350f5a20fd
独立模块 cmd/picoclaw-voice，实现端到端语音对话链路（ASR → LLM → TTS）。 协议层基于 xiaozhi WebSocket 协议，做了适当扩展。 - 流式 ASR：豆包 / FunASR（本地） - 流式 TTS：豆包 / Fish Speech（本地） - Docker Compose 一键部署本地 ASR + TTS 服务 - Python demo 客户端用于联调
58 lines
4.1 KiB
Text
58 lines
4.1 KiB
Text
# picoclaw-voice 配置样例
|
||
# 启动方式（推荐）：
|
||
# cp .env.example .env # 编辑 .env 填入实际值
|
||
# ./picoclaw-voice
|
||
|
||
# ─── 监听 ──────────────────────────────────────────────────────────────────────
|
||
PICOCLAW_VOICE_LISTEN=:8765
|
||
|
||
# ─── picoclaw 配置路径（LLM / 记忆 / MCP 工具） ───────────────────────────────
|
||
# 默认读取 ~/.picoclaw/config.json，与 picoclaw 进程共享同一份配置。
|
||
# PICOCLAW_CONFIG=/path/to/config.json
|
||
# PICOCLAW_HOME=/path/to/.picoclaw
|
||
|
||
# ─── 多轮记忆：固定 owner_id ──────────────────────────────────────────────────
|
||
# 设置后所有设备共享同一份对话记忆；留空则每台设备独立记忆。
|
||
# PICOCLAW_VOICE_OWNER_ID=alice
|
||
|
||
# ══════════════════════════════════════════════════════════════════════════════
|
||
# 方案 A：doubao（火山引擎）供应商（默认）
|
||
# ══════════════════════════════════════════════════════════════════════════════
|
||
# ASR + TTS 使用同一套火山引擎账号时，只需填共享凭证：
|
||
PICOCLAW_VOICE_APPID=your_volcengine_appid
|
||
PICOCLAW_VOICE_TOKEN=your_volcengine_token
|
||
|
||
# ASR（可选覆盖）
|
||
PICOCLAW_VOICE_ASR_PROVIDER=doubao
|
||
# PICOCLAW_VOICE_ASR_APPID= # 覆盖共享 AppID
|
||
# PICOCLAW_VOICE_ASR_TOKEN= # 覆盖共享 Token
|
||
PICOCLAW_VOICE_ASR_CLUSTER=bigmodel_transcribe
|
||
PICOCLAW_VOICE_ASR_RESOURCE_ID=volc.bigasr.sauc.duration
|
||
|
||
# TTS（可选覆盖）
|
||
PICOCLAW_VOICE_TTS_PROVIDER=doubao
|
||
# PICOCLAW_VOICE_TTS_APPID= # 覆盖共享 AppID
|
||
# PICOCLAW_VOICE_TTS_TOKEN= # 覆盖共享 Token
|
||
PICOCLAW_VOICE_TTS_CLUSTER=volcano_tts
|
||
PICOCLAW_VOICE_TTS_VOICE=zh_female_wanwanxiaohe_moon_bigtts
|
||
|
||
# ══════════════════════════════════════════════════════════════════════════════
|
||
# 方案 B：本地免费方案（FunASR + Fish Speech）
|
||
# 需要在本机先启动 FunASR（端口 10095）和 Fish Speech（端口 8080）服务
|
||
# 参见：docker/docker-compose.asr-tts.yml
|
||
# ══════════════════════════════════════════════════════════════════════════════
|
||
# ASR：FunASR 本地 WebSocket（支持中文，2pass 模式实时转写+高精度）
|
||
# PICOCLAW_VOICE_ASR_PROVIDER=funasr
|
||
# PICOCLAW_VOICE_ASR_WS_URL=wss://127.0.0.1:10095 # 内置自签名证书，跳过校验
|
||
# PICOCLAW_VOICE_ASR_MODE=2pass # 可选：2pass（默认）| online | offline
|
||
|
||
# TTS：Fish Speech 本地 HTTP（PCM 输出，中文效果好，推荐 GPU）
|
||
# PICOCLAW_VOICE_TTS_PROVIDER=fishspeech
|
||
# PICOCLAW_VOICE_TTS_API_URL=http://127.0.0.1:8080
|
||
# PICOCLAW_VOICE_TTS_REFERENCE_ID= # 留空用默认音色
|
||
# PICOCLAW_VOICE_TTS_SAMPLE_RATE=44100
|
||
|
||
# ══════════════════════════════════════════════════════════════════════════════
|
||
# 方案 C：混搭（例如 doubao ASR + Fish Speech TTS）
|
||
# ══════════════════════════════════════════════════════════════════════════════
|
||
# 只需分别设置 ASR/TTS 对应的 PROVIDER 及其凭证字段，两套配置互不干扰。
|