xiaozhi-sdk 0.2.5__tar.gz → 0.2.8__tar.gz

{xiaozhi_sdk-0.2.5/xiaozhi_sdk.egg-info → xiaozhi_sdk-0.2.8}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: xiaozhi-sdk
-Version: 0.2.5
+Version: 0.2.8
 Summary: 一个用于连接和控制小智智能设备的Python SDK，支持实时音频通信、MCP工具集成和设备管理功能。
 Author-email: dairoot <623815825@qq.com>
 License-Expression: MIT
@@ -43,7 +43,8 @@ Dynamic: license-file
 ## 📦 安装
 
 ```bash
-pip install xiaozhi-sdk
+pip install uv
+uv pip install xiaozhi-sdk -U
 ```
 
 ---
@@ -60,10 +61,21 @@ pip install xiaozhi-sdk
 python -m xiaozhi_sdk --help
 ```
 
-#### 连接设备（需要提供 MAC 地址）
+#### 连接设备
 
 ```bash
+# 默认本机 mac 地址
+python -m xiaozhi_sdk
+
+# 指定 mac 地址
 python -m xiaozhi_sdk 00:22:44:66:88:00
+
+# 更多常用操作
+## --url 指定服务端 websocket 地址
+## --wake_word 指定唤醒词
+python -m xiaozhi_sdk 00:22:44:66:88:00 \
+  --url ws://127.0.0.1:8180 \
+  --wake_word="你好啊"
 ```
 
 ### 2. 编程使用 (高阶用法)

{xiaozhi_sdk-0.2.5 → xiaozhi_sdk-0.2.8}/README.md

@@ -16,7 +16,8 @@
 ## 📦 安装
 
 ```bash
-pip install xiaozhi-sdk
+pip install uv
+uv pip install xiaozhi-sdk -U
 ```
 
 ---
@@ -33,10 +34,21 @@ pip install xiaozhi-sdk
 python -m xiaozhi_sdk --help
 ```
 
-#### 连接设备（需要提供 MAC 地址）
+#### 连接设备
 
 ```bash
+# 默认本机 mac 地址
+python -m xiaozhi_sdk
+
+# 指定 mac 地址
 python -m xiaozhi_sdk 00:22:44:66:88:00
+
+# 更多常用操作
+## --url 指定服务端 websocket 地址
+## --wake_word 指定唤醒词
+python -m xiaozhi_sdk 00:22:44:66:88:00 \
+  --url ws://127.0.0.1:8180 \
+  --wake_word="你好啊"
 ```
 
 ### 2. 编程使用 (高阶用法)

{xiaozhi_sdk-0.2.5 → xiaozhi_sdk-0.2.8}/pyproject.toml

@@ -43,6 +43,7 @@ include = ["xiaozhi_sdk*"]
 xiaozhi_sdk = ["../file/**/*"]
 
 [tool.uv]
+index-url = "https://pypi.tuna.tsinghua.edu.cn/simple"
 dev-dependencies = [
     "black>=24.8.0",
     "flake8>=5.0.4",

xiaozhi_sdk-0.2.8/tests/test_wake_word.py

@@ -0,0 +1,55 @@
+import asyncio
+import os
+import sys
+
+import pytest
+
+sys.path.insert(0, os.path.join(os.path.dirname(__file__), ".."))
+
+from xiaozhi_sdk import XiaoZhiWebsocket
+from xiaozhi_sdk.utils import read_audio_file
+
+sample_rate = 16000
+frame_duration = 60
+
+MAC_ADDR = "00:22:44:66:88:00"
+
+URL = None
+ota_url = None
+
+
+async def test_main():
+    is_end = asyncio.Event()
+    async def message_handler_callback(message):
+        if message.get("state") == "stop":
+            is_end.set()
+        print("message received:", message)
+
+    xiaozhi = XiaoZhiWebsocket(
+        message_handler_callback, url=URL, ota_url=ota_url,
+        audio_sample_rate=sample_rate, audio_frame_duration=frame_duration)
+    await xiaozhi.init_connection(MAC_ADDR)
+
+    await xiaozhi.send_wake_word("你好")
+    await asyncio.sleep(5)
+
+    # await xiaozhi.send_wake_word("1+1")
+    # await asyncio.sleep(5)
+    #
+    # await xiaozhi.send_wake_word("你是什么大语言模型")
+    # await asyncio.sleep(5)
+
+    # say hellow
+    for pcm in read_audio_file("./file/audio/16k_say_hello.wav", sample_rate, frame_duration):
+        await xiaozhi.send_audio(pcm)
+    await xiaozhi.send_silence_audio()
+    await asyncio.sleep(5)
+
+    await xiaozhi.close()
+
+
+if __name__ == "__main__":
+    asyncio.run(test_main())
+
+
+

{xiaozhi_sdk-0.2.5 → xiaozhi_sdk-0.2.8}/tests/test_xiaozhi.py

@@ -11,11 +11,12 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), ".."))
 
 from xiaozhi_sdk import XiaoZhiWebsocket
 from xiaozhi_sdk.utils import read_audio_file
-
+sample_rate = 16000
+frame_duration = 60
 
 async def assistant_audio_play(audio_queue, wait_time=5):
     # 创建一个持续播放的流
-    stream = sd.OutputStream(samplerate=16000, channels=1, dtype=np.int16)
+    stream = sd.OutputStream(samplerate=sample_rate, channels=1, dtype=np.int16)
     stream.start()
     last_time = int(time.time())
     while True:
@@ -78,29 +79,32 @@ URL = None
 # URL = None
 
 
-@pytest.mark.asyncio
 async def test_main():
-    xiaozhi = XiaoZhiWebsocket(message_handler_callback, url=URL, ota_url=ota_url)
+    xiaozhi = XiaoZhiWebsocket(message_handler_callback, url=URL, ota_url=ota_url, audio_sample_rate=sample_rate, audio_frame_duration=frame_duration)
 
     await xiaozhi.set_mcp_tool(mcp_tool_func())
     await xiaozhi.init_connection(MAC_ADDR)
 
     # # say hellow
-    for pcm in read_audio_file("./file/audio/say_hello.wav"):
+    for pcm in read_audio_file("./file/audio/16k_say_hello.wav", sample_rate, frame_duration):
         await xiaozhi.send_audio(pcm)
     await xiaozhi.send_silence_audio()
     await assistant_audio_play(xiaozhi.output_audio_queue)
 
     # say take photo
-    for pcm in read_audio_file("./file/audio/take_photo.wav"):
+    for pcm in read_audio_file("./file/audio/16k_take_photo.wav", sample_rate, frame_duration):
         await xiaozhi.send_audio(pcm)
     await xiaozhi.send_silence_audio()
     await assistant_audio_play(xiaozhi.output_audio_queue, 5)
 
     # play music
-    # for pcm in read_audio_file("./file/audio/play_music.wav"):
-    #     await xiaozhi.send_audio(pcm)
-    # await xiaozhi.send_silence_audio()
-    # await assistant_audio_play(xiaozhi.output_audio_queue, 500)
+    for pcm in read_audio_file("./file/audio/16k_play_music.wav", sample_rate, frame_duration):
+        await xiaozhi.send_audio(pcm)
+    await xiaozhi.send_silence_audio()
+    await assistant_audio_play(xiaozhi.output_audio_queue, 500)
 
     await xiaozhi.close()
+
+
+if __name__ == "__main__":
+    asyncio.run(test_main())

xiaozhi_sdk-0.2.8/tests/test_xiaozhi_opus.py

@@ -0,0 +1,88 @@
+import asyncio
+import os
+import sys
+import time
+
+import numpy as np
+import sounddevice as sd
+
+sys.path.insert(0, os.path.join(os.path.dirname(__file__), ".."))
+
+from xiaozhi_sdk import XiaoZhiWebsocket
+from xiaozhi_sdk.utils import read_audio_file
+
+sample_rate = 48000
+frame_duration = 60
+MAC_ADDR = "00:22:44:66:88:00"
+
+
+async def assistant_audio_play(audio_queue, wait_time=5):
+    # 创建一个持续播放的流
+    stream = sd.OutputStream(samplerate=sample_rate, channels=1, dtype=np.int16)
+    stream.start()
+    last_time = int(time.time())
+    while True:
+        if not audio_queue:
+            await asyncio.sleep(0.01)
+            if last_time and time.time() - last_time > wait_time:
+                break
+
+            continue
+
+        pcm_data = audio_queue.popleft()
+
+        # 将字节数据转换为 numpy int16 数组
+        audio_array = pcm_data
+
+        stream.write(audio_array)
+        last_time = time.time()
+
+    stream.stop()
+    stream.close()
+
+
+async def message_handler_callback(message):
+    print("message received:", message)
+    if message["type"] == "music":
+        print("music:", message["text"])
+
+
+async def test_main():
+    xiaozhi = XiaoZhiWebsocket(message_handler_callback, audio_sample_rate=sample_rate,
+                               audio_frame_duration=frame_duration)
+
+    await xiaozhi.init_connection(MAC_ADDR)
+    current_dir = os.path.dirname(os.path.abspath(__file__))
+    test_audio_file = "../file/audio/test_16k.wav"
+
+    if sample_rate == 24000:
+        test_audio_file = "../file/audio/test_24k.wav"
+    elif sample_rate == 48000:
+        test_audio_file = "../file/audio/test_48k.wav"
+    wav_path = os.path.join(current_dir, test_audio_file)
+
+    for pcm in read_audio_file(wav_path, sample_rate, frame_duration):
+        await xiaozhi.send_audio(pcm)
+    await xiaozhi.send_silence_audio()
+
+    await assistant_audio_play(xiaozhi.output_audio_queue)
+
+    for pcm in read_audio_file(wav_path, sample_rate, frame_duration):
+        await xiaozhi.send_audio(pcm)
+    await xiaozhi.send_silence_audio()
+
+    await assistant_audio_play(xiaozhi.output_audio_queue)
+
+    for pcm in read_audio_file(wav_path, sample_rate, frame_duration):
+        await xiaozhi.send_audio(pcm)
+    await xiaozhi.send_silence_audio()
+
+    await assistant_audio_play(xiaozhi.output_audio_queue)
+
+    time.sleep(10)
+
+    await xiaozhi.close()
+
+
+if __name__ == "__main__":
+    asyncio.run(test_main())

{xiaozhi_sdk-0.2.5 → xiaozhi_sdk-0.2.8}/xiaozhi_sdk/__init__.py

@@ -1,3 +1,3 @@
-__version__ = "0.2.5"
+__version__ = "0.2.8"
 
 from xiaozhi_sdk.core import XiaoZhiWebsocket  # noqa

{xiaozhi_sdk-0.2.5 → xiaozhi_sdk-0.2.8}/xiaozhi_sdk/cli.py

@@ -2,7 +2,9 @@ import asyncio
 import io
 import logging
 import time
+import uuid
 from collections import deque
+from tkinter import NO
 from typing import Optional
 
 import click
@@ -12,7 +14,11 @@ import sounddevice as sd
 from PIL import ImageGrab
 
 from xiaozhi_sdk import XiaoZhiWebsocket
-from xiaozhi_sdk.config import INPUT_SERVER_AUDIO_SAMPLE_RATE
+from xiaozhi_sdk.config import (
+    INPUT_AUDIO_CHANNELS,
+    INPUT_AUDIO_FRAME_DURATION,
+    INPUT_AUDIO_SAMPLE_RATE,
+)
 
 # 定义自定义日志级别
 INFO1 = 21
@@ -50,7 +56,7 @@ logging.Logger.info3 = info3
 handler = colorlog.StreamHandler()
 handler.setFormatter(
     colorlog.ColoredFormatter(
-        "%(log_color)s%(asctime)s - %(name)s - %(levelname)s - %(message)s",
+        "%(log_color)s%(asctime)s - %(name)s - %(levelname)-5s - %(message)s",
         datefmt="%Y-%m-%d %H:%M:%S",
         log_colors={
             "DEBUG": "white",
@@ -71,8 +77,10 @@ logger.setLevel(logging.DEBUG)
 
 # 全局状态
 input_audio_buffer: deque[bytes] = deque()
-is_playing_audio = False
+device_stauts = "listen"  # "speak" or "listen"
+
 is_end = False
+human_speak_time = None
 
 
 def get_image_byte(data):
@@ -103,16 +111,26 @@ def get_image_byte(data):
 
 
 async def handle_message(message):
+    global device_stauts
+    global human_speak_time
+
     """处理接收到的消息"""
     global is_end
-    if message["type"] == "stt":  # 人类语音
+    if message["type"] == "tts" and message["state"] == "start":  # start
+        pass
+
+    elif message["type"] == "stt":  # 人类语音
+        human_speak_time = time.time()
         logger.info1("human: %s", message["text"])
+
     elif message["type"] == "tts" and message["state"] == "sentence_start":  # AI语音
+        device_stauts = "speak"  # 防止打断
         logger.info2("AI: %s", message["text"])
+
     elif message["type"] == "tts" and message["state"] == "stop":
-        pass
+        device_stauts = "listen"
         # logger.info2("播放结束")
-        # logger.info("聆听中...")
+        logger.info("聆听中...")
     elif message["type"] == "llm":  # 表情
         logger.info3("emotion: %s", message["text"])
     else:  # 其他消息
@@ -123,31 +141,42 @@ async def handle_message(message):
         is_end = True
 
 
-async def play_assistant_audio(audio_queue: deque[bytes], enable_audio):
+async def play_assistant_audio(audio_queue: deque[bytes], enable_audio, audio_samplerate):
     """播放音频流"""
-    global is_playing_audio
+    global device_stauts
+    global human_speak_time
 
     stream = None
     if enable_audio:
-        stream = sd.OutputStream(samplerate=INPUT_SERVER_AUDIO_SAMPLE_RATE, channels=1, dtype=np.int16)
+        stream = sd.OutputStream(samplerate=audio_samplerate, channels=INPUT_AUDIO_CHANNELS, dtype=np.int16)
         stream.start()
+
     last_audio_time = None
 
     while True:
         if is_end:
             return
 
+        if device_stauts == "listen":
+            last_audio_time = None
+
         if not audio_queue:
+            # 空音频 超过 2s ，将device_stauts 设置为listen，代表聆听中
+            if device_stauts == "speak" and last_audio_time and time.time() - last_audio_time > 2:
+                device_stauts = "listen"
+
             await asyncio.sleep(0.01)
-            if last_audio_time and time.time() - last_audio_time > 1:
-                is_playing_audio = False
             continue
 
-        is_playing_audio = True
+        last_audio_time = time.time()
+
+        if human_speak_time:
+            logger.debug("首个音频包响应时间：%s 秒", time.time() - human_speak_time)
+            human_speak_time = None
+
         pcm_data = audio_queue.popleft()
         if stream:
             stream.write(pcm_data)
-        last_audio_time = time.time()
 
 
 class XiaoZhiClient:
@@ -165,10 +194,16 @@ class XiaoZhiClient:
         self.mac_address = ""
         self.wake_word = wake_word
 
-    async def start(self, mac_address: str, serial_number: str, license_key: str, enable_audio):
+    async def start(self, mac_address: str, serial_number: str, license_key: str, enable_audio, audio_samplerate):
         """启动客户端连接"""
         self.mac_address = mac_address
-        self.xiaozhi = XiaoZhiWebsocket(handle_message, url=self.url, ota_url=self.ota_url, wake_word=self.wake_word)
+        self.xiaozhi = XiaoZhiWebsocket(
+            handle_message,
+            url=self.url,
+            ota_url=self.ota_url,
+            wake_word=self.wake_word,
+            audio_sample_rate=audio_samplerate,
+        )
         from xiaozhi_sdk.utils.mcp_tool import take_photo
 
         take_photo["tool_func"] = get_image_byte
@@ -178,7 +213,7 @@ class XiaoZhiClient:
             self.mac_address, aec=False, serial_number=serial_number, license_key=license_key
         )
 
-        asyncio.create_task(play_assistant_audio(self.xiaozhi.output_audio_queue, enable_audio))
+        asyncio.create_task(play_assistant_audio(self.xiaozhi.output_audio_queue, enable_audio, audio_samplerate))
 
     def audio_callback(self, indata, frames, time, status):
         """音频输入回调函数"""
@@ -193,28 +228,49 @@ class XiaoZhiClient:
                 return
 
             if not input_audio_buffer:
-                await asyncio.sleep(0.02)
+                await asyncio.sleep(0.01)
                 continue
 
             pcm_data = input_audio_buffer.popleft()
-            if not is_playing_audio:
+            if device_stauts == "listen":
+
                 await self.xiaozhi.send_audio(pcm_data)
+            else:
+                input_audio_buffer.clear()
 
 
 async def run_client(
-    mac_address: str, url: str, ota_url: str, serial_number: str, license_key: str, enable_audio: bool, wake_word: str
+    mac_address: str,
+    url: str,
+    ota_url: str,
+    serial_number: str,
+    license_key: str,
+    enable_audio: bool,
+    wake_word: str,
 ):
     """运行客户端的异步函数"""
     logger.debug("Recording... Press Ctrl+C to stop.")
     client = XiaoZhiClient(url, ota_url, wake_word)
-    await client.start(mac_address, serial_number, license_key, enable_audio)
-
-    with sd.InputStream(callback=client.audio_callback, channels=1, samplerate=16000, blocksize=960):
+    await client.start(mac_address, serial_number, license_key, enable_audio, INPUT_AUDIO_SAMPLE_RATE)
+    blocksize = INPUT_AUDIO_SAMPLE_RATE * INPUT_AUDIO_FRAME_DURATION // 1000
+    with sd.InputStream(
+        callback=client.audio_callback,
+        channels=INPUT_AUDIO_CHANNELS,
+        samplerate=INPUT_AUDIO_SAMPLE_RATE,
+        blocksize=blocksize,
+    ):
+        logger.info("聆听中...")
         await client.process_audio_input()
 
 
+def get_mac_address():
+    mac = uuid.getnode()
+    mac_addr = ":".join(["%02x" % ((mac >> ele) & 0xFF) for ele in range(40, -8, -8)])
+    return mac_addr
+
+
 @click.command()
-@click.argument("mac_address")
+@click.argument("mac_address", required=False)
 @click.option("--url", help="服务端websocket地址")
 @click.option("--ota_url", help="OTA地址")
 @click.option("--serial_number", default="", help="设备的序列号")
@@ -222,10 +278,17 @@ async def run_client(
 @click.option("--enable_audio", default=True, help="是否开启音频播放")
 @click.option("--wake_word", default="", help="唤醒词")
 def main(
-    mac_address: str, url: str, ota_url: str, serial_number: str, license_key: str, enable_audio: bool, wake_word: str
+    mac_address: str,
+    url: str,
+    ota_url: str,
+    serial_number: str,
+    license_key: str,
+    enable_audio: bool,
+    wake_word: str,
 ):
     """小智SDK客户端
 
     MAC_ADDRESS: 设备的MAC地址 (格式: XX:XX:XX:XX:XX:XX)
     """
+    mac_address = mac_address or get_mac_address()
     asyncio.run(run_client(mac_address, url, ota_url, serial_number, license_key, enable_audio, wake_word))

xiaozhi_sdk-0.2.8/xiaozhi_sdk/config.py

@@ -0,0 +1,7 @@
+XIAOZHI_SAMPLE_RATE = 16000  # 固定不变动
+
+INPUT_AUDIO_SAMPLE_RATE = 24000
+INPUT_AUDIO_CHANNELS = 1
+INPUT_AUDIO_FRAME_DURATION = 60
+
+OTA_URL = "https://api.tenclass.net/xiaozhi/ota"

{xiaozhi_sdk-0.2.5 → xiaozhi_sdk-0.2.8}/xiaozhi_sdk/core.py

@@ -1,7 +1,6 @@
 import asyncio
 import json
 import logging
-import os
 import re
 import uuid
 from collections import deque
@@ -9,10 +8,15 @@ from typing import Any, Callable, Deque, Dict, Optional
 
 import websockets
 
-from xiaozhi_sdk.config import INPUT_SERVER_AUDIO_SAMPLE_RATE
+from xiaozhi_sdk.config import (
+    INPUT_AUDIO_CHANNELS,
+    INPUT_AUDIO_FRAME_DURATION,
+    INPUT_AUDIO_SAMPLE_RATE,
+    XIAOZHI_SAMPLE_RATE,
+)
 from xiaozhi_sdk.iot import OtaDevice
 from xiaozhi_sdk.mcp import McpTool
-from xiaozhi_sdk.utils import get_wav_info, read_audio_file, setup_opus
+from xiaozhi_sdk.utils import setup_opus
 
 setup_opus()
 from xiaozhi_sdk.opus import AudioOpus
@@ -27,15 +31,17 @@ class XiaoZhiWebsocket(McpTool):
         message_handler_callback: Optional[Callable] = None,
         url: Optional[str] = None,
         ota_url: Optional[str] = None,
-        audio_sample_rate: int = 16000,
-        audio_channels: int = 1,
+        audio_sample_rate: int = INPUT_AUDIO_SAMPLE_RATE,
+        audio_channels: int = INPUT_AUDIO_CHANNELS,
+        audio_frame_duration=INPUT_AUDIO_FRAME_DURATION,
         wake_word: str = "",
     ):
         super().__init__()
         self.url = url
         self.ota_url = ota_url
         self.audio_channels = audio_channels
-        self.audio_opus = AudioOpus(audio_sample_rate, audio_channels)
+        self.audio_frame_duration = audio_frame_duration
+        self.audio_opus = AudioOpus(audio_sample_rate, audio_channels, audio_frame_duration)
         self.wake_word = wake_word
 
         # 客户端标识
@@ -70,13 +76,13 @@ class XiaoZhiWebsocket(McpTool):
         hello_message = {
             "type": "hello",
             "version": 1,
-            "features": {"mcp": True, "aec": aec},
+            "features": {"mcp": True, "aec": aec, "consistent_sample_rate": False},
             "transport": "websocket",
             "audio_params": {
                 "format": "opus",
-                "sample_rate": 16000,
+                "sample_rate": XIAOZHI_SAMPLE_RATE,
                 "channels": 1,
-                "frame_duration": 60,
+                "frame_duration": self.audio_opus.input_frame_duration,
             },
         }
         await self.websocket.send(json.dumps(hello_message))
@@ -108,17 +114,17 @@ class XiaoZhiWebsocket(McpTool):
                 break
             await asyncio.sleep(3)
 
-    async def _send_demo_audio(self) -> None:
-        """发送演示音频"""
-        current_dir = os.path.dirname(os.path.abspath(__file__))
-        wav_path = os.path.join(current_dir, "../file/audio/greet.wav")
-        framerate, channels = get_wav_info(wav_path)
-        audio_opus = AudioOpus(framerate, channels)
-
-        for pcm_data in read_audio_file(wav_path):
-            opus_data = await audio_opus.pcm_to_opus(pcm_data)
-            await self.websocket.send(opus_data)
-        await self.send_silence_audio()
+    # async def _send_demo_audio(self) -> None:
+    #     """发送演示音频"""
+    #     current_dir = os.path.dirname(os.path.abspath(__file__))
+    #     wav_path = os.path.join(current_dir, "../file/audio/16k_greet.wav")
+    #     framerate, channels = get_wav_info(wav_path)
+    #     audio_opus = AudioOpus(framerate, channels, self.audio_frame_duration)
+    #
+    #     for pcm_data in read_audio_file(wav_path, 16000, self.audio_frame_duration):
+    #         opus_data = await audio_opus.pcm_to_opus(pcm_data)
+    #         await self.websocket.send(opus_data)
+    #     await self.send_silence_audio()
 
     async def send_wake_word(self, wake_word: str) -> bool:
         """发送唤醒词"""
@@ -137,8 +143,8 @@ class XiaoZhiWebsocket(McpTool):
 
     async def send_silence_audio(self, duration_seconds: float = 1.2) -> None:
         """发送静音音频"""
-        frames_count = int(duration_seconds * 1000 / 60)
-        pcm_frame = b"\x00\x00" * int(INPUT_SERVER_AUDIO_SAMPLE_RATE / 1000 * 60)
+        frames_count = int(duration_seconds * 1000 / self.audio_opus.input_frame_duration)
+        pcm_frame = b"\x00\x00" * int(self.audio_opus.input_sample_rate / 1000 * self.audio_opus.input_frame_duration)
 
         for _ in range(frames_count):
             await self.send_audio(pcm_frame)
@@ -159,6 +165,7 @@ class XiaoZhiWebsocket(McpTool):
         data = json.loads(message)
         message_type = data["type"]
         if message_type == "hello":
+            self.audio_opus.set_out_audio_frame(data["audio_params"])
             self.hello_received.set()
             self.session_id = data["session_id"]
             return
@@ -219,7 +226,7 @@ class XiaoZhiWebsocket(McpTool):
 
         await self._send_hello(self.aec)
         await self._start_listen()
-        logger.debug("[websocket] Connection successful")
+        logger.debug("[websocket] Connection successful. mac_addr: %s", self.mac_addr)
         await asyncio.sleep(0.5)
 
     async def init_connection(
@@ -250,7 +257,9 @@ class XiaoZhiWebsocket(McpTool):
 
         if not await self.is_activate(ota_info):
             self.iot_task = asyncio.create_task(self._activate_iot_device(license_key, ota_info))
+            await self.send_wake_word("hi")
             logger.debug("[IOT] 设备未激活")
+            return
 
         if self.wake_word:
             await self.send_wake_word(self.wake_word)

xiaozhi_sdk-0.2.8/xiaozhi_sdk/opus.py

@@ -0,0 +1,74 @@
+import av
+import numpy as np
+import opuslib
+
+from xiaozhi_sdk.config import XIAOZHI_SAMPLE_RATE
+
+
+class AudioOpus:
+
+    def __init__(self, sample_rate, channels, frame_duration):
+        self.input_frame_duration = frame_duration
+        self.input_sample_rate = sample_rate
+        self.input_channels = channels
+        self.input_frame_size = self.input_sample_rate * self.input_frame_duration // 1000
+
+        # 创建 Opus 编码器
+        self.opus_encoder_16k = opuslib.Encoder(
+            fs=XIAOZHI_SAMPLE_RATE, channels=1, application=opuslib.APPLICATION_VOIP
+        )
+
+        self.resampler = av.AudioResampler(format="s16", layout="mono", rate=sample_rate)
+        self.resampler_16k = av.AudioResampler(format="s16", layout="mono", rate=16000)
+
+    def set_out_audio_frame(self, audio_params):
+        # 小智服务端 的 音频信息
+        self.out_sample_rate = audio_params["sample_rate"]
+        self.out_frame_size = self.out_sample_rate * audio_params["frame_duration"] // 1000
+
+        # 创建 Opus 解码器
+        self.opus_decoder = opuslib.Decoder(
+            fs=self.out_sample_rate,  # 采样率
+            channels=audio_params["channels"],  # 单声道
+        )
+
+    def to_16k_samplerate_pcm(self, pcm_array):
+        layout = "mono" if self.input_channels == 1 else "stereo"
+        frame = av.AudioFrame.from_ndarray(pcm_array.reshape(1, -1), format="s16", layout=layout)
+        frame.sample_rate = self.input_sample_rate
+        resampled_frames = self.resampler_16k.resample(frame)
+        samples = resampled_frames[0].to_ndarray().flatten()
+        return samples
+
+    async def pcm_to_opus(self, pcm):
+        pcm_array = np.frombuffer(pcm, dtype=np.int16)
+        pcm_bytes = pcm_array.tobytes()
+        if self.input_sample_rate != XIAOZHI_SAMPLE_RATE:
+            # 小智服务端仅支持 16000 采样率， 将 pcm_array 转 16k 采样率
+            pcm_array = self.to_16k_samplerate_pcm(pcm_array)
+            pcm_bytes = pcm_array.tobytes()
+
+        frame_size = XIAOZHI_SAMPLE_RATE * self.input_frame_duration // 1000
+        return self.opus_encoder_16k.encode(pcm_bytes, frame_size)
+
+    async def change_sample_rate(self, pcm_array) -> np.ndarray:
+        # 采样率 变更
+        frame = av.AudioFrame.from_ndarray(np.array(pcm_array).reshape(1, -1), format="s16", layout="mono")
+        frame.sample_rate = self.out_sample_rate
+        resampled_frames = self.resampler.resample(frame)
+        samples = resampled_frames[0].to_ndarray().flatten()
+        return samples
+
+    def padding(self, samples):
+        # 不足 self.frame_size 补 0
+        samples_padded = np.pad(samples, (0, self.input_frame_size - samples.size), mode="constant", constant_values=0)
+        return samples_padded.reshape(1, self.input_frame_size)
+
+    async def opus_to_pcm(self, opus) -> np.ndarray:
+        pcm_data = self.opus_decoder.decode(opus, frame_size=self.out_frame_size)
+        pcm_array = np.frombuffer(pcm_data, dtype=np.int16)
+        if self.input_sample_rate != self.out_sample_rate:
+            pcm_array = await self.change_sample_rate(pcm_array)
+
+        pcm_array = self.padding(pcm_array)
+        return pcm_array

{xiaozhi_sdk-0.2.5 → xiaozhi_sdk-0.2.8}/xiaozhi_sdk/utils/__init__.py

@@ -9,7 +9,7 @@ def get_wav_info(file_path):
         return wav_file.getframerate(), wav_file.getnchannels()
 
 
-def read_audio_file(file_path):
+def read_audio_file(file_path, sample_rate, frame_duration):
     """
     读取音频文件并通过yield返回PCM流
 
@@ -19,9 +19,10 @@ def read_audio_file(file_path):
     Yields:
         bytes: PCM音频数据块
     """
+    frame_size = sample_rate * frame_duration // 1000
     with wave.open(file_path, "rb") as wav_file:
         while True:
-            pcm = wav_file.readframes(960)  # 每次读取960帧（60ms的音频数据）
+            pcm = wav_file.readframes(frame_size)
             if not pcm:
                 break
             yield pcm

{xiaozhi_sdk-0.2.5 → xiaozhi_sdk-0.2.8/xiaozhi_sdk.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: xiaozhi-sdk
-Version: 0.2.5
+Version: 0.2.8
 Summary: 一个用于连接和控制小智智能设备的Python SDK，支持实时音频通信、MCP工具集成和设备管理功能。
 Author-email: dairoot <623815825@qq.com>
 License-Expression: MIT
@@ -43,7 +43,8 @@ Dynamic: license-file
 ## 📦 安装
 
 ```bash
-pip install xiaozhi-sdk
+pip install uv
+uv pip install xiaozhi-sdk -U
 ```
 
 ---
@@ -60,10 +61,21 @@ pip install xiaozhi-sdk
 python -m xiaozhi_sdk --help
 ```
 
-#### 连接设备（需要提供 MAC 地址）
+#### 连接设备
 
 ```bash
+# 默认本机 mac 地址
+python -m xiaozhi_sdk
+
+# 指定 mac 地址
 python -m xiaozhi_sdk 00:22:44:66:88:00
+
+# 更多常用操作
+## --url 指定服务端 websocket 地址
+## --wake_word 指定唤醒词
+python -m xiaozhi_sdk 00:22:44:66:88:00 \
+  --url ws://127.0.0.1:8180 \
+  --wake_word="你好啊"
 ```
 
 ### 2. 编程使用 (高阶用法)

{xiaozhi_sdk-0.2.5 → xiaozhi_sdk-0.2.8}/xiaozhi_sdk.egg-info/SOURCES.txt

@@ -2,10 +2,13 @@ LICENSE
 MANIFEST.in
 README.md
 pyproject.toml
-file/audio/greet.wav
-file/audio/play_music.wav
-file/audio/say_hello.wav
-file/audio/take_photo.wav
+file/audio/16k_greet.wav
+file/audio/16k_play_music.wav
+file/audio/16k_say_hello.wav
+file/audio/16k_take_photo.wav
+file/audio/test_16k.wav
+file/audio/test_24k.wav
+file/audio/test_48k.wav
 file/image/leijun.jpg
 file/opus/linux-arm64-libopus.so
 file/opus/linux-x64-libopus.so
@@ -16,6 +19,7 @@ tests/test_iot.py
 tests/test_pic.py
 tests/test_wake_word.py
 tests/test_xiaozhi.py
+tests/test_xiaozhi_opus.py
 xiaozhi_sdk/__init__.py
 xiaozhi_sdk/__main__.py
 xiaozhi_sdk/cli.py
@@ -29,10 +33,13 @@ xiaozhi_sdk.egg-info/SOURCES.txt
 xiaozhi_sdk.egg-info/dependency_links.txt
 xiaozhi_sdk.egg-info/requires.txt
 xiaozhi_sdk.egg-info/top_level.txt
-xiaozhi_sdk/../file/audio/greet.wav
-xiaozhi_sdk/../file/audio/play_music.wav
-xiaozhi_sdk/../file/audio/say_hello.wav
-xiaozhi_sdk/../file/audio/take_photo.wav
+xiaozhi_sdk/../file/audio/16k_greet.wav
+xiaozhi_sdk/../file/audio/16k_play_music.wav
+xiaozhi_sdk/../file/audio/16k_say_hello.wav
+xiaozhi_sdk/../file/audio/16k_take_photo.wav
+xiaozhi_sdk/../file/audio/test_16k.wav
+xiaozhi_sdk/../file/audio/test_24k.wav
+xiaozhi_sdk/../file/audio/test_48k.wav
 xiaozhi_sdk/../file/image/leijun.jpg
 xiaozhi_sdk/../file/opus/linux-arm64-libopus.so
 xiaozhi_sdk/../file/opus/linux-x64-libopus.so

xiaozhi_sdk-0.2.5/tests/test_wake_word.py

@@ -1,33 +0,0 @@
-import asyncio
-import os
-import sys
-
-import pytest
-
-sys.path.insert(0, os.path.join(os.path.dirname(__file__), ".."))
-
-from xiaozhi_sdk import XiaoZhiWebsocket
-
-
-MAC_ADDR = "00:22:44:66:88:00"
-ota_url = None
-URL = None
-
-
-@pytest.mark.asyncio
-async def test_main():
-    is_end = asyncio.Event()
-    async def message_handler_callback(message):
-        if message.get("state") == "stop":
-            is_end.set()
-        print("message received:", message)
-
-    xiaozhi = XiaoZhiWebsocket(message_handler_callback, url=URL, ota_url=ota_url)
-    await xiaozhi.init_connection(MAC_ADDR)
-
-    await xiaozhi.send_wake_word("退下，拜拜不聊了")
-    await asyncio.wait_for(is_end.wait(), timeout=20.0)
-    await xiaozhi.send_wake_word("你好")
-
-    await asyncio.wait_for(is_end.wait(), timeout=20.0)
-    await xiaozhi.close()

xiaozhi_sdk-0.2.5/xiaozhi_sdk/config.py

@@ -1,3 +0,0 @@
-INPUT_SERVER_AUDIO_SAMPLE_RATE = 16000
-
-OTA_URL = "https://api.tenclass.net/xiaozhi/ota"

xiaozhi_sdk-0.2.5/xiaozhi_sdk/opus.py

@@ -1,61 +0,0 @@
-import math
-
-import av
-import numpy as np
-import opuslib
-
-from xiaozhi_sdk.config import INPUT_SERVER_AUDIO_SAMPLE_RATE
-
-
-class AudioOpus:
-
-    def __init__(self, sample_rate, channels):
-        self.sample_rate = sample_rate
-        self.channels = channels
-
-        # 创建 Opus 编码器
-        self.opus_encoder = opuslib.Encoder(
-            fs=sample_rate, channels=channels, application=opuslib.APPLICATION_VOIP  # 采样率  # 单声道  # 语音应用
-        )
-
-        # 创建 Opus 解码器
-        self.opus_decoder = opuslib.Decoder(
-            fs=INPUT_SERVER_AUDIO_SAMPLE_RATE,  # 采样率
-            channels=1,  # 单声道
-        )
-
-        self.resampler = av.AudioResampler(format="s16", layout="mono", rate=sample_rate)
-
-    async def pcm_to_opus(self, pcm):
-        pcm_array = np.frombuffer(pcm, dtype=np.int16)
-        pcm_bytes = pcm_array.tobytes()
-        return self.opus_encoder.encode(pcm_bytes, 960)
-
-    @staticmethod
-    def to_n_960(samples) -> np.ndarray:
-        n = math.ceil(samples.shape[0] / 960)
-        arr_padded = np.pad(samples, (0, 960 * n - samples.shape[0]), mode="constant", constant_values=0)
-        return arr_padded.reshape(n, 960)
-
-    async def change_sample_rate(self, pcm_array) -> np.ndarray:
-        if self.sample_rate == INPUT_SERVER_AUDIO_SAMPLE_RATE:
-            return self.to_n_960(pcm_array)
-
-        frame = av.AudioFrame.from_ndarray(np.array(pcm_array).reshape(1, -1), format="s16", layout="mono")
-        frame.sample_rate = INPUT_SERVER_AUDIO_SAMPLE_RATE  # Assuming input is 16kHz
-        resampled_frames = self.resampler.resample(frame)
-        samples = resampled_frames[0].to_ndarray().flatten()
-        new_frame = av.AudioFrame.from_ndarray(
-            samples.reshape(1, -1),
-            format="s16",
-            layout="mono",
-        )
-        new_frame.sample_rate = self.sample_rate
-        new_samples = new_frame.to_ndarray().flatten()
-        return self.to_n_960(new_samples)
-
-    async def opus_to_pcm(self, opus) -> np.ndarray:
-        pcm_data = self.opus_decoder.decode(opus, 960)
-        pcm_array = np.frombuffer(pcm_data, dtype=np.int16)
-        samples = await self.change_sample_rate(pcm_array)
-        return samples