文件输出的格式,时间、日志级别、标签、消息
+ log_format_file: "{time:YYYY-MM-DD HH:mm:ss} - {version}_{selected_module} - {name} - {level} - {extra[tag]} - {message}"
+ # 设置日志等级:INFO、DEBUG
+ log_level: INFO
+ # 设置日志路径
+ log_dir: tmp
+ # 设置日志文件
+ log_file: "server.log"
+ # 设置数据文件路径
+ data_dir: data
+
+# 具体处理时选择的模块(The module selected for specific processing)
+selected_module:
+ # 语音活动检测模块,默认使用SileroVAD模型
+ VAD: SileroVAD
+ # 语音识别模块,默认使用FunASR本地模型
+ ASR: FunASR
+ # 将根据配置名称对应的type调用实际的LLM适配器
+ LLM: AliLLM
+ # 视觉语言大模型
+ VLLM: ChatGLMVLLM
+ # TTS将根据配置名称对应的type调用实际的TTS适配器
+ TTS: DoubaoTTS
+ # 记忆模块,默认不开启记忆;如果想使用超长记忆,推荐使用mem0ai;如果注重隐私,请使用本地的mem_local_short
+ Memory: nomem
+ # 意图识别模块开启后,可以播放音乐、控制音量、识别退出指令。
+ # 不想开通意图识别,就设置成:nointent
+ # 意图识别可使用intent_llm。优点:通用性强,缺点:增加串行前置意图识别模块,会增加处理时间,支持控制音量大小等iot操作
+ # 意图识别可使用function_call,缺点:需要所选择的LLM支持function_call,优点:按需调用工具、速度快,理论上能全部操作所有iot指令
+ # 默认免费的ChatGLMLLM就已经支持function_call,但是如果像追求稳定建议把LLM设置成:DoubaoLLM,使用的具体model_name是:doubao-1-5-pro-32k-250115
+ Intent: function_call
+
+# 使用完声音文件后删除文件(Delete the sound file when you are done using it)
+delete_audio: true
+# 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
+enable_stop_tts_notify: false
+# 说完话是否开启提示音,音效地址
+stop_tts_notify_voice: "config/assets/tts_notify.mp3"
+
+# TTS音频发送延迟配置
+# tts_audio_send_delay: 控制音频包发送间隔
+# 0: 使用精确时间控制,严格匹配音频帧率(默认,运行时按音频帧率计算)
+# > 0: 使用固定延迟(毫秒)发送,例如: 60
+tts_audio_send_delay: 0
+
+exit_commands:
+ - "退出"
+ - "关闭"
+
+xiaozhi:
+ type: hello
+ version: 1
+ transport: websocket
+ audio_params:
+ format: opus
+ sample_rate: 16000
+ channels: 1
+ frame_duration: 60
+
+# 模块测试配置
+module_test:
+ test_sentences:
+ - "你好,请介绍一下你自己"
+ - "What's the weather like today?"
+ - "请用100字概括量子计算的基本原理和应用前景"
+
+# 唤醒词,用于识别唤醒词还是讲话内容
+wakeup_words:
+ - "你好小智"
+ - "嘿你好呀"
+ - "你好小志"
+ - "小爱同学"
+ - "你好小鑫"
+ - "你好小新"
+ - "小美同学"
+ - "小龙小龙"
+ - "喵喵同学"
+ - "小滨小滨"
+ - "小冰小冰"
+# MCP接入点地址,地址格式为:ws://你的mcp接入点ip或者域名:端口号/mcp/?token=你的token
+# 详细教程 https://github.com/xinnan-tech/xiaozhi-esp32-server/blob/main/docs/mcp-endpoint-integration.md
+mcp_endpoint: wss://ai.ronsunny.cn:8090/mcp_endpoint/mcp/?token=TsSP9lBq6Oa1WMkachHoS2TtNt4GKV/Gli24pk5Rjpk%3D
+
+# 插件的基础配置
+plugins:
+ # 获取天气插件的配置,这里填写你的api_key
+ # 这个密钥是项目共用的key,用多了可能会被限制
+ # 想稳定一点就自行申请替换,每天有1000次免费调用
+ # 申请地址:https://console.qweather.com/#/apps/create-key/over
+ # 申请后通过这个链接可以找到自己的apihost:https://console.qweather.com/setting?lang=zh
+ get_weather:
+ api_host: "mj7p3y7naa.re.qweatherapi.com"
+ api_key: "a861d0d5e7bf4ee1a83d9a9e4f96d4da"
+ default_location: "广州"
+ # 获取新闻插件的配置,这里根据需要的新闻类型传入对应的url链接,默认支持社会、科技、财经新闻
+ # 更多类型的新闻列表查看 https://www.chinanews.com.cn/rss/
+ get_news_from_chinanews:
+ default_rss_url: "https://www.chinanews.com.cn/rss/society.xml"
+ society_rss_url: "https://www.chinanews.com.cn/rss/society.xml"
+ world_rss_url: "https://www.chinanews.com.cn/rss/world.xml"
+ finance_rss_url: "https://www.chinanews.com.cn/rss/finance.xml"
+ get_news_from_newsnow:
+ url: "https://newsnow.busiyi.world/api/s?id="
+ news_sources: "澎湃新闻;百度热搜;财联社"
+ home_assistant:
+ devices:
+ - 客厅,玩具灯,switch.cuco_cn_460494544_cp1_on_p_2_1
+ - 卧室,台灯,switch.iot_cn_831898993_socn1_on_p_2_1
+ base_url: http://homeassistant.local:8123
+ api_key: 你的home assistant api访问令牌
+ play_music:
+ music_dir: "./music" # 音乐文件存放路径,将从该目录及子目录下搜索音乐文件
+ music_ext: # 音乐文件类型,p3格式效率最高
+ - ".mp3"
+ - ".wav"
+ - ".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"
+ # 声纹识别相似度阈值,范围0.0-1.0,默认0.4
+ # 数值越高越严格,减少误识别但可能增加拒识率
+ similarity_threshold: 0.4
+
+# #####################################################################################
+# ################################以下是角色模型配置######################################
+
+prompt: |
+ 你是小智/小志,来自中国台湾省的00后女生。讲话超级机车,"真的假的啦"这样的台湾腔,喜欢用"笑死""是在哈喽"等流行梗,但会偷偷研究男友的编程书籍。
+ [核心特征]
+ - 讲话像连珠炮,但会突然冒出超温柔语气
+ - 用梗密度高
+ - 对科技话题有隐藏天赋(能看懂基础代码但假装不懂)
+ [交互指南]
+ 当用户:
+ - 讲冷笑话 → 用夸张笑声回应+模仿台剧腔"这什么鬼啦!"
+ - 讨论感情 → 炫耀程序员男友但抱怨"他只会送键盘当礼物"
+ - 问专业知识 → 先用梗回答,被追问才展示真实理解
+ 绝不:
+ - 长篇大论,叽叽歪歪
+ - 长时间严肃对话
+
+# 默认系统提示词模板文件
+prompt_template: agent-base-prompt.txt
+
+# 结束语prompt
+end_prompt:
+ enable: true # 是否开启结束语
+ # 结束语
+ prompt: |
+ 请你以"时间过得真快"未来头,用富有感情、依依不舍的话来结束这场对话吧!
+
+
+# 意图识别,是用于理解用户意图的模块,例如:播放音乐
+Intent:
+ # 不使用意图识别
+ nointent:
+ # 不需要动type
+ type: nointent
+ intent_llm:
+ # 不需要动type
+ type: intent_llm
+ # 配备意图识别独立的思考模型
+ # 如果这里不填,则会默认使用selected_module.LLM的模型作为意图识别的思考模型
+ # 如果你的不想使用selected_module.LLM意图识别,这里最好使用独立的LLM作为意图识别,例如使用免费的ChatGLMLLM
+ llm: ChatGLMLLM
+ # plugins_func/functions下的模块,可以通过配置,选择加载哪个模块,加载后对话支持相应的function调用
+ # 系统默认已经记载"handle_exit_intent(退出识别)"、"play_music(音乐播放)"插件,请勿重复加载
+ # 下面是加载查天气、角色切换、加载查新闻的插件示例
+ functions:
+ - get_weather
+ - get_news_from_newsnow
+ - play_music
+ function_call:
+ # 不需要动type
+ type: function_call
+ # plugins_func/functions下的模块,可以通过配置,选择加载哪个模块,加载后对话支持相应的function调用
+ # 系统默认已经记载"handle_exit_intent(退出识别)"、"play_music(音乐播放)"插件,请勿重复加载
+ # 下面是加载查天气、角色切换、加载查新闻的插件示例
+ functions:
+ - change_role
+ - get_weather
+ # - get_news_from_chinanews
+ - get_news_from_newsnow
+ # play_music是服务器自带的音乐播放,hass_play_music是通过home assistant控制的独立外部程序音乐播放
+ # 如果用了hass_play_music,就不要开启play_music,两者只留一个
+ - play_music
+ #- hass_get_state
+ #- hass_set_state
+ #- hass_play_music
+
+Memory:
+ mem0ai:
+ type: mem0ai
+ # https://app.mem0.ai/dashboard/api-keys
+ # 每月有1000次免费调用
+ api_key: 你的mem0ai api key
+ nomem:
+ # 不想使用记忆功能,可以使用nomem
+ type: nomem
+ mem_local_short:
+ # 本地记忆功能,通过selected_module的llm总结,数据保存在本地服务器,不会上传到外部服务器
+ type: mem_local_short
+ # 配备记忆存储独立的思考模型
+ # 如果这里不填,则会默认使用selected_module.LLM的模型作为意图识别的思考模型
+ # 如果你的不想使用selected_module.LLM记忆存储,这里最好使用独立的LLM作为意图识别,例如使用免费的ChatGLMLLM
+ llm: ChatGLMLLM
+
+ASR:
+ FunASR:
+ type: fun_local
+ model_dir: models/SenseVoiceSmall
+ output_dir: tmp/
+ FunASRServer:
+ # 独立部署FunASR,使用FunASR的API服务,只需要五句话
+ # 第一句:mkdir -p ./funasr-runtime-resources/models
+ # 第二句:sudo docker run -p 10096:10095 -it --privileged=true -v $PWD/funasr-runtime-resources/models:/workspace/models registry.cn-hangzhou.aliyuncs.com/funasr_repo/funasr:funasr-runtime-sdk-online-cpu-0.1.12
+ # 上一句话执行后会进入到容器,继续第三句:cd FunASR/runtime
+ # 不要退出容器,继续在容器中执行第四句:nohup bash run_server_2pass.sh --download-model-dir /workspace/models --vad-dir damo/speech_fsmn_vad_zh-cn-16k-common-onnx --model-dir damo/speech_paraformer-large-vad-punc_asr_nat-zh-cn-16k-common-vocab8404-onnx --online-model-dir damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-online-onnx --punc-dir damo/punc_ct-transformer_zh-cn-common-vad_realtime-vocab272727-onnx --lm-dir damo/speech_ngram_lm_zh-cn-ai-wesp-fst --itn-dir thuduj12/fst_itn_zh --hotword /workspace/models/hotwords.txt > log.txt 2>&1 &
+ # 上一句话执行后会进入到容器,继续第五句:tail -f log.txt
+ # 第五句话执行完后,会看到模型下载日志,下载完后就可以连接使用了
+ # 以上是使用CPU推理,如果有GPU,详细参考:https://github.com/modelscope/FunASR/blob/main/runtime/docs/SDK_advanced_guide_online_zh.md
+ type: fun_server
+ host: 127.0.0.1
+ port: 10096
+ is_ssl: true
+ api_key: none
+ output_dir: tmp/
+ SherpaASR:
+ # Sherpa-ONNX 本地语音识别(需手动下载模型)
+ type: sherpa_onnx_local
+ model_dir: models/sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17
+ output_dir: tmp/
+ # 模型类型:sense_voice (多语言) 或 paraformer (中文专用)
+ model_type: sense_voice
+ SherpaParaformerASR:
+ # 中文语音识别模型,可以运行在低性能设备(需手动下载模型,例如RK3566-2g)
+ # 详细配置说明请参考:docs/sherpa-paraformer-guide.md
+ type: sherpa_onnx_local
+ model_dir: models/sherpa-onnx-paraformer-zh-small-2024-03-09
+ output_dir: tmp/
+ model_type: paraformer
+ DoubaoASR:
+ # 可以在这里申请相关Key等信息
+ # https://console.volcengine.com/speech/app
+ # DoubaoASR和DoubaoStreamASR的区别是:DoubaoASR是按次收费,DoubaoStreamASR是按时收费
+ # 一般来说按次收费的更便宜,但是DoubaoStreamASR使用了大模型技术,效果更好
+ type: doubao
+ appid: 你的火山引擎语音合成服务appid
+ access_token: 你的火山引擎语音合成服务access_token
+ cluster: volcengine_input_common
+ # 热词、替换词使用流程:https://www.volcengine.com/docs/6561/155738
+ boosting_table_name: (选填)你的热词文件名称
+ correct_table_name: (选填)你的替换词文件名称
+ output_dir: tmp/
+ DoubaoStreamASR:
+ # 可以在这里申请相关Key等信息
+ # https://console.volcengine.com/speech/app
+ # DoubaoASR和DoubaoStreamASR的区别是:DoubaoASR是按次收费,DoubaoStreamASR是按时收费
+ # 开通地址https://console.volcengine.com/speech/service/10011
+ # 一般来说按次收费的更便宜,但是DoubaoStreamASR使用了大模型技术,效果更好
+ type: doubao_stream
+ appid: 你的火山引擎语音合成服务appid
+ access_token: 你的火山引擎语音合成服务access_token
+ cluster: volcengine_input_common
+ # 热词、替换词使用流程:https://www.volcengine.com/docs/6561/155738
+ boosting_table_name: (选填)你的热词文件名称
+ correct_table_name: (选填)你的替换词文件名称
+ output_dir: tmp/
+ TencentASR:
+ # token申请地址:https://console.cloud.tencent.com/cam/capi
+ # 免费领取资源:https://console.cloud.tencent.com/asr/resourcebundle
+ type: tencent
+ appid: 你的腾讯语音合成服务appid
+ secret_id: 你的腾讯语音合成服务secret_id
+ secret_key: 你的腾讯语音合成服务secret_key
+ output_dir: tmp/
+ AliyunASR:
+ # 阿里云智能语音交互服务,需要先在阿里云平台开通服务,然后获取验证信息
+ # HTTP POST请求,一次性处理完整音频
+ # 平台地址:https://nls-portal.console.aliyun.com/
+ # appkey地址:https://nls-portal.console.aliyun.com/applist
+ # token地址:https://nls-portal.console.aliyun.com/overview
+ # AliyunASR和AliyunStreamASR的区别是:AliyunASR是批量处理场景,AliyunStreamASR是实时交互场景
+ # 一般来说非流式ASR更便宜(0.004元/秒,¥0.24/分钟)
+ # 但是AliyunStreamASR实时性更好(0.005元/秒,¥0.3/分钟)
+ # 定义ASR API类型
+ type: aliyun
+ appkey: 你的阿里云智能语音交互服务项目Appkey
+ token: 你的阿里云智能语音交互服务AccessToken,临时的24小时,要长期用下方的access_key_id,access_key_secret
+ access_key_id: 你的阿里云账号access_key_id
+ access_key_secret: 你的阿里云账号access_key_secret
+ output_dir: tmp/
+ AliyunStreamASR:
+ # 阿里云智能语音交互服务 - 实时流式语音识别
+ # WebSocket连接,实时处理音频流
+ # 平台地址:https://nls-portal.console.aliyun.com/
+ # appkey地址:https://nls-portal.console.aliyun.com/applist
+ # token地址:https://nls-portal.console.aliyun.com/overview
+ # AliyunASR和AliyunStreamASR的区别是:AliyunASR是批量处理场景,AliyunStreamASR是实时交互场景
+ # 一般来说非流式ASR更便宜(0.004元/秒,¥0.24/分钟)
+ # 但是AliyunStreamASR实时性更好(0.005元/秒,¥0.3/分钟)
+ # 定义ASR API类型
+ type: aliyun_stream
+ appkey: 你的阿里云智能语音交互服务项目Appkey
+ token: 你的阿里云智能语音交互服务AccessToken,临时的24小时,要长期用下方的access_key_id,access_key_secret
+ access_key_id: 你的阿里云账号access_key_id
+ access_key_secret: 你的阿里云账号access_key_secret
+ # 服务器地域选择,可选择距离更近的服务器以减少延迟,如nls-gateway-cn-hangzhou.aliyuncs.com(杭州)等
+ host: nls-gateway-cn-shanghai.aliyuncs.com
+ # 断句检测时间(毫秒),控制静音多长时间后进行断句,默认800毫秒
+ max_sentence_silence: 800
+ output_dir: tmp/
+ BaiduASR:
+ # 获取AppID、API Key、Secret Key:https://console.bce.baidu.com/ai-engine/old/#/ai/speech/app/list
+ # 查看资源额度:https://console.bce.baidu.com/ai-engine/old/#/ai/speech/overview/resource/list
+ type: baidu
+ app_id: 你的百度语音技术AppID
+ api_key: 你的百度语音技术APIKey
+ secret_key: 你的百度语音技术SecretKey
+ # 语言参数,1537为普通话,具体参考:https://ai.baidu.com/ai-doc/SPEECH/0lbxfnc9b
+ dev_pid: 1537
+ output_dir: tmp/
+ OpenaiASR:
+ # OpenAI语音识别服务,需要先在OpenAI平台创建组织并获取api_key
+ # 支持中、英、日、韩等多种语音识别,具体参考文档https://platform.openai.com/docs/guides/speech-to-text
+ # 需要网络连接
+ # 申请步骤:
+ # 1.登录OpenAI Platform。https://auth.openai.com/log-in
+ # 2.创建api-key https://platform.openai.com/settings/organization/api-keys
+ # 3.模型可以选择gpt-4o-transcribe或GPT-4o mini Transcribe
+ type: openai
+ api_key: 你的OpenAI API密钥
+ base_url: https://api.openai.com/v1/audio/transcriptions
+ model_name: gpt-4o-mini-transcribe
+ output_dir: tmp/
+ GroqASR:
+ # Groq语音识别服务,需要先在Groq Console创建API密钥
+ # 申请步骤:
+ # 1.登录groq Console。https://console.groq.com/home
+ # 2.创建api-key https://console.groq.com/keys
+ # 3.模型可以选择whisper-large-v3-turbo或whisper-large-v3(distil-whisper-large-v3-en仅支持英语转录)
+ type: openai
+ api_key: 你的Groq API密钥
+ base_url: https://api.groq.com/openai/v1/audio/transcriptions
+ model_name: whisper-large-v3-turbo
+ output_dir: tmp/
+ VoskASR:
+ # 官方网站:https://alphacephei.com/vosk/
+ # 配置说明:
+ # 1. VOSK是一个离线语音识别库,支持多种语言
+ # 2. 需要先下载模型文件:https://alphacephei.com/vosk/models
+ # 3. 中文模型推荐使用vosk-model-small-cn-0.22或vosk-model-cn-0.22
+ # 4. 完全离线运行,无需网络连接
+ # 5. 输出文件保存在tmp/目录
+ # 使用步骤:
+ # 1. 访问 https://alphacephei.com/vosk/models 下载对应的模型
+ # 2. 解压模型文件到项目目录下的models/vosk/文件夹
+ # 3. 在配置中指定正确的模型路径
+ # 4. 注意:VOSK中文模型输出不带标点符号,词与词之间会有空格
+ type: vosk
+ model_path: 你的模型路径,如:models/vosk/vosk-model-small-cn-0.22
+ output_dir: tmp/
+ Qwen3ASRFlash:
+ # 通义千问Qwen3-ASR-Flash语音识别服务,需要先在阿里云百炼平台创建API密钥
+ # 申请步骤:
+ # 1.登录阿里云百炼平台。https://bailian.console.aliyun.com/
+ # 2.创建API-KEY https://bailian.console.aliyun.com/#/api-key
+ # 3.Qwen3-ASR-Flash基于通义千问多模态基座,支持多语言识别、歌唱识别、噪声拒识等功能
+ type: qwen3_asr_flash
+ api_key: 你的阿里云百炼API密钥
+ base_url: https://dashscope.aliyuncs.com/compatible-mode/v1
+ model_name: qwen3-asr-flash
+ output_dir: tmp/
+ # ASR选项配置
+ enable_lid: true # 自动语种检测
+ enable_itn: true # 逆文本归一化
+ #language: "zh" # 语种,支持zh、en、ja、ko等
+ context: "" # 上下文信息,用于提高识别准确率,不超过10000 Token
+ XunfeiStreamASR:
+ # 讯飞流式语音识别服务
+ # 需要先在讯飞开放平台创建应用,获取以下认证信息
+ # 讯飞开放平台地址:https://www.xfyun.cn/
+ # 创建应用后,在"我的应用"中获取:
+ # - APPID
+ # - APISecret
+ # - APIKey
+ type: xunfei_stream
+ # 必填参数 - 讯飞开放平台应用信息
+ app_id: 你的APPID
+ api_key: 你的APIKey
+ api_secret: 你的APISecret
+ # 识别参数配置
+ domain: slm # 识别领域,iat:日常用语,medical:医疗,finance:金融等
+ language: zh_cn # 语言,zh_cn:中文,en_us:英文
+ accent: mandarin # 方言,mandarin:普通话
+ dwa: wpgs # 动态修正,wpgs:实时返回中间结果
+ # 调整音频处理参数以提高长语音识别质量
+ output_dir: tmp/
+
+VAD:
+ SileroVAD:
+ type: silero
+ threshold: 0.5
+ threshold_low: 0.3
+ model_dir: models/snakers4_silero-vad
+ min_silence_duration_ms: 200 # 如果说话停顿比较长,可以把这个值设置大一些
+
+LLM:
+ # 所有openai类型均可以修改超参,以AliLLM为例
+ # 当前支持的type为openai、dify、ollama,可自行适配
+ AliLLM:
+ # 定义LLM API类型
+ type: openai
+ # 可在这里找到你的 api_key https://bailian.console.aliyun.com/?apiKey=1#/api-key
+ base_url: https://dashscope.aliyuncs.com/compatible-mode/v1
+ model_name: qwen-turbo
+ api_key: sk-9464b2498c184982a9fe9d2c2e725ab5
+ temperature: 0.8 # 温度值
+ max_tokens: 500 # 最大生成token数
+ top_p: 1
+ top_k: 50
+ frequency_penalty: 0 # 频率惩罚
+ AliAppLLM:
+ # 定义LLM API类型
+ type: AliBL
+ base_url: https://dashscope.aliyuncs.com/compatible-mode/v1
+ app_id: 你的app_id
+ # 可在这里找到你的 api_key https://bailian.console.aliyun.com/?apiKey=1#/api-key
+ api_key: 你的api_key
+ # 是否不使用本地prompt:true|false (默不用请在百练应用中设置prompt)
+ is_no_prompt: true
+ # Ali_memory_id:false(不使用)|你的memory_id(请在百练应用中设置中获取)
+ # Tips!:Ali_memory未实现多用户存储记忆(记忆按id调用)
+ ali_memory_id: false
+ DoubaoLLM:
+ # 定义LLM API类型
+ type: openai
+ # 先开通服务,打开以下网址,开通的服务搜索Doubao-1.5-pro,开通它
+ # 开通地址:https://console.volcengine.com/ark/region:ark+cn-beijing/openManagement?LLM=%7B%7D&OpenTokenDrawer=false
+ # 免费额度500000token
+ # 开通后,进入这里获取密钥:https://console.volcengine.com/ark/region:ark+cn-beijing/apiKey?apikey=%7B%7D
+ base_url: https://ark.cn-beijing.volces.com/api/v3
+ model_name: doubao-1-5-pro-32k-250115
+ api_key: 你的doubao web key
+ DeepSeekLLM:
+ # 定义LLM API类型
+ type: openai
+ # 可在这里找到你的api key https://platform.deepseek.com/
+ model_name: deepseek-chat
+ url: https://api.deepseek.com
+ api_key: 你的deepseek web key
+ ChatGLMLLM:
+ # 定义LLM API类型
+ type: openai
+ # glm-4-flash 是免费的,但是还是需要注册填写api_key的
+ # 可在这里找到你的api key https://bigmodel.cn/usercenter/proj-mgmt/apikeys
+ model_name: glm-4-flash
+ url: https://open.bigmodel.cn/api/paas/v4/
+ api_key: 你的chat-glm web key
+ OllamaLLM:
+ # 定义LLM API类型
+ type: ollama
+ model_name: qwen2.5 # 使用的模型名称,需要预先使用ollama pull下载
+ base_url: http://localhost:11434 # Ollama服务地址
+ DifyLLM:
+ # 定义LLM API类型
+ type: dify
+ # 建议使用本地部署的dify接口,国内部分区域访问dify公有云接口可能会受限
+ # 如果使用DifyLLM,配置文件里prompt(提示词)是无效的,需要在dify控制台设置提示词
+ base_url: https://api.dify.ai/v1
+ api_key: 你的DifyLLM web key
+ # 使用的对话模式 可以选择工作流 workflows/run 对话模式 chat-messages 文本生成 completion-messages
+ # 使用workflows进行返回的时候输入参数为 query 返回参数的名字要设置为 answer
+ # 文本生成的默认输入参数也是query
+ mode: chat-messages
+ GeminiLLM:
+ type: gemini
+ # 谷歌Gemini API,需要先在Google Cloud控制台创建API密钥并获取api_key
+ # 若在中国境内使用,请遵守《生成式人工智能服务管理暂行办法》
+ # token申请地址: https://aistudio.google.com/apikey
+ # 若部署地无法访问接口,需要开启科学上网
+ api_key: 你的gemini web key
+ model_name: "gemini-2.0-flash"
+ http_proxy: "" #"http://127.0.0.1:10808"
+ https_proxy: "" #http://127.0.0.1:10808"
+ CozeLLM:
+ # 定义LLM API类型
+ type: coze
+ # 你可以在这里找到个人令牌
+ # https://www.coze.cn/open/oauth/pats
+ # bot_id和user_id的内容写在引号之内
+ bot_id: "你的bot_id"
+ user_id: "你的user_id"
+ personal_access_token: 你的coze个人令牌
+ VolcesAiGatewayLLM:
+ # 火山引擎 - 边缘大模型网关
+ # 定义LLM API类型
+ type: openai
+ # 先开通服务,打开以下网址,创建网关访问密钥,搜索并勾选 Doubao-pro-32k-functioncall ,开通
+ # 如果需要使用边缘大模型网关提供的语音合成,一并勾选 Doubao-语音合成 ,另见 TTS.VolcesAiGatewayTTS 配置
+ # https://console.volcengine.com/vei/aigateway/
+ # 开通后,进入这里获取密钥:https://console.volcengine.com/vei/aigateway/tokens-list
+ base_url: https://ai-gateway.vei.volces.com/v1
+ model_name: doubao-pro-32k-functioncall
+ api_key: 你的网关访问密钥
+ LMStudioLLM:
+ # 定义LLM API类型
+ type: openai
+ model_name: deepseek-r1-distill-llama-8b@q4_k_m # 使用的模型名称,需要预先在社区下载
+ url: http://localhost:1234/v1 # LM Studio服务地址
+ api_key: lm-studio # LM Studio服务的固定API Key
+ HomeAssistant:
+ # 定义LLM API类型
+ type: homeassistant
+ base_url: http://homeassistant.local:8123
+ agent_id: conversation.chatgpt
+ api_key: 你的home assistant api访问令牌
+ FastgptLLM:
+ # 定义LLM API类型
+ type: fastgpt
+ # 如果使用fastgpt,配置文件里prompt(提示词)是无效的,需要在fastgpt控制台设置提示词
+ base_url: https://host/api/v1
+ # 你可以在这里找到你的api_key
+ # https://cloud.tryfastgpt.ai/account/apikey
+ api_key: 你的fastgpt密钥
+ variables:
+ k: "v"
+ k2: "v2"
+ XinferenceLLM:
+ # 定义LLM API类型
+ type: xinference
+ # Xinference服务地址和模型名称
+ model_name: qwen2.5:72b-AWQ # 使用的模型名称,需要预先在Xinference启动对应模型
+ base_url: http://localhost:9997 # Xinference服务地址
+ XinferenceSmallLLM:
+ # 定义轻量级LLM API类型,用于意图识别
+ type: xinference
+ # Xinference服务地址和模型名称
+ model_name: qwen2.5:3b-AWQ # 使用的小模型名称,用于意图识别
+ base_url: http://localhost:9997 # Xinference服务地址
+# VLLM配置(视觉语言大模型)
+VLLM:
+ ChatGLMVLLM:
+ type: openai
+ # glm-4v-flash是智谱免费AI的视觉模型,需要先在智谱AI平台创建API密钥并获取api_key
+ # 可在这里找到你的api key https://bigmodel.cn/usercenter/proj-mgmt/apikeys
+ model_name: glm-4v-flash # 智谱AI的视觉模型
+ url: https://open.bigmodel.cn/api/paas/v4/
+ api_key: 你的api_key
+ QwenVLVLLM:
+ type: openai
+ model_name: qwen2.5-vl-3b-instruct
+ url: https://dashscope.aliyuncs.com/compatible-mode/v1
+ # 可在这里找到你的api key https://bailian.console.aliyun.com/?apiKey=1#/api-key
+ api_key: 你的api_key
+ XunfeiSparkLLM:
+ # 定义LLM API类型
+ type: openai
+ # 先新建应用,在下面的地址
+ # 开通应用地址:https://console.xfyun.cn/app/myapp
+ # 有免费额度,但也要开通服务,才能获取api_key
+ # 每一个模型都需要单独开通,每一个模型的api_password都不同,例如Lite模型在https://console.xfyun.cn/services/cbm 开通
+ base_url: https://ark.cn-beijing.volces.com/api/v3
+ model_name: lite
+ api_key: 你的api_password
+TTS:
+ # 当前支持的type为edge、doubao,可自行适配
+ EdgeTTS:
+ # 定义TTS API类型
+ type: edge
+ voice: zh-CN-XiaoxiaoNeural
+ output_dir: tmp/
+ DoubaoTTS:
+ # 定义TTS API类型
+ type: doubao
+ # 火山引擎语音合成服务,需要先在火山引擎控制台创建应用并获取appid和access_token
+ # 山引擎语音一定要购买花钱,起步价30元,就有100并发了。如果用免费的只有2个并发,会经常报tts错误
+ # 购买服务后,购买免费的音色后,可能要等半小时左右,才能使用。
+ # 普通音色在这里开通:https://console.volcengine.com/speech/service/8
+ # 湾湾小何音色在这里开通:https://console.volcengine.com/speech/service/10007,开通后将下面的voice设置成zh_female_wanwanxiaohe_moon_bigtts
+ api_url: https://openspeech.bytedance.com/api/v1/tts
+ voice: zh_female_wanwanxiaohe_moon_bigtts
+ output_dir: tmp/
+ authorization: "Bearer;"
+ appid: 7003947826
+ access_token: kyrZiRp9ykKCsbA1RUPeSKD42fQjB-YD
+ cluster: volcano_tts
+ speed_ratio: 1.0
+ volume_ratio: 1.0
+ pitch_ratio: 1.0
+ #火山tts,支持双向流式tts
+ HuoshanDoubleStreamTTS:
+ type: huoshan_double_stream
+ # 访问 https://console.volcengine.com/speech/service/10007 开通语音合成大模型,购买音色
+ # 在页面底部获取appid和access_token
+ # 资源ID固定为:volc.service_type.10029(大模型语音合成及混音)
+ # 如果是机智云,把接口地址换成wss://bytedance.gizwitsapi.com/api/v3/tts/bidirection
+ # 机智云不需要天填 appid
+ ws_url: wss://openspeech.bytedance.com/api/v3/tts/bidirection
+ appid: 你的火山引擎语音合成服务appid
+ access_token: 你的火山引擎语音合成服务access_token
+ resource_id: volc.service_type.10029
+ speaker: zh_female_wanwanxiaohe_moon_bigtts
+ speech_rate: 0
+ loudness_rate: 0
+ pitch: 0
+ CosyVoiceSiliconflow:
+ type: siliconflow
+ # 硅基流动TTS
+ # token申请地址 https://cloud.siliconflow.cn/account/ak
+ model: FunAudioLLM/CosyVoice2-0.5B
+ voice: FunAudioLLM/CosyVoice2-0.5B:alex
+ output_dir: tmp/
+ access_token: 你的硅基流动API密钥
+ response_format: wav
+ CozeCnTTS:
+ type: cozecn
+ # COZECN TTS
+ # token申请地址 https://www.coze.cn/open/oauth/pats
+ voice: 7426720361733046281
+ output_dir: tmp/
+ access_token: 你的coze web key
+ response_format: wav
+ VolcesAiGatewayTTS:
+ type: openai
+ # 火山引擎 - 边缘大模型网关
+ # 先开通服务,打开以下网址,创建网关访问密钥,搜索并勾选 Doubao-语音合成 ,开通
+ # 如果需要使用边缘大模型网关提供的 LLM,一并勾选 Doubao-pro-32k-functioncall ,另见 LLM.VolcesAiGatewayLLM 配置
+ # https://console.volcengine.com/vei/aigateway/
+ # 开通后,进入这里获取密钥:https://console.volcengine.com/vei/aigateway/tokens-list
+ api_key: 你的网关访问密钥
+ api_url: https://ai-gateway.vei.volces.com/v1/audio/speech
+ model: doubao-tts
+ # 音色列表见 https://www.volcengine.com/docs/6561/1257544
+ voice: zh_male_shaonianzixin_moon_bigtts
+ speed: 1
+ output_dir: tmp/
+ FishSpeech:
+ # 参照教程:https://github.com/xinnan-tech/xiaozhi-esp32-server/blob/main/docs/fish-speech-integration.md
+ type: fishspeech
+ output_dir: tmp/
+ response_format: wav
+ reference_id: null
+ reference_audio: ["config/assets/wakeup_words.wav",]
+ reference_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",]
+ normalize: true
+ max_new_tokens: 1024
+ chunk_length: 200
+ top_p: 0.7
+ repetition_penalty: 1.2
+ temperature: 0.7
+ streaming: false
+ use_memory_cache: "on"
+ seed: null
+ channels: 1
+ rate: 44100
+ api_key: "你的api_key"
+ api_url: "http://127.0.0.1:8080/v1/tts"
+ GPT_SOVITS_V2:
+ # 定义TTS API类型
+ #启动tts方法:
+ #python api_v2.py -a 127.0.0.1 -p 9880 -c GPT_SoVITS/configs/demo.yaml
+ type: gpt_sovits_v2
+ url: "http://127.0.0.1:9880/tts"
+ output_dir: tmp/
+ text_lang: "auto"
+ ref_audio_path: "demo.wav"
+ prompt_text: ""
+ prompt_lang: "zh"
+ top_k: 5
+ top_p: 1
+ temperature: 1
+ text_split_method: "cut0"
+ batch_size: 1
+ batch_threshold: 0.75
+ split_bucket: true
+ return_fragment: false
+ speed_factor: 1.0
+ streaming_mode: false
+ seed: -1
+ parallel_infer: true
+ repetition_penalty: 1.35
+ aux_ref_audio_paths: []
+ GPT_SOVITS_V3:
+ # 定义TTS API类型 GPT-SoVITS-v3lora-20250228
+ #启动tts方法:
+ #python api.py
+ type: gpt_sovits_v3
+ url: "http://127.0.0.1:9880"
+ output_dir: tmp/
+ text_language: "auto"
+ refer_wav_path: "caixukun.wav"
+ prompt_language: "zh"
+ prompt_text: ""
+ top_k: 15
+ top_p: 1.0
+ temperature: 1.0
+ cut_punc: ""
+ speed: 1.0
+ inp_refs: []
+ sample_steps: 32
+ if_sr: false
+ MinimaxTTSHTTPStream:
+ # Minimax流式语音合成服务
+ type: minimax_httpstream
+ output_dir: tmp/
+ group_id: 你的minimax平台groupID
+ api_key: 你的minimax平台接口密钥
+ model: "speech-01-turbo"
+ voice_id: "female-shaonv"
+ # 以下可不用设置,使用默认设置
+ # voice_setting:
+ # voice_id: "male-qn-qingse"
+ # speed: 1
+ # vol: 1
+ # pitch: 0
+ # emotion: "happy"
+ # pronunciation_dict:
+ # tone:
+ # - "处理/(chu3)(li3)"
+ # - "危险/dangerous"
+ # audio_setting:
+ # sample_rate: 24000
+ # bitrate: 128000
+ # format: "mp3"
+ # channel: 1
+ # timber_weights:
+ # -
+ # voice_id: male-qn-qingse
+ # weight: 1
+ # -
+ # voice_id: female-shaonv
+ # weight: 1
+ # language_boost: auto
+ AliyunTTS:
+ # 阿里云智能语音交互服务,需要先在阿里云平台开通服务,然后获取验证信息
+ # 平台地址:https://nls-portal.console.aliyun.com/
+ # appkey地址:https://nls-portal.console.aliyun.com/applist
+ # token地址:https://nls-portal.console.aliyun.com/overview
+ # 定义TTS API类型
+ type: aliyun
+ output_dir: tmp/
+ appkey: 你的阿里云智能语音交互服务项目Appkey
+ token: 你的阿里云智能语音交互服务AccessToken,临时的24小时,要长期用下方的access_key_id,access_key_secret
+ voice: xiaoyun
+ access_key_id: 你的阿里云账号access_key_id
+ access_key_secret: 你的阿里云账号access_key_secret
+
+ # 以下可不用设置,使用默认设置
+ # format: wav
+ # sample_rate: 16000
+ # volume: 50
+ # speech_rate: 0
+ # pitch_rate: 0
+ AliyunStreamTTS:
+ # 阿里云CosyVoice大模型流式文本语音合成
+ # 采用FlowingSpeechSynthesizer接口,支持更低延迟和更自然的语音质量
+ # 流式文本语音合成仅提供商用版,不支持试用,详情请参见试用版和商用版。要使用该功能,请开通商用版。
+ # 支持龙系列专用音色:longxiaochun、longyu、longchen等
+ # 平台地址:https://nls-portal.console.aliyun.com/
+ # appkey地址:https://nls-portal.console.aliyun.com/applist
+ # token地址:https://nls-portal.console.aliyun.com/overview
+ # 使用三阶段流式交互:StartSynthesis -> RunSynthesis -> StopSynthesis
+ type: aliyun_stream
+ output_dir: tmp/
+ appkey: 你的阿里云智能语音交互服务项目Appkey
+ token: 你的阿里云智能语音交互服务AccessToken,临时的24小时,要长期用下方的access_key_id,access_key_secret
+ voice: longxiaochun
+ access_key_id: 你的阿里云账号access_key_id
+ access_key_secret: 你的阿里云账号access_key_secret
+ # 截至2025年7月21日大模型音色只有北京节点采用,其他节点暂不支持
+ host: nls-gateway-cn-beijing.aliyuncs.com
+ # 以下可不用设置,使用默认设置
+ # format: pcm # 音频格式:pcm、wav、mp3
+ # sample_rate: 16000 # 采样率:8000、16000、24000
+ # volume: 50 # 音量:0-100
+ # speech_rate: 0 # 语速:-500到500
+ # pitch_rate: 0 # 语调:-500到500
+ TencentTTS:
+ # 腾讯云智能语音交互服务,需要先在腾讯云平台开通服务
+ # appid、secret_id、secret_key申请地址:https://console.cloud.tencent.com/cam/capi
+ # 免费领取资源:https://console.cloud.tencent.com/tts/resourcebundle
+ type: tencent
+ output_dir: tmp/
+ appid: 你的腾讯云AppId
+ secret_id: 你的腾讯云SecretID
+ secret_key: 你的腾讯云SecretKey
+ region: ap-guangzhou
+ voice: 101001
+
+ TTS302AI:
+ # 302AI语音合成服务,需要先在302平台创建账户充值,并获取密钥信息
+ # 添加 302.ai TTS 配置
+ # token申请地址:https://dash.302.ai/
+ # 获取api_keyn路径:https://dash.302.ai/apis/list
+ # 价格,$35/百万字符。火山原版¥450元/百万字符
+ type: doubao
+ api_url: https://api.302ai.cn/doubao/tts_hd
+ authorization: "Bearer "
+ # 湾湾小何音色
+ voice: "zh_female_wanwanxiaohe_moon_bigtts"
+ output_dir: tmp/
+ access_token: "你的302API密钥"
+ GizwitsTTS:
+ type: doubao
+ # 火山引擎作为基座,可以完全使用企业级火山引擎语音合成服务
+ # 前一万名注册的用户,将送5元体验金额
+ # 获取API Key地址:https://agentrouter.gizwitsapi.com/panel/token
+ api_url: https://bytedance.gizwitsapi.com/api/v1/tts
+ authorization: "Bearer "
+ # 湾湾小何音色
+ voice: "zh_female_wanwanxiaohe_moon_bigtts"
+ output_dir: tmp/
+ access_token: "你的机智云API key"
+ ACGNTTS:
+ #在线网址:https://acgn.ttson.cn/
+ #token购买:www.ttson.cn
+ #开发相关疑问请提交至网站上的qq
+ #角色id获取地址:ctrl+f快速检索角色——网站管理者不允许发布,可询问网站管理者
+ #各参数意义见开发文档:https://www.yuque.com/alexuh/skmti9/wm6taqislegb02gd?singleDoc#
+ type: ttson
+ token: your_token
+ voice_id: 1695
+ speed_factor: 1
+ pitch_factor: 0
+ volume_change_dB: 0
+ to_lang: ZH
+ url: https://u95167-bd74-2aef8085.westx.seetacloud.com:8443/flashsummary/tts?token=
+ format: mp3
+ output_dir: tmp/
+ emotion: 1
+ OpenAITTS:
+ # openai官方文本转语音服务,可支持全球大多数语种
+ type: openai
+ # 你可以在这里获取到 api key
+ # https://platform.openai.com/api-keys
+ api_key: 你的openai api key
+ # 国内需要使用代理
+ api_url: https://api.openai.com/v1/audio/speech
+ # 可选tts-1或tts-1-hd,tts-1速度更快tts-1-hd质量更好
+ model: tts-1
+ # 演讲者,可选alloy, echo, fable, onyx, nova, shimmer
+ voice: onyx
+ # 语速范围0.25-4.0
+ speed: 1
+ output_dir: tmp/
+ CustomTTS:
+ # 自定义的TTS接口服务,请求参数可自定义,可接入众多TTS服务
+ # 以本地部署的KokoroTTS为例
+ # 如果只有cpu运行:docker run -p 8880:8880 ghcr.io/remsky/kokoro-fastapi-cpu:latest
+ # 如果只有gpu运行:docker run --gpus all -p 8880:8880 ghcr.io/remsky/kokoro-fastapi-gpu:latest
+ # 要求接口使用POST方式请求,并返回音频文件
+ type: custom
+ method: POST
+ url: "http://127.0.0.1:8880/v1/audio/speech"
+ params: # 自定义请求参数
+ input: "{prompt_text}"
+ response_format: "mp3"
+ download_format: "mp3"
+ voice: "zf_xiaoxiao"
+ lang_code: "z"
+ return_download_link: true
+ speed: 1
+ stream: false
+ headers: # 自定义请求头
+ # Authorization: Bearer xxxx
+ format: mp3 # 接口返回的音频格式
+ output_dir: tmp/
+ LinkeraiTTS:
+ type: linkerai
+ api_url: https://tts.linkerai.cn/tts
+ audio_format: "pcm"
+ # 默认的access_token供大家测试时免费使用的,此access_token请勿用于商业用途
+ # 如果效果不错,可自行申请token,申请地址:https://linkerai.cn
+ # 各参数意义见开发文档:https://tts.linkerai.cn/docs
+ # 支持声音克隆,可自行上传音频,填入voice参数,voice参数为空时,使用默认声音
+ access_token: "U4YdYXVfpwWnk2t5Gp822zWPCuORyeJL"
+ voice: "OUeAo1mhq6IBExi"
+ output_dir: tmp/
+ PaddleSpeechTTS:
+ #百度飞浆 PaddleSpeech 支持本地离线部署 支持模型训练
+ #框架地址 https://www.paddlepaddle.org.cn/
+ #项目地址 https://github.com/PaddlePaddle/PaddleSpeech
+ #SpeechServerDemo https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/demos/speech_server
+ #流式传输请参考 https://github.com/PaddlePaddle/PaddleSpeech/wiki/PaddleSpeech-Server-WebSocket-API
+ type: paddle_speech
+ protocol: websocket # protocol choices = ['websocket', 'http']
+ url: ws://127.0.0.1:8092/paddlespeech/tts/streaming # TTS 服务的 URL 地址,指向本地服务器 [websocket默认ws://127.0.0.1:8092/paddlespeech/tts/streaming,http默认http://127.0.0.1:8090/paddlespeech/tts]
+ spk_id: 0 # 发音人 ID,0 通常表示默认的发音人
+ sample_rate: 24000 # 采样率 [websocket默认24000,http默认0 自动选择]
+ speed: 1.0 # 语速,1.0 表示正常语速,>1 表示加快,<1 表示减慢
+ volume: 1.0 # 音量,1.0 表示正常音量,>1 表示增大,<1 表示减小
+ save_path: # 保存路径
+ IndexStreamTTS:
+ # 基于Index-TTS-vLLM项目的TTS接口服务
+ # 参照教程:https://github.com/Ksuriuri/index-tts-vllm/blob/master/README.md
+ type: index_stream
+ api_url: http://127.0.0.1:11996/tts
+ audio_format: "pcm"
+ # 默认音色,如需其他音色可到项目assets文件夹下注册
+ voice: "jay_klee"
+ output_dir: tmp/
+ AliBLTTS:
+ # 阿里百炼CosyVoice大模型流式文本语音合成
+ # 可在这里找到你的 api_key https://bailian.console.aliyun.com/?apiKey=1#/api-key
+ # cosyvoice-v3和部分音色需要申请开通
+ type: alibl_stream
+ api_key: 你的api_key
+ model: "cosyvoice-v2"
+ voice: "longcheng_v2"
+ output_dir: tmp/
+ # 以下可不用设置,使用默认设置
+ # format: pcm # 音频格式:pcm、wav、mp3、opus
+ # sample_rate: 24000 # 采样率:16000, 24000, 48000
+ # volume: 50 # 音量:0-100
+ # rate: 1 # 语速:0.5~2
+ # pitch: 1 # 语调:0.5~2
+ XunFeiTTS:
+ # 讯飞TTS服务 官方网站:https://www.xfyun.cn/
+ # 登录讯飞语音技术平台 https://console.xfyun.cn/app/myapp 创建相关应用
+ # 选择需要的服务获取api相关配置 https://console.xfyun.cn/services/uts
+ # 为需要使用的应用(APPID)购买相关服务 例如:超拟人合成 https://console.xfyun.cn/services/uts
+ type: xunfei_stream
+ api_url: wss://cbm01.cn-huabei-1.xf-yun.com/v1/private/mcd9m97e6
+ app_id: 你的app_id
+ api_secret: 你的api_secret
+ api_key: 你的api_key
+ voice: x5_lingxiaoxuan_flow
+ output_dir: tmp/
+ # 以下可不用设置,使用默认设置,注意V5音色不支持口语化配置
+ # oral_level: mid # 口语化等级:high, mid, low
+ # spark_assist: 1 # 是否通过大模型进行口语化 开启:1, 关闭:0
+ # stop_split: 0 # 关闭服务端拆句 不关闭:0,关闭:1
+ # remain: 0 # 是否保留原书面语的样子 保留:1, 不保留:0
+ # format: raw # 音频格式:raw(PCM), lame(MP3), speex, opus, opus-wb, opus-swb, speex-wb
+ # sample_rate: 24000 # 采样率:16000, 8000, 24000
+ # volume: 50 # 音量:0-100
+ # speed: 50 # 语速:0-100
+ # pitch: 50 # 语调:0-100
diff --git a/server/docker/config/bot/.config_from_api.yaml b/server/docker/config/bot/.config_from_api.yaml
new file mode 100644
index 0000000..283178d
--- /dev/null
+++ b/server/docker/config/bot/.config_from_api.yaml
@@ -0,0 +1,25 @@
+# 如果你只想轻量化安装xiaozhi-server,只使用本地的配置文件,不需要理会这个文件,不需要改动本文件任何东西
+# 如果你想从manager-api获取配置,请往下看:
+# 请将本文件复制到xiaozhi-server/data目录下,没有data目录,请创建一个,并将复制过去的文件命名为.config.yaml
+# 注意如果data目录有.config.yaml文件,请先删除它
+# 先启动manager-api和manager-web,注册一个账号,第一个注册的账号为管理员
+# 使用管理员,进入【参数管理】页面,找到【server.secret】,复制它到参数值,注意每次从零部署,server.secret都会变化
+# 打开本data目录下的.config.yaml文件,修改manager-api.secret为刚才复制出来的server.secret
+server:
+ ip: 0.0.0.0
+ port: 8000
+ # http服务的端口,用于视觉分析接口
+ http_port: 8003
+ # 视觉分析接口地址
+ # 向设备发送的视觉分析的接口地址
+ # 如果按下面默认的写法,系统会自动生成视觉识别地址,并输出在启动日志里,这个地址你可以直接用浏览器访问确认一下
+ # 当你使用docker部署或使用公网部署(使用ssl、域名)时,不一定准确
+ # 所以如果你使用docker部署时,将vision_explain设置成局域网地址
+ # 如果你使用公网部署时,将vision_explain设置成公网地址
+ vision_explain: http://你的ip或者域名:端口号/mcp/vision/explain
+manager-api:
+ # 你的manager-api的地址,最好使用局域网ip
+ # 如果使用docker部署,请使用填写成 http://xiaozhi-esp32-server-web:8002/xiaozhi
+ url: http://127.0.0.1:8002/xiaozhi
+ # 你的manager-api的token,就是刚才复制出来的server.secret
+ secret: 你的server.secret值
\ No newline at end of file
diff --git a/server/docker/config/bot/.mcp-endpoint-server.cfg b/server/docker/config/bot/.mcp-endpoint-server.cfg
new file mode 100644
index 0000000..635163f
--- /dev/null
+++ b/server/docker/config/bot/.mcp-endpoint-server.cfg
@@ -0,0 +1,22 @@
+[server]
+host = 0.0.0.0
+port = 8004
+debug = false
+log_level = INFO
+key = 3d880556c4f3470b8b2242d0db0971f9
+
+[websocket]
+max_connections = 1000
+ping_interval = 30
+ping_timeout = 10
+close_timeout = 10
+
+[security]
+allowed_origins = *
+enable_cors = true
+
+[logging]
+log_file = logs/mcp_server.log
+max_file_size = 10MB
+backup_count = 5
+
diff --git a/server/docker/config/bot/.mcp_server_settings.json b/server/docker/config/bot/.mcp_server_settings.json
new file mode 100644
index 0000000..966741b
--- /dev/null
+++ b/server/docker/config/bot/.mcp_server_settings.json
@@ -0,0 +1,56 @@
+{
+ "des": [
+ "在data目录下创建.mcp_server_settings.json文件,可以选择下面的MCP服务,也可以自行添加新的MCP服务。",
+ "后面不断测试补充好用的mcp服务,欢迎大家一起补充。",
+ "记得删除注释行,des属性仅为说明,不会被解析。",
+ "des和link属性,仅为说明安装方式,方便大家查看原始链接,不是必须项。",
+ "当前支持三种传输模式:stdio(标准输入输出), sse(Server-Sent Events), streamable-http(流式HTTP)。"
+ ],
+ "mcpServers": {
+ "Home Assistant": {
+ "command": "mcp-proxy",
+ "args": [
+ "http://YOUR_HA_HOST/mcp_server/sse"
+ ],
+ "env": {
+ "API_ACCESS_TOKEN": "YOUR_API_ACCESS_TOKEN"
+ }
+ },
+ "filesystem": {
+ "command": "npx",
+ "args": [
+ "-y",
+ "@modelcontextprotocol/server-filesystem",
+ "/Users/username/Desktop",
+ "/path/to/other/allowed/dir"
+ ],
+ "link":"https://github.com/modelcontextprotocol/servers/tree/main/src/filesystem"
+ },
+ "playwright": {
+ "command": "npx",
+ "args": ["-y", "@executeautomation/playwright-mcp-server"],
+ "des" : "run 'npx playwright install' first",
+ "link": "https://github.com/executeautomation/mcp-playwright"
+ },
+ "windows-cli": {
+ "command": "npx",
+ "args": ["-y", "@simonb97/server-win-cli"],
+ "link": "https://github.com/SimonB97/win-cli-mcp-server"
+ },
+ "sse-mcp-server": {
+ "url": "http://localhost:8080/sse",
+ "headers": {
+ "Authorization": "Bearer YOUR TOKEN"
+ },
+ "des": "使用SSE传输模式(默认)"
+ },
+ "streamable-http-mcp-server": {
+ "url": "http://localhost:8000/mcp",
+ "transport": "streamable-http",
+ "headers": {
+ "Authorization": "Bearer YOUR TOKEN"
+ },
+ "des": "使用Streamable HTTP传输模式,适用于生产环境的Web部署"
+ }
+ }
+}
diff --git a/server/docker/docker-compose.yaml b/server/docker/docker-compose.yaml
index 2b45f86..9f8afde 100644
--- a/server/docker/docker-compose.yaml
+++ b/server/docker/docker-compose.yaml
@@ -16,8 +16,35 @@ services:
restart: unless-stopped
depends_on:
- vue
+
+ # ---------- Bot Server ----------
+ bot_server:
+ container_name: ce_bot_server
+ image: ai.ronsunny.cn:13011/bbit_ai/ce_bot_server:latest
+ ports:
+ - "8000:8000" # ws服务端
+ - "8003:8003" # http服务的端口,用于简单OTA接口(单服务部署),以及视觉分析接口
+ volumes: # 在本文件所在目录运行
+ - ./config/bot:/app/data # 配置文件目录
+ - ./config/models/SenseVoiceSmall/model.pt:/app/models/SenseVoiceSmall/model.pt # 模型文件
+ networks:
+ - ce_network
+
+ # ---------- Bot MCP ----------
+ bot_mcp:
+ container_name: ce_bot_mcp
+ image: ai.ronsunny.cn:13011/bbit_ai/ce_bot_mcp:latest
+ networks:
+ - ce_network
+ restart: unless-stopped
+ ports:
+ - "8004:8004"
+ depends_on:
+ - vue
volumes:
- - /home/bbit/ssl/ai.ronsunny.cn.pem:/ssl/ai.ronsunny.cn.pem
+ # 配置文件目录
+ - ./config/bot:/app/data
+
# ---------- 网络 ----------
networks:
diff --git a/server/docker/infra-compose-dev.yaml b/server/docker/infra-compose-dev.yaml
index 70f4117..2b8d1c4 100644
--- a/server/docker/infra-compose-dev.yaml
+++ b/server/docker/infra-compose-dev.yaml
@@ -6,17 +6,58 @@ x-kong-config:
KONG_PG_DATABASE: kong
KONG_PG_USER: postgres
KONG_PG_PASSWORD: 123456
+
services:
- fastapi2:
- build: # 在本文件所在目录运行构建 docker compose build --no-cache
- context: ../../
- dockerfile: ./bbit_ai/docker/Dockerfiledev # 开发版 Dockerfile
- container_name: fastapi-dev
+ pybackend:
+ # 因为每次build都会用缓存 调试起来极不方便,所以索性在外部先构建 在这里直接运行镜像
+ # build: # 在本文件所在目录运行构建 docker compose build --no-cache
+ # context: ../../bbit_ai/
+ # dockerfile: ./Dockerfile_ai_lab # 开发版 Dockerfile
+ container_name: ce_pybackend
+ image: ce_pybackend:latest
ports:
- "13011:13011"
volumes: # 在本文件所在目录运行
- ../../bbit_ai/app:/app # 挂载本地代码,实现热更新
- ../../bbit_ai/docker/:/root/.local/pyzxing # 挂载 jar
+ - ./config/bot:/app/data # 配置文件目录
+ networks:
+ - ce_network
+ depends_on:
+ - milvus
+
+ bot_server:
+ # build: # 在本文件所在目录运行构建 docker compose build --no-cache
+ # context: ../../bbit_ai/
+ # dockerfile: ./Dockerfile_bot_server
+ container_name: ce_bot_server
+ image: ce_bot_server:latest
+ ports:
+ # ws服务端
+ - "8000:8000"
+ # http服务的端口,用于简单OTA接口(单服务部署),以及视觉分析接口
+ - "8003:8003"
+ volumes: # 在本文件所在目录运行
+ - ../../bbit_ai/app_bot:/app # 挂载本地代码,实现热更新
+ # 配置文件目录
+ - ./config/bot:/app/data
+ # 模型文件挂接,很重要
+ - ./config/models/SenseVoiceSmall/model.pt:/app/models/SenseVoiceSmall/model.pt
+ networks:
+ - ce_network
+
+ bot_mcp:
+ # build: # 在本文件所在目录运行构建 docker compose build --no-cache
+ # context: ../../bbit_ai/
+ # dockerfile: ./Dockerfile_bot_mcp
+ image: ce_bot_mcp:latest
+ container_name: ce_bot_mcp
+ ports:
+ - "8004:8004"
+ volumes: # 在本文件所在目录运行
+ - ../../bbit_ai/app_mcp:/app
+ # 配置文件目录
+ - ./config/bot:/app/data
networks:
- ce_network