┌─────────────┐     ┌──────────────┐     ┌─────────────┐
│  小爱音箱   │ ←→  │ Bridge Server│ ←→  │ OpenClaw    │
│  (Client)   │     │  (Python)    │     │ Gateway     │
└─────────────┘     └──────────────┘     └─────────────┘
                           ↓                    ↓
                    ┌──────────────┐     ┌─────────────┐
                    │  小智 AI      │     │  AI Agent   │
                    │  (可选)      │     │  (自定义)   │
                    └──────────────┘     └─────────────┘

麦克风 → client → server → XiaoAI → GlobalStream → KWS/小爱 ASR
→ WakeupSessionManager → before_wakeup() → VAD speech/silence
→ XiaoZhi start/stop listening → xiaozhi-esp32-server

小爱指令 "让龙虾 xxx" → before_wakeup() → send_to_openclaw()
→ OpenClawManager → Gateway → Agent
→ 自动 TTS 播报 或 Agent 主动调用 xiaoai-tts skill

唤醒词 "你好龙虾" → WakeupSessionManager → OpenClawConversationController
→ 循环：VAD 检测语音 → SherpaASR 离线识别 → OpenClaw → TTS 播放
→ 说"退出"/"再见"退出

curl POST /api/play/text → API Server → SpeakerManager → 小爱音箱

# 创建项目目录
mkdir -p open-xiaoai-bridge
cd open-xiaoai-bridge

# 下载配置文件
curl -O https://raw.githubusercontent.com/coderzc/open-xiaoai-bridge/main/config.py
curl -O https://raw.githubusercontent.com/coderzc/open-xiaoai-bridge/main/docker-compose.yml

# 创建模型目录
mkdir -p models

# 将下载的模型文件解压到 models 目录
# 确保模型文件在 ./models 路径下

APP_CONFIG = {
    # OpenClaw 配置
    "openclaw": {
        "url": "ws://127.0.0.1:18789",  # 修改为你的 OpenClaw WebSocket 地址
        "token": "your_openclaw_token",  # 修改为你的认证令牌
        "session_key": "agent:main:open-xiaoai-bridge",
        "identity_path": "/app/openclaw/identity/device.json",
        "tts_speaker": "xiaoai",  # "xiaoai" = 小爱原生 TTS；填豆包音色 ID 则用豆包 TTS
    },

    # 小智 AI 配置（如不需要可禁用）
    "xiaozhi": {
        "OTA_URL": "http://127.0.0.1:8003/xiaozhi/ota/",
        "WEBSOCKET_URL": "ws://127.0.0.1:8000/xiaozhi/v1/",
        "WEBSOCKET_ACCESS_TOKEN": "",
    },

    # 豆包 TTS 配置（可选）
    "tts": {
        "doubao": {
            "app_id": "你的 App ID",
            "access_key": "你的 Access Key",
            "default_speaker": "zh_female_vv_uranus_bigtts",
        }
    },
}

# 启动所有服务
docker compose up -d

# 查看日志
docker compose logs -f

# 停止服务
docker compose down

# 检查服务状态
docker compose ps

# 测试 API 端点
curl http://localhost:9092/api/health

# 克隆项目
git clone https://github.com/coderzc/open-xiaoai-bridge.git
cd open-xiaoai-bridge

# 安装 Python 依赖（需要 uv）
uv sync

# 安装 Rust 依赖（需要 Rust）
cargo build --release

# Ubuntu/Debian
sudo apt-get install pkg-config patchelf

# CentOS/RHEL
sudo yum install pkg-config patchelf

# 创建模型目录
mkdir -p core/models

# 将模型文件解压到 core/models/ 目录

# 设置环境变量并启动
API_SERVER_ENABLE=1 XIAOZHI_ENABLE=1 OPENCLAW_ENABLED=1 ./scripts/start.sh

curl -X POST http://localhost:9092/api/play/text \
  -H "Content-Type: application/json" \
  -d '{"text": "你好，我是小爱同学"}'

curl -X POST http://localhost:9092/api/play/url \
  -H "Content-Type: application/json" \
  -d '{"url": "https://example.com/audio.mp3"}'

curl -X POST http://localhost:9092/api/play/file \
  -F "file=@/path/to/audio.mp3"

curl -X POST http://localhost:9092/api/tts/doubao \
  -H "Content-Type: application/json" \
  -d '{"text": "你好", "speaker_id": "zh_female_cancan_mars_bigtts"}'

curl -X POST http://localhost:9092/api/interrupt

唤醒词 "你好龙虾" → 提示音 → 你说话 → [VAD 检测] → [本地 ASR 识别] 
→ OpenClaw Agent → [TTS 播报] → 提示音 → 你继续说 → ...

# config.py 中的 before_wakeup
if "让龙虾" in text:
    await speaker.abort_xiaoai()
    await app.send_to_openclaw_and_play_reply(text.replace("让龙虾", ""))
    return None  # 框架不做额外处理

if "告诉龙虾" in text:
    await speaker.abort_xiaoai()
    await app.send_to_openclaw(text.replace("告诉龙虾", ""))
    return None

"wakeup": {
    "keywords": [
        "你好小智",        # 中文
        "小智小智",
        "hi open claw",    # 英文（全小写）
        "你好龙虾",
        "龙虾你好",
    ],
},

async def before_wakeup(speaker, text, source, app):
    if source == "kws":          # 唤醒词触发
        if "龙虾" in text:
            await speaker.play(text="龙虾来了")
            return "openclaw"    # → OpenClaw 连续对话
        if "小智" in text:
            await speaker.play(text="小智来了")
            return "xiaozhi"     # → 小智 AI
        return None              # → 不处理

    if source == "xiaoai":       # 小爱语音指令
        if text == "召唤龙虾":
            await speaker.abort_xiaoai()
            return "openclaw"
        if text == "召唤小智":
            await speaker.abort_xiaoai()
            return "xiaozhi"
    # 返回 None → 交给小爱原生处理

# config.py

AGENT_SESSIONS = {
    "龙虾": "agent:assistant:open-xiaoai-bridge",
    "小美": "agent:xiaomei:open-xiaoai-bridge",
    "管家": "agent:butler:open-xiaoai-bridge",
}

async def before_wakeup(speaker, text, source, app):
    if source == "kws":
        for keyword, session_key in AGENT_SESSIONS.items():
            if keyword in text:
                app.set_openclaw_session_key(session_key)  # 切换到对应 Agent
                await speaker.play(text=f"{keyword}来了")
                return "openclaw"                          # 进入连续对话

    if source == "xiaoai":
        for keyword, session_key in AGENT_SESSIONS.items():
            if f"召唤{keyword}" in text:
                app.set_openclaw_session_key(session_key)
                await speaker.abort_xiaoai()
                return "openclaw"

"wakeup": {
    "keywords": [
        "你好龙虾", "你好小美", "你好管家",
    ],
},

async def after_wakeup(speaker, source=None, session_key=None):
    if source == "openclaw":
        # session_key 格式：agent:<agentId>:<rest>，第二段即 agentId
        agent_id = session_key.split(":")[1] if session_key else None
        if agent_id == "assistant":
            await speaker.play(text="龙虾，再见")
        elif agent_id == "xiaomei":
            await speaker.play(text="小美，再见")
        elif agent_id == "butler":
            await speaker.play(text="管家，再见")
        else:
            await speaker.play(text="再见")
    if source == "xiaozhi":
        await speaker.play(text="小智，再见")

"openclaw": {
    # 自动播放/连续对话用：约束输出格式
    "rule_prompt": "注意：将结果处理成纯文字版，不要返回任何 markdown 格式，也不要包含任何代码块，并将字数控制在 300 字以内",

    # Agent 自主播报用：告诉 Agent 需要调用 skill
    "rule_prompt_for_skill": "注意：这条消息是主人通过小爱音箱发送的，他看不到你回复的文字，调用 `xiaoai-tts` skill 播报出来。字数控制在 300 字以内",
}

"tts": {
    "doubao": {
        "app_id": "你的 App ID",
        "access_key": "你的 Access Key",
        "default_speaker": "zh_female_cancan_mars_bigtts",  # 默认音色
        "stream": True,           # 流式播放，首音延迟更低
        "audio_format": "pcm",    # 局域网推荐，首音更快
    }
}

"tts": {
    "doubao": {
        "default_speaker": "S_xxxxxxxx",
    }
}

"tts": {
    "doubao": {
        "stream": True,           # 流式播放，首音延迟更低
        "audio_format": "pcm",    # 局域网推荐，首音更快
        # "audio_format": "auto", # 短文本 PCM，长文本 MP3
    }
}

APP_CONFIG = {
    "vad": {
        "min_silence_duration": 1000,  # 毫秒
    },
}

# docker-compose.yml
volumes:
  - ./openclaw:/app/openclaw

agent:<agentId>:<rest>

agent:main:open-xiaoai-bridge          # 默认值（本项目）
agent:main:main                        # OpenClaw 原生默认主会话
agent:assistant:open-xiaoai-bridge     # 指定其他 Agent
agent:main:direct:alice                # 按用户隔离

APP_CONFIG = {
    "xiaozhi": {
        "VERIFICATION_CODE": "首次登录时，验证码会在这里更新",
    },
}

APP_CONFIG = {
    "xiaozhi": {
        "OTA_URL": "https://your-server/xiaozhi/ota/",
        "WEBSOCKET_URL": "wss://your-server/xiaozhi/v1/",
    },
}