xiaozhi-sdk 0.0.6__tar.gz → 0.0.7__tar.gz

{xiaozhi_sdk-0.0.6/xiaozhi_sdk.egg-info → xiaozhi_sdk-0.0.7}/PKG-INFO

@@ -1,14 +1,13 @@
 Metadata-Version: 2.4
 Name: xiaozhi-sdk
-Version: 0.0.6
+Version: 0.0.7
 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: MIT
+License-Expression: MIT
 Project-URL: Homepage, https://github.com/dairoot/xiaozhi-sdk
 Classifier: Programming Language :: Python :: 3
-Classifier: License :: OSI Approved :: MIT License
 Classifier: Operating System :: OS Independent
-Requires-Python: >=3.8.1
+Requires-Python: >=3.9
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: numpy
@@ -21,6 +20,8 @@ Requires-Dist: sounddevice
 Requires-Dist: python-socks
 Requires-Dist: click
 Requires-Dist: colorlog
+Requires-Dist: soundfile>=0.13.1
+Requires-Dist: pydub>=0.25.1
 Dynamic: license-file
 
 # 小智SDK (XiaoZhi SDK)
@@ -73,7 +74,11 @@ python -m xiaozhi_sdk 00:22:44:66:88:00
 ## ✅ 运行测试
 
 ```bash
-pytest tests/
+# 安装开发依赖
+uv sync --group dev
+
+# 运行测试
+uv run pytest
 ```
 
 

{xiaozhi_sdk-0.0.6 → xiaozhi_sdk-0.0.7}/README.md

@@ -48,7 +48,11 @@ python -m xiaozhi_sdk 00:22:44:66:88:00
 ## ✅ 运行测试
 
 ```bash
-pytest tests/
+# 安装开发依赖
+uv sync --group dev
+
+# 运行测试
+uv run pytest
 ```
 
 

{xiaozhi_sdk-0.0.6 → xiaozhi_sdk-0.0.7}/pyproject.toml

@@ -8,8 +8,8 @@ dynamic = ["version"]
 description = "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。"
 readme = "README.md"
 authors = [{name = "dairoot", email = "623815825@qq.com"}]
-license = {text = "MIT"}
-requires-python = ">=3.8.1"
+license = "MIT"
+requires-python = ">=3.9"
 dependencies = [
     "numpy",
     "websockets>=15.0.1",
@@ -21,10 +21,11 @@ dependencies = [
     "python-socks",
     "click",
     "colorlog",
+    "soundfile>=0.13.1",
+    "pydub>=0.25.1",
 ]
 classifiers = [
     "Programming Language :: Python :: 3",
-    "License :: OSI Approved :: MIT License",
     "Operating System :: OS Independent",
 ]
 

{xiaozhi_sdk-0.0.6 → xiaozhi_sdk-0.0.7}/tests/test_xiaozhi.py

@@ -27,7 +27,11 @@ async def assistant_audio_play(audio_queue, wait_time=5):
             continue
 
         pcm_data = audio_queue.popleft()
-        stream.write(pcm_data)
+
+        # 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()
@@ -59,10 +63,11 @@ def mcp_tool_func():
 
 async def message_handler_callback(message):
     print("message received:", message)
-    pass
+    if message["type"] == "music":
+        print("music:", message["text"])
 
 
-MAC_ADDR = "64:e8:13:18:21:1c"
+MAC_ADDR = "00:22:44:66:88:00"
 
 ota_url = "http://localhost:3080/api/ota"
 URL = "ws://120.79.156.134:8380"
@@ -80,16 +85,22 @@ async def test_main():
     await xiaozhi.set_mcp_tool_callback(mcp_tool_func())
     await xiaozhi.init_connection(MAC_ADDR)
 
-    # say hellow
-    for pcm in read_audio_file("./file/audio/say_hello.wav"):
-        await xiaozhi.send_audio(pcm)
-    await xiaozhi.send_silence_audio()
-    await assistant_audio_play(xiaozhi.output_audio_queue)
+    # # say hellow
+    # for pcm in read_audio_file("./file/audio/say_hello.wav"):
+    #     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"):
+    #     await xiaozhi.send_audio(pcm)
+    # await xiaozhi.send_silence_audio()
+    # await assistant_audio_play(xiaozhi.output_audio_queue, 5)
 
-    # say take photo
-    for pcm in read_audio_file("./file/audio/take_photo.wav"):
+    # 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, 5)
+    await assistant_audio_play(xiaozhi.output_audio_queue, 500)
 
     await xiaozhi.close()

{xiaozhi_sdk-0.0.6 → xiaozhi_sdk-0.0.7}/xiaozhi_sdk/__init__.py

@@ -1,3 +1,3 @@
-__version__ = "0.0.6"
+__version__ = "0.0.7"
 
 from xiaozhi_sdk.core import XiaoZhiWebsocket  # noqa

{xiaozhi_sdk-0.0.6 → xiaozhi_sdk-0.0.7}/xiaozhi_sdk/cli.py

@@ -87,9 +87,11 @@ class XiaoZhiClient:
         """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, send_wake=True)
+
         await self.xiaozhi.init_connection(
             self.mac_address, aec=False, serial_number=serial_number, license_key=license_key
         )
+
         asyncio.create_task(play_assistant_audio(self.xiaozhi.output_audio_queue))
 
     def audio_callback(self, indata, frames, time, status):

{xiaozhi_sdk-0.0.6 → xiaozhi_sdk-0.0.7}/xiaozhi_sdk/core.py

@@ -5,7 +5,7 @@ import os
 import re
 import uuid
 from collections import deque
-from typing import Any, Callable, Dict, Optional
+from typing import Any, Callable, Deque, Dict, Optional
 
 import websockets
 
@@ -13,6 +13,7 @@ from xiaozhi_sdk.config import INPUT_SERVER_AUDIO_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.mcp_tool import async_mcp_play_music, async_search_custom_music
 
 setup_opus()
 from xiaozhi_sdk.opus import AudioOpus
@@ -54,12 +55,17 @@ class XiaoZhiWebsocket(McpTool):
         self.message_handler_task: Optional[asyncio.Task] = None
 
         # 输出音频
-        self.output_audio_queue: deque[bytes] = deque()
+        self.output_audio_queue: Deque[bytes] = deque()
+        self.is_playing: bool = False
 
         # OTA设备
         self.ota: Optional[OtaDevice] = None
         self.iot_task: Optional[asyncio.Task] = None
         self.wait_device_activated: bool = False
+        self.tool_func = {
+            "async_play_custom_music": async_mcp_play_music,
+            "async_search_custom_music": async_search_custom_music,
+        }
 
     async def _send_hello(self, aec: bool) -> None:
         """发送hello消息"""
@@ -148,6 +154,13 @@ class XiaoZhiWebsocket(McpTool):
         elif message_type == "mcp":
             await self.mcp(data)
         elif self.message_handler_callback:
+            if data["type"] == "tts":
+                if data["state"] == "sentence_start":
+                    self.is_playing = True
+                    # self.output_audio_queue.clear()
+                else:
+                    self.is_playing = False
+
             await self.message_handler_callback(data)
 
     async def _message_handler(self) -> None:
@@ -164,7 +177,7 @@ class XiaoZhiWebsocket(McpTool):
 
     async def set_mcp_tool_callback(self, tool_func: Dict[str, Callable[..., Any]]) -> None:
         """设置MCP工具回调函数"""
-        self.tool_func = tool_func
+        self.tool_func.update(tool_func)
 
     async def connect_websocket(self, websocket_token):
         """连接websocket"""

{xiaozhi_sdk-0.0.6 → xiaozhi_sdk-0.0.7}/xiaozhi_sdk/mcp.py

@@ -1,9 +1,12 @@
+import asyncio
 import json
 import logging
 
+import numpy as np
 import requests
 
-from xiaozhi_sdk.data import mcp_initialize_payload, mcp_tool_conf, mcp_tools_payload
+from xiaozhi_sdk.utils.mcp_data import mcp_initialize_payload, mcp_tool_conf, mcp_tools_payload
+from xiaozhi_sdk.utils.mcp_tool import _get_random_music_info
 
 logger = logging.getLogger("xiaozhi_sdk")
 
@@ -16,6 +19,7 @@ class McpTool(object):
         self.explain_token = ""
         self.websocket = None
         self.tool_func = {}
+        self.is_playing = False
 
     def get_mcp_json(self, payload: dict):
         return json.dumps({"session_id": self.session_id, "type": "mcp", "payload": payload})
@@ -45,14 +49,46 @@ class McpTool(object):
             return res_json, True
         return res_json, False
 
+    async def play_custom_music(self, tool_func, arguments):
+        pcm_array, is_error = await tool_func(arguments)
+        while True:
+            if not self.is_playing:
+                break
+            await asyncio.sleep(0.1)
+        pcm_array = await self.audio_opus.change_sample_rate(np.array(pcm_array))
+        self.output_audio_queue.extend(pcm_array)
+
     async def mcp_tool_call(self, mcp_json: dict):
         tool_name = mcp_json["params"]["name"]
         tool_func = self.tool_func[tool_name]
+        arguments = mcp_json["params"]["arguments"]
         try:
-            tool_res, is_error = tool_func(mcp_json["params"]["arguments"])
+            if tool_name == "async_play_custom_music":
+
+                # v1 返回 url
+                music_info = await _get_random_music_info(arguments["id_list"])
+                if not music_info.get("url"):
+                    tool_res, is_error = {"message": "播放失败"}, True
+                else:
+                    tool_res, is_error = {"message": "正在为你播放: {}".format(arguments["music_name"])}, False
+                    data = {
+                        "type": "music", "state": "start",
+                        "url": music_info["url"],
+                        "text": arguments["music_name"],
+                        "source": "sdk.mcp_music_tool"
+                    }
+                    await self.message_handler_callback(data)
+
+                # v2 音频放到输出
+                # asyncio.create_task(self.play_custom_music(tool_func, arguments))
+
+            elif tool_name.startswith("async_"):
+                tool_res, is_error = await tool_func(arguments)
+            else:
+                tool_res, is_error = tool_func(arguments)
         except Exception as e:
             logger.error("[MCP] tool_func error: %s", e)
-            return
+            return self._build_response(mcp_json["id"], "工具调用失败", True)
 
         if tool_name == "take_photo":
             tool_res, is_error = await self.analyze_image(tool_res, mcp_json["params"]["arguments"]["question"])
@@ -84,8 +120,9 @@ class McpTool(object):
             for name, func in self.tool_func.items():
                 if func:
                     tool_list.append(name)
-                    mcp_tool_conf[name]["name"] = name
-                    mcp_tools_payload["result"]["tools"].append(mcp_tool_conf[name])
+                    target_name = name.removeprefix("async_")
+                    mcp_tool_conf[target_name]["name"] = name
+                    mcp_tools_payload["result"]["tools"].append(mcp_tool_conf[target_name])
             await self.websocket.send(self.get_mcp_json(mcp_tools_payload))
             logger.debug("[MCP] 加载成功，当前可用工具列表为：%s", tool_list)
 

{xiaozhi_sdk-0.0.6 → xiaozhi_sdk-0.0.7}/xiaozhi_sdk/opus.py

@@ -1,3 +1,5 @@
+import math
+
 import av
 import numpy as np
 import opuslib
@@ -29,11 +31,16 @@ class AudioOpus:
         pcm_bytes = pcm_array.tobytes()
         return self.opus_encoder.encode(pcm_bytes, 960)
 
-    async def change_sample_rate(self, pcm_array):
+    @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 pcm_array.reshape(1, 960)
+            return self.to_n_960(pcm_array)
 
-        c = int(self.sample_rate / INPUT_SERVER_AUDIO_SAMPLE_RATE)
         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)
@@ -45,10 +52,9 @@ class AudioOpus:
         )
         new_frame.sample_rate = self.sample_rate
         new_samples = new_frame.to_ndarray().flatten()
-        arr_padded = np.pad(new_samples, (0, 960 * c - new_samples.shape[0]), mode="constant", constant_values=0)
-        return arr_padded.reshape(c, 960)
+        return self.to_n_960(new_samples)
 
-    async def opus_to_pcm(self, opus):
+    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)

xiaozhi_sdk-0.0.6/xiaozhi_sdk/utils.py → xiaozhi_sdk-0.0.7/xiaozhi_sdk/utils/__init__.py

@@ -45,11 +45,11 @@ def setup_opus():
                 arch = "x64"
 
             if system == "darwin":  # macOS
-                return f"{current_dir}/../file/opus/macos-{arch}-libopus.dylib"
+                return f"{current_dir}/../../file/opus/macos-{arch}-libopus.dylib"
             elif system == "windows":  # Windows
-                return f"{current_dir}/../file/opus/windows-{arch}-opus.dll"
+                return f"{current_dir}/../../file/opus/windows-opus.dll"
             elif system == "linux":  # Linux
-                return f"{current_dir}/../file/opus/linux-{arch}-libopus.so"
+                return f"{current_dir}/../../file/opus/linux-{arch}-libopus.so"
             else:
                 # 默认情况，尝试系统查找
                 return ctypes.util.find_library(name)

xiaozhi_sdk-0.0.6/xiaozhi_sdk/data.py → xiaozhi_sdk-0.0.7/xiaozhi_sdk/utils/mcp_data.py

@@ -11,6 +11,22 @@ mcp_initialize_payload: Dict[str, Any] = {
 }
 
 mcp_tool_conf: Dict[str, Dict[str, Any]] = {
+    "search_custom_music": {
+        "description": "Search music and get music IDs. Use this tool when the user asks to search or play music. This tool returns a list of music with their IDs, which are required for playing music. Args:\n  `music_name`: The name of the music to search\n  `author_name`: The name of the music author (optional)",
+        "inputSchema": {
+            "type": "object",
+            "properties": {"music_name": {"type": "string"}, "author_name": {"type": "string"}},
+            "required": ["music_name"],
+        },
+    },
+    "play_custom_music": {
+        "description": "Play music using music IDs. IMPORTANT: You must call `search_custom_music` first to get the music IDs before using this tool. Use this tool after getting music IDs from search results. Args:\n  `id_list`: The id list of the music to play (obtained from search_custom_music results). The list must contain more than 2 music IDs, and the system will randomly select one to play.\n  `music_name`: The name of the music (obtained from search_custom_music results)",
+        "inputSchema": {
+            "type": "object",
+            "properties": {"music_name": {"type": "string"}, "id_list": {"type": "array", "items": {"type": "string"}, "minItems": 3}},
+            "required": ["music_name", "id_list"],
+        },
+    },
     "get_device_status": {
         "description": "Provides the real-time information of the device, including the current status of the audio speaker, screen, battery, network, etc.\nUse this tool for: \n1. Answering questions about current condition (e.g. what is the current volume of the audio speaker?)\n2. As the first step to control the device (e.g. turn up / down the volume of the audio speaker, etc.)",
         "inputSchema": {"type": "object", "properties": {}},

xiaozhi_sdk-0.0.7/xiaozhi_sdk/utils/mcp_tool.py

@@ -0,0 +1,88 @@
+import io
+import random
+
+import aiohttp
+import numpy as np
+from pydub import AudioSegment
+
+
+async def async_search_custom_music(data) -> tuple[dict, bool]:
+    search_url = f"https://music-api.gdstudio.xyz/api.php?types=search&name={data['music_name']}&count=100&pages=1"
+
+    # 为搜索请求设置 10 秒超时
+    timeout = aiohttp.ClientTimeout(total=10)
+    async with aiohttp.ClientSession(timeout=timeout) as session:
+        async with session.get(search_url) as response:
+            response_json = await response.json()
+
+    music_list = []
+    first_music_list = []
+    other_music_list1 = []
+    other_music_list2 = []
+    for line in response_json:
+        if data.get("author_name") and data["author_name"] in line["artist"][0]:
+            first_music_list.append(line)
+        elif data.get("author_name") and (data["author_name"] in line["artist"] or data["author_name"] in line["name"]):
+            other_music_list1.append(line)
+        else:
+            other_music_list2.append(line)
+
+    if len(first_music_list) <= 10:
+        music_list = first_music_list
+        random.shuffle(other_music_list2)
+        music_list = music_list + other_music_list1[: 20 - len(music_list)]
+        music_list = music_list + other_music_list2[: 20 - len(music_list)]
+
+    # print(data)
+    # print("找到音乐，数量：", len(first_music_list), len(music_list))
+
+    if not music_list:
+        return {}, False
+    return {"message": "已找到歌曲", "music_list": music_list}, False
+
+
+async def _get_random_music_info(id_list: list) -> dict:
+    timeout = aiohttp.ClientTimeout(total=10)
+    async with aiohttp.ClientSession(timeout=timeout) as session:
+        random.shuffle(id_list)
+
+        for music_id in id_list:
+            url = f"https://music-api.gdstudio.xyz/api.php?types=url&id={music_id}&br=128"
+            async with session.get(url) as response:
+                res_json = await response.json()
+            if res_json.get("url"):
+                break
+
+    return res_json
+
+
+async def async_mcp_play_music(data) -> tuple[list, bool]:
+    id_list = data["id_list"]
+    res_json = await _get_random_music_info(id_list)
+
+    if not res_json:
+        return [], False
+
+    pcm_list = []
+    buffer = io.BytesIO()
+    # 为下载音乐文件设置 60 秒超时（音乐文件可能比较大）
+    download_timeout = aiohttp.ClientTimeout(total=60)
+    async with aiohttp.ClientSession(timeout=download_timeout) as session:
+        async with session.get(res_json["url"]) as resp:
+            async for chunk in resp.content.iter_chunked(1024):
+                buffer.write(chunk)
+
+    buffer.seek(0)
+    audio = AudioSegment.from_mp3(buffer)
+    audio = audio.set_frame_rate(16000).set_channels(1).set_sample_width(2)  # 2 bytes = 16 bits
+    pcm_data = audio.raw_data
+
+    chunk_size = 960 * 2
+    for i in range(0, len(pcm_data), chunk_size):
+        chunk = pcm_data[i: i + chunk_size]
+
+        if chunk:  # 确保不添加空块
+            chunk = np.frombuffer(chunk, dtype=np.int16)
+            pcm_list.extend(chunk)
+
+    return pcm_list, False

{xiaozhi_sdk-0.0.6 → xiaozhi_sdk-0.0.7/xiaozhi_sdk.egg-info}/PKG-INFO

@@ -1,14 +1,13 @@
 Metadata-Version: 2.4
 Name: xiaozhi-sdk
-Version: 0.0.6
+Version: 0.0.7
 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: MIT
+License-Expression: MIT
 Project-URL: Homepage, https://github.com/dairoot/xiaozhi-sdk
 Classifier: Programming Language :: Python :: 3
-Classifier: License :: OSI Approved :: MIT License
 Classifier: Operating System :: OS Independent
-Requires-Python: >=3.8.1
+Requires-Python: >=3.9
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: numpy
@@ -21,6 +20,8 @@ Requires-Dist: sounddevice
 Requires-Dist: python-socks
 Requires-Dist: click
 Requires-Dist: colorlog
+Requires-Dist: soundfile>=0.13.1
+Requires-Dist: pydub>=0.25.1
 Dynamic: license-file
 
 # 小智SDK (XiaoZhi SDK)
@@ -73,7 +74,11 @@ python -m xiaozhi_sdk 00:22:44:66:88:00
 ## ✅ 运行测试
 
 ```bash
-pytest tests/
+# 安装开发依赖
+uv sync --group dev
+
+# 运行测试
+uv run pytest
 ```
 
 

{xiaozhi_sdk-0.0.6 → xiaozhi_sdk-0.0.7}/xiaozhi_sdk.egg-info/SOURCES.txt

@@ -3,6 +3,7 @@ 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/image/leijun.jpg
@@ -10,7 +11,7 @@ file/opus/linux-arm64-libopus.so
 file/opus/linux-x64-libopus.so
 file/opus/macos-arm64-libopus.dylib
 file/opus/macos-x64-libopus.dylib
-file/opus/windows-x86_64-opus.dll
+file/opus/windows-opus.dll
 tests/test_iot.py
 tests/test_pic.py
 tests/test_xiaozhi.py
@@ -19,17 +20,16 @@ xiaozhi_sdk/__main__.py
 xiaozhi_sdk/cli.py
 xiaozhi_sdk/config.py
 xiaozhi_sdk/core.py
-xiaozhi_sdk/data.py
 xiaozhi_sdk/iot.py
 xiaozhi_sdk/mcp.py
 xiaozhi_sdk/opus.py
-xiaozhi_sdk/utils.py
 xiaozhi_sdk.egg-info/PKG-INFO
 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/image/leijun.jpg
@@ -37,4 +37,7 @@ xiaozhi_sdk/../file/opus/linux-arm64-libopus.so
 xiaozhi_sdk/../file/opus/linux-x64-libopus.so
 xiaozhi_sdk/../file/opus/macos-arm64-libopus.dylib
 xiaozhi_sdk/../file/opus/macos-x64-libopus.dylib
-xiaozhi_sdk/../file/opus/windows-x86_64-opus.dll
+xiaozhi_sdk/../file/opus/windows-opus.dll
+xiaozhi_sdk/utils/__init__.py
+xiaozhi_sdk/utils/mcp_data.py
+xiaozhi_sdk/utils/mcp_tool.py

{xiaozhi_sdk-0.0.6 → xiaozhi_sdk-0.0.7}/xiaozhi_sdk.egg-info/requires.txt

@@ -8,3 +8,5 @@ sounddevice
 python-socks
 click
 colorlog
+soundfile>=0.13.1
+pydub>=0.25.1