Skip to main content

A Python client library for Xiaozhi AI assistant

Project description

Xiaozhi Client

这是一个用于连接小智AI助手的Python客户端库。它提供了简单的接口来进行语音对话和文本交互。

小智前端-硬件版 小智后端-本地部署版

源码安装

git clone https://github.com/Eric0308/xiaozhi-client.git
cd xiaozhi-client
conda create -n xiaozhi-client python=3.10 -y
conda activate xiaozhi-client 
pip install -e .
python examples/simple_client.py

pip 安装

conda create -n xiaozhi-client python=3.10 -y
conda activate xiaozhi-client 
pip install xiaozhi-client
python examples/simple_client.py

快速开始

这是一个基础的文本对话示例:

import asyncio
import sys
from xiaozhi_client import XiaozhiClient, ClientConfig
from concurrent.futures import ThreadPoolExecutor

async def main():
    client = XiaozhiClient(ClientConfig(ws_url="ws://localhost:8000",))
    
    client.on_tts_end = lambda msg: print("\n请输入消息(输入q退出):")
    
    # 创建输入循环
    executor = ThreadPoolExecutor(max_workers=1)
    loop = asyncio.get_event_loop()
    
    print("\n请输入消息(输入q退出):")
    
    try:
        await client.connect()
        while True:
            message = await loop.run_in_executor(executor, sys.stdin.readline)
            message = message.strip()
            
            if message.lower() == 'q':
                break
                
            if message:
                await client.send_txt_message(message)
                # 等待音频播放完成
                while client.is_playing.is_set():
                    await asyncio.sleep(0.1)
    finally:
        await client.close()

asyncio.run(main())

完整示例代码请参考 simple_client.py

语音对话示例

这是一个基础的语音对话示例:

import asyncio
from xiaozhi_client import XiaozhiClient, ClientConfig, AudioConfig

async def start_voice_chat(
    ws_url: str = "ws://localhost:8000",
    sample_rate: int = 16000,
    channels: int = 1
):
    """启动语音对话"""
    config = ClientConfig(ws_url=ws_url)
    audio_config = AudioConfig(sample_rate=sample_rate, channels=channels)
    client = XiaozhiClient(config, audio_config)
    
    async def on_tts_end(msg):
        print("\n[系统] 继续聆听中... (q:退出)")
        client.resume_voice_input()
        
    client.on_tts_end = on_tts_end
    
    try:
        await client.connect()
        # 启动语音输入
        await client.start_voice_input()
        print("[系统] 开始对话... (q:退出)\n")
        
        # 命令处理循环
        while True:
            cmd = await asyncio.get_event_loop().run_in_executor(None, input, "")
            cmd = cmd.strip().lower()
            
            if cmd == 'q': break

    finally:
        await client.close()
        print("\n[系统] 程序已退出")

asyncio.run(start_voice_chat())

完整示例代码请参考 audio_chat.py

特性

  • WebSocket连接管理
  • 音频编解码(Opus格式)
  • 支持实时语音对话
  • 支持文本消息交互
  • 内置设备标识和认证
  • 支持不同的语音识别模式

配置项

ClientConfig

  • ws_url: WebSocket服务器地址
  • device_token: 设备认证token
  • enable_token: 是否启用token认证
  • protocol_version: 协议版本(默认1)

AudioConfig

  • sample_rate: 采样率(默认16000)
  • channels: 声道数(默认1)
  • frame_size: 帧大小(默认960)
  • frame_duration: 帧时长(默认20ms)
  • format: 音频格式(默认"opus")

支持的消息类型

语音识别

# 开始监听
{
    "type": "listen",
    "state": "start",
    "mode": "auto"  # auto/manual/realtime
}

# 停止监听
{
    "type": "listen",
    "state": "stop"
}

TTS状态回调

{
    "type": "tts",
    "state": "start|stop|sentence_start",
    "text": "要说的文本"  # 仅在 sentence_start 时存在
}

示例

  1. 基础文本对话 - examples/simple_client.py
  2. 实时语音对话 - examples/audio_chat.py

开发说明

音频处理

客户端发送和接收的音频数据都使用Opus编码:

  • 采样率:16000Hz
  • 声道数:1(单声道)
  • 帧大小:960样本/帧
  • 帧时长:20ms

错误处理

客户端会自动处理连接断开等错误:

  • WebSocket连接断开时会触发重连
  • 音频解码错误会被捕获并记录
  • 网络错误会抛出相应异常

依赖

  • websockets==10.4
  • opuslib==3.0.1
  • numpy==2.2.3
  • sounddevice==0.5.1
  • loguru==0.7.3
  • pyaudio==0.2.14

协议文档

完整的协议文档请参考API文档。主要包括:

  • 设备认证
  • 音频流协议
  • 消息类型定义
  • 错误处理机制

致谢

本项目参考和借鉴了以下优秀的开源项目:

xiaozhi-py - Python版本的小智客户端实现

xiaozhi-web-client - Web版本的小智客户端实现

感谢所有为此项目做出贡献的开发者和社区成员。

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

xiaozhi_client-0.1.5.tar.gz (15.9 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

xiaozhi_client-0.1.5-py3-none-any.whl (14.3 kB view details)

Uploaded Python 3

File details

Details for the file xiaozhi_client-0.1.5.tar.gz.

File metadata

  • Download URL: xiaozhi_client-0.1.5.tar.gz
  • Upload date:
  • Size: 15.9 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.10.11

File hashes

Hashes for xiaozhi_client-0.1.5.tar.gz
Algorithm Hash digest
SHA256 7a4f8de4b568860c936838f004c4602b7b739272fb7ecdb36edd1fc7d30ac448
MD5 abc4948fe303f96ea0e043370921502a
BLAKE2b-256 fa703c2dccc1ad672fe9756fe277ed383287382331353cd5f3153db0dda28a6f

See more details on using hashes here.

File details

Details for the file xiaozhi_client-0.1.5-py3-none-any.whl.

File metadata

  • Download URL: xiaozhi_client-0.1.5-py3-none-any.whl
  • Upload date:
  • Size: 14.3 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.10.11

File hashes

Hashes for xiaozhi_client-0.1.5-py3-none-any.whl
Algorithm Hash digest
SHA256 30d9da9808658ecb727437b9e4ba894177fc50d3b1e447c08a8dd01a82af7688
MD5 e72614813d52cd78d6efc570d1848cd6
BLAKE2b-256 208529349004d28de848c7073a5d1deed79791266904759ba47d1512324cd2e8

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page