use pybind

build.sh

@@ -1,17 +1,9 @@
-g++ -DBOOST_DEBUG_PYTHON -shared -fPIC \
-    -I/usr/include/python3.10 -I/usr/include/python3.10/numpy -I./include \
+g++ -shared -fPIC -std=c++11 \
+    $(python3 -m pybind11 --includes) \  # 自动获取 pybind11 和 Python 头文件路径
+    -I./include \                        # 保留自定义头文件路径
     -L./lib -L/usr/lib/x86_64-linux-gnu \
-    -DRTC_NUMPY_IMPL \
-    rtc_plugins.cpp util/RTCContext.cpp \
-    -lMRTCEngine -lboost_python310 -lboost_numpy310 -lpython3.10 \
-    -Wl,-rpath='$ORIGIN/lib' \
-    -o rtc_plugins.so
-#g++ -shared -fPIC \
-#    -I/usr/include/python3.10 -I/usr/include/python3.10/numpy -I./include \
-#    -I/usr/include/python3.10 -I/usr/include/numpy \
-#    -Wl,-rpath='$ORIGIN/lib' \
-#    -lboost_python310 -lpython3.10 \
-#    -lMRTCEngine -lboost_python310 -lpython3.10 \
-#    -L$(python3 -c "import numpy; print(numpy.get_include())") \
-#    rtc_plugins.cpp util/RTCContext.cpp \
-#    -o rtc_plugins.so
+    -lMRTCEngine \                       # 保留其他必要库
+    -lpython3.10 \                       # 仍然需要链接 Python
+    -Wl,-rpath='$ORIGIN/lib' \           # 保留运行时库路径
+    -o rtc_plugins$(python3-config --extension-suffix) \  # 标准化输出文件名
+    rtc_plugins.cpp util/RTCContext.cpp

rtc_plugins.cpp

@@ -1,7 +1,10 @@
-// rtc_plugins.cpp
+#define IMPLEMENT_NUMPY_API
+#include "util/RTCContext.h"
+#include <pybind11/pybind11.h>
+#include <pybind11/numpy.h>  // pybind11 的 NumPy 支持
+#include <pybind11/detail/common.h>
 
-#define IMPLEMENT_NUMPY_API  // 标记这是实现文件
-#include "util/numpyStub.h"
+namespace py = pybind11;
 #include "util/RTCContext.h"
 
 // 提供转换接口
@@ -9,10 +12,15 @@ void** get_numpy_api() {
     return (void**)RTC_PLUGINS_ARRAY_API;
 }
 
+// 初始化 NumPy（适配 pybind11）
+void init_numpy() {
+    if (import_array() < 0) {
+        throw py::bind_already_set();  // 自动捕获 NumPy 初始化错误
+    }
+    std::cout << "NumPy API addr: " << PyArray_API << std::endl;
+}
 
-namespace py = boost::python;
-
-int init(const char* selfUserId, const char* selfDisplayName, const char* selfRoomId, boost::python::object callback) {
+int init(const char* selfUserId, const char* selfDisplayName, const char* selfRoomId, py::object callback) {
     if (!PyArray_API) {
         std::cout << "PyArray_API is null in outer init" << std::endl;
     } else {
@@ -47,178 +55,57 @@ int initSend(const char* srcRoomId, const char* destRoomId, const int16_t destCh
         return -1;
     }
 }
-int getSize() {
-    return RTCContext::instance().getSize();
+// NumPy 数据交互（关键修改）
+py::array_t<int16_t> getNumpyData() {
+    return py::array_t<int16_t>(
+            RTCContext::instance().getNumpyData()  // 假设返回的是已有 NumPy 数组
+    );
 }
 
-namespace bp = boost::python;
-namespace np = boost::python::numpy;
-np::ndarray getNumpyData() {
-    return RTCContext::instance().getNumpyData();
+int sendCustomAudioData(int16_t destChannelIndex, py::array_t<int16_t> inputArray,
+                        int32_t sampleRate, uint64_t channelNum, uint64_t dataLen) {
+    py::gil_scoped_release release;
+
+    auto buf = inputArray.request();
+    if (buf.size != dataLen) {
+        throw py::value_error("Array length does not match dataLen");
     }
-bp::list getListData() {
+
+    return RTCContext::instance().sendCustomAudioData(
+            destChannelIndex, buf.ptr, sampleRate, channelNum, dataLen
+    );
+}
+py::list getListData() {
     return RTCContext::instance().getListData();
 }
-int16_t getDataCount() {
-    return RTCContext::instance().getDataCount();
-}
-py::object create_int16_array() {
-    // 1. 定义数组维度（1维，长度为 4）
-    npy_intp dims[1] = {4};
-
-    // 2. 创建原生 C 数组（int16_t 数据）
-    int16_t data[4] = {1, 2, -3, 4};  // 示例数据
-
-    // 3. 通过 NumPy C API 创建 PyObject*
-    PyObject* py_array = PyArray_SimpleNewFromData(
-            1,              // 维度数
-            dims,           // 各维度大小
-            NPY_INT16,      // 数据类型（np.int16）
-            data            // 数据指针
-    );
-
-    if (!py_array) {
-        throw std::runtime_error("Failed to create NumPy array");
-    }
-
-    // 4. 转换为 py::object（自动管理引用计数）
-    return py::object(py::handle<>(py_array));
-}
-
-int sendCustomAudioData(int16_t destChannelIndex, py::object pD,
-                        int32_t sampleRate, uint64_t channelNum, uint64_t dataLen) {
-    try {
-        // 强制转换为 int16 连续数组
-        PyObject* py_array = PyArray_FROM_OTF(
-                pD.ptr(),
-                NPY_INT16,
-                NPY_ARRAY_IN_ARRAY | NPY_ARRAY_FORCECAST
-        );
-        if (!py_array) {
-            throw std::runtime_error("Failed to convert input to int16 array");
-        }
-
-        // 修复点：使用花括号初始化
-        py::object arr{py::handle<>(py_array)};
-
-        // 检查数据长度
-        PyArrayObject* npArray = reinterpret_cast<PyArrayObject*>(arr.ptr());
-        if (PyArray_SIZE(npArray) != static_cast<npy_intp>(dataLen)) {
-            Py_DECREF(py_array);
-            throw std::runtime_error("Array length does not match dataLen");
-        }
-
-        // 处理数据...
-        void* dataPtr = PyArray_DATA(npArray);
-        int ret = RTCContext::instance().sendCustomAudioData(
-                destChannelIndex, dataPtr, sampleRate, channelNum, dataLen
-        );
-        Py_DECREF(py_array);  // 释放临时数组
-        return ret;
-    } catch (...) {
-        PyErr_SetString(PyExc_RuntimeError, "Invalid audio data");
-        return -1;
-    }
+int getSize() {
+    return RTCContext::instance().getSize();
 }
 RetAudioFrame getData() {
     return RTCContext::instance().getData();
 }
-
-/*
-int sendCustomAudioData(const int16_t destChannelIndex, py::object pyData, int32_t sampleRate, uint64_t channelNum,
-                        uint64_t dataLen) {
-    try {
-        //py::object pyData = create_int16_array();
-        std::cout << "step 1" << std::endl;
-        // 1. 检查输入有效性
-        if (pyData.ptr() == nullptr) {
-            throw std::runtime_error("Input data is NULL");
+int16_t getDataCount() {
+    return RTCContext::instance().getDataCount();
 }
 
-        std::cout << "step 2" << std::endl;
-        std::cout << "pyData ptr is:" << pyData.ptr() << std::endl;
-        if (!pyData.ptr() || !Py_IsInitialized() || !PyObject_TypeCheck(pyData.ptr(), &PyBaseObject_Type)) {
-            throw std::runtime_error("Invalid Python object");
-        }
-        std::cout << "step 2" << std::endl;
-        // 2. 检查是否是 numpy 数组
-        if (!PyArray_Check(pyData.ptr())) {
-            std::cout << "input is notnumpy" << std::endl;
-            throw std::runtime_error("Input is not a numpy array");
-        }
-
-        std::cout << "step 3" << std::endl;
-        // 3. 转换为 PyArrayObject
-        PyArrayObject* npArray = reinterpret_cast<PyArrayObject*>(pyData.ptr());
-
-        std::cout << "step 4" << std::endl;
-        // 4. 检查数据类型是否为 int16
-        if (PyArray_TYPE(npArray) != NPY_INT16) {
-            throw std::runtime_error("Array must be of type int16 (np.int16)");
-        }
-
-        std::cout << "step 5" << std::endl;
-        // 5. 检查数据是否连续
-        if (!PyArray_ISCONTIGUOUS(npArray)) {
-            throw std::runtime_error("Array must be contiguous in memory");
-        }
-        std::cout << "step 6" << std::endl;
-
-        // 6. 获取数据指针
-        void* dataPtr = PyArray_DATA(npArray);
-        if (dataPtr == nullptr) {
-            throw std::runtime_error("Invalid data pointer");
-        }
-        std::cout << "step 7" << std::endl;
-        return RTCContext::instance().sendCustomAudioData(destChannelIndex, dataPtr, sampleRate, channelNum, dataLen);
-
-    } catch (const std::exception& e) {
-        std::cout << "error:" << e.what() << std::endl;
-        return -1;
-    }
-}
- */
-
-
-void init_numpy() {
-    // 直接调用底层函数，绕过宏的问题
-    if (_import_array() < 0) {
-        PyErr_SetString(PyExc_ImportError, "numpy.core.multiarray failed to import");
-        throw std::runtime_error("NumPy initialization failed");
-    }
-    std::cout << "NumPy API addr: " << PyArray_API << std::endl;
-}
-BOOST_PYTHON_MODULE(rtc_plugins) {
-    try {
+PYBIND11_MODULE(rtc_plugins, m) {
     init_numpy();
-        void** numpyApi = (void**)PyArray_API;
-        if (!numpyApi || !numpyApi[93]) { // 93是PyArray_SimpleNew的偏移量
-            std::cout << "NumPy API corrupt! Key functions missing." << std::endl;
-            PyErr_Print();
-            throw std::runtime_error("Invalid NumPy API state");
-        } else {
-            RTCContext::instance().setNumpyApi(numpyApi);
-            std::cout << "set numpyApi succ:" << numpyApi[93] << std::endl;
-        }
+    // 可选：暴露 RetAudioFrame 类（需额外绑定）
+    py::class_<RetAudioFrame>(m, "RetAudioFrame")
+        .def_readwrite("data", &RetAudioFrame::data)
+    .def_readwrite("dataCount", &RetAudioFrame::dataCount)
+    .def_readwrite("sampleRate", &RetAudioFrame::sampleRate)
+    .def_readwrite("numChannels", &RetAudioFrame::numChannels)
+    .def_readwrite("channelIndex", &RetAudioFrame::channelIndex);
+    m.def("init", &init);
+    m.def("initRecv", &initRecv);
+    m.def("initSend", &initSend);
+    m.def("sendCustomAudioData", &sendCustomAudioData);
+    m.def("getSize", &getSize);
+    m.def("getData", &getData);
+    m.def("getNumpyData", &getNumpyData);
+    m.def("getListData", &getListData);
+    m.def("getDataCount", &getDataCount);
 
-        /*
-        if (!PyArray_API) {
-            std::cout << "PyArray_API is null" << std::endl;
-        } else {
-            std::cout << "PyArray_API is not null" << std::endl;
-        }
-         */
 
-        py::def("init", &init);
-        py::def("initRecv", &initRecv);
-        py::def("initSend", &initSend);
-        py::def("sendCustomAudioData", &sendCustomAudioData);
-        py::def("getSize", &getSize);
-        py::def("getData", &getData);
-        py::def("getNumpyData", &getNumpyData);
-        py::def("getListData", &getListData);
-        py::def("getDataCount", &getDataCount);
-    } catch (...) {
-        PyErr_SetString(PyExc_RuntimeError, "Module initialization failed");
-    }
 }

test_audio_r.py

@@ -0,0 +1,45 @@
+import rtc_plugins
+import time
+import sounddevice as sd
+import numpy as np
+import mmap
+import os
+from ctypes import c_int16
+import struct
+
+srcUserId = "srcUser12"
+destUserId = "destUser12"
+
+srcDisplayName = "srcDisplayName12"
+destDisplayName = "destDisplayName12"
+srcRoomId = "srcRoom12"
+#destRoomId = "destRoomId12"
+destRoomId = srcRoomId
+srcChannelIndex = 46
+destChannelIndex = 47
+def my_callback_r(shmName, dataSize, dataCount, sampleRate, numChannels, channelIndex):
+    print(f"my_callback_r, dataSize:{dataSize}, dataCount:{dataCount}, sampleRate:{sampleRate}, numChannels:{numChannels}, channelIndex:{channelIndex}")
+    print(f"data:{shmName}")
+    print("after my_callback_r")
+
+ret = rtc_plugins.init(destUserId, destDisplayName, destRoomId, my_callback_r)
+if ret != 0:
+    print(f"init fail, ret:{ret}")
+    exit(1)
+ret = rtc_plugins.initRecv(destRoomId, srcUserId, destChannelIndex)
+if ret != 0:
+    print(f"initRecv fail, ret:{ret}")
+    exit(1)
+#ret = rtc_plugins.initSend(destRoomId, srcRoomId, srcChannelIndex, 1)
+#if ret != 0:
+#    print(f"initSend fail, ret:{ret}")
+#    exit(1)
+
+sampleRate = 16000
+while True:
+    frame = rtc_plugins.getListData()
+    sd.play(frame, sampleRate)
+    sd.wait()
+    print(f"get frame:{frame}")
+
+    time.sleep(0.005)

test_audio_s.py

@@ -36,14 +36,14 @@ if audioData.dtype != np.int16:
 ret = rtc_plugins.sendCustomAudioData(destChannelIndex, audioData, sampleRate, 1, len(audioData))
 if ret != 0:
     print(f"send fail, ret:{ret}")
-ret = rtc_plugins.initRecv(srcRoomId, srcUserId, srcChannelIndex)
-if ret != 0:
-    print(f"initRecv fail, ret:{ret}")
-    exit(1)
-for i in range(100):
-    #ret = rtc_plugins.sendCustomAudioData(destChannelIndex, audioData, sampleRate, 1, len(audioData))
+#ret = rtc_plugins.initRecv(srcRoomId, srcUserId, srcChannelIndex)
 #if ret != 0:
-    #    print(f"send fail, ret:{ret}")
+#    print(f"initRecv fail, ret:{ret}")
+#    exit(1)
+for i in range(100):
+    ret = rtc_plugins.sendCustomAudioData(destChannelIndex, audioData, sampleRate, 1, len(audioData))
+    if ret != 0:
+        print(f"send fail, ret:{ret}")
 
     #size = rtc_plugins.getSize()
     #print(f"data size:{size}")

util/RTCContext.cpp

@@ -56,242 +56,7 @@ void RTCContext::onSoundLevelUpdate(const char* roomId, const char* peerId, uint
 {
     std::cout << "RTCContext::onSoundLevelUpdate()" << std::endl;
 }
-/*
-void RTCContext::onAudioProcess(const char* roomId, const char* peerId,
-                                mrtc::MRTCAudioFrame& audioFrame,
-                                mrtc::MRTCAudioSourceType audioSourceType)
-{
-    namespace py = boost::python;
 
-    PyGILState_STATE gstate = PyGILState_Ensure();
-
-    try {
-        std::cout << "-----------------------------------" << std::endl;
-        std::cout << "dataCount:" << audioFrame.dataCount << std::endl;
-        std::cout << "dataCount value: " << audioFrame.dataCount
-                  << " (max: " << std::numeric_limits<npy_intp>::max() << ")" << std::endl;
-
-
-        std::cout << "onAudioProcess, numpyApi_:" << numpyApi_[93] << std::endl;
-        if (!numpyApi_ || !numpyApi_[93]) { // 93是PyArray_SimpleNew的偏移量
-            std::cout << "numpyApi_ is null in onAudioProcess" << std::endl;
-        } else {
-            std::cout << "numpyApi_ is not null in onAudioProcess:" << numpyApi_[93] << std::endl;
-        }
-        //auto numpyApi = RTCContext::numpy_api();
-        std::cout << "step1" << std::endl;
-        if (!numpyApi_) {
-            PyGILState_Release(gstate);
-            throw std::runtime_error("NumPy C-API not initialized. Call import_array() in module init");
-        }
-        std::cout << "step2" << std::endl;
-
-        using PyArray_SimpleNew_t = PyObject*(*)(int, npy_intp*, int);
-
-        void* func_ptr = numpyApi_[93];
-        std::cout << "Raw function pointer: " << func_ptr << std::endl;
-
-        auto ptmp = (PyObject*(*)(int, npy_intp*, int))numpyApi_[93];
-        std::cout << "ptmp is:" << ptmp << std::endl;
-        std::cout << "Pointer sizes:\n"
-                  << "void*: " << sizeof(void*) << "\n"
-                  << "FunctionPtr: " << sizeof(PyObject*(*)(int, npy_intp*, int)) << std::endl;
-
-        // 2. 使用memcpy避免编译器优化问题
-        PyArray_SimpleNew_t PyArray_SimpleNew;
-        static_assert(sizeof(func_ptr) == sizeof(PyArray_SimpleNew),
-                      "Pointer size mismatch");
-        std::cout << "step3" << std::endl;
-        memcpy(&PyArray_SimpleNew, &func_ptr, sizeof(func_ptr));
-
-        //auto PyArray_SimpleNew = reinterpret_cast<PyArray_SimpleNew_t>(numpyApi_[93]);
-        std::cout << "step4, PyArray_SimpleNew:" << PyArray_SimpleNew << std::endl;
-
-        // 3. 严格校验输入数据
-        if (!audioFrame.data || audioFrame.dataCount <= 0) {
-            PyGILState_Release(gstate);
-            throw std::invalid_argument("Invalid audio frame data");
-        }
-        std::cout << "step5" << std::endl;
-
-        // 4. 安全创建维度数组（带边界检查）
-        if (audioFrame.dataCount > std::numeric_limits<npy_intp>::max()) {
-            PyGILState_Release(gstate);
-            throw std::overflow_error("Audio frame size exceeds maximum limit");
-        }
-        std::cout << "step6" << std::endl;
-        npy_intp dims[1] = {static_cast<npy_intp>(audioFrame.dataCount)};
-
-        std::cout << "step7" << std::endl;
-        // 5. 创建NumPy数组（带内存保护）
-        PyObject* pyArray = nullptr;
-        pyArray = PyArray_SimpleNew(1, dims, NPY_INT16);
-        std::cout << "step8" << std::endl;
-        if (!pyArray) {
-            PyGILState_Release(gstate);
-            throw std::bad_alloc();
-        }
-        std::cout << "step9" << std::endl;
-
-        // 6. 安全拷贝数据（带对齐检查）
-        if (reinterpret_cast<uintptr_t>(audioFrame.data) % alignof(int16_t) != 0) {
-            Py_DECREF(pyArray);
-            PyGILState_Release(gstate);
-            throw std::runtime_error("Unaligned audio data pointer");
-        }
-        std::cout << "step10" << std::endl;
-        std::memcpy(PyArray_DATA(reinterpret_cast<PyArrayObject*>(pyArray)),
-                    audioFrame.data,
-                    audioFrame.dataCount * sizeof(int16_t));
-
-        std::cout << "step11" << std::endl;
-        // 7. 执行回调（带引用计数保护）
-        if (!pyCallback_.is_none()) {
-            try {
-                pyCallback_(
-                        py::handle<>(pyArray),  // 自动管理引用
-                        audioFrame.dataCount,
-                        audioFrame.sampleRate,
-                        audioFrame.numChannels,
-                        audioFrame.channelIndex
-                );
-            } catch (...) {
-                Py_DECREF(pyArray);
-                throw;  // 重新抛出异常
-            }
-        }
-        std::cout << "step12" << std::endl;
-
-        // 8. 释放资源
-        Py_DECREF(pyArray);
-        std::cout << "step13" << std::endl;
-        PyGILState_Release(gstate);
-        std::cout << "step14" << std::endl;
-
-    } catch (const std::exception& e) {
-        std::cerr << "Audio process error: " << e.what() << std::endl;
-        PyErr_Print();
-    }
-    exit(0);
-}
- */
-/*
-void RTCContext::onAudioProcess(const char* roomId, const char* peerId,
-                                mrtc::MRTCAudioFrame& audioFrame,
-                                mrtc::MRTCAudioSourceType audioSourceType)
-{
-    namespace py = boost::python;
-    std::cout << "=== 开始音频处理 ===" << std::endl;
-
-    // 1. 获取GIL
-    std::cout << "[1] 获取GIL锁..." << std::endl;
-    PyGILState_STATE gstate = PyGILState_Ensure();
-
-    try {
-        // 2. 输入参数校验
-        std::cout << "[2] 检查输入参数..." << std::endl;
-        std::cout << "    dataCount: " << audioFrame.dataCount
-                  << " (max: " << std::numeric_limits<npy_intp>::max() << ")" << std::endl;
-
-        if (!audioFrame.data || audioFrame.dataCount <= 0) {
-            std::cout << "[ERROR] 无效音频数据指针或长度" << std::endl;
-            throw std::invalid_argument("Invalid audio frame data");
-        }
-
-        if (audioFrame.dataCount > std::numeric_limits<npy_intp>::max()) {
-            std::cout << "[ERROR] 数据长度超过最大值" << std::endl;
-            throw std::overflow_error("Audio frame size exceeds maximum limit");
-        }
-
-        // 3. 准备数组维度
-        std::cout << "[3] 准备数组维度..." << std::endl;
-        npy_intp dims[1] = {static_cast<npy_intp>(audioFrame.dataCount)};
-        std::cout << "    维度设置完成: [" << dims[0] << "]" << std::endl;
-
-        // 4. 检查NumPy API状态
-        std::cout << "[4] 检查NumPy API状态..." << std::endl;
-        std::cout << "    numpyApi_ 地址: " << numpyApi_ << std::endl;
-        if (!numpyApi_) {
-            throw std::runtime_error("NumPy C-API not initialized");
-        }
-
-        // 5. 获取PyArray_SimpleNew函数
-        std::cout << "[5] 获取PyArray_SimpleNew函数..." << std::endl;
-        using PyArray_SimpleNew_t = PyObject*(*)(int, npy_intp*, int);
-        PyArray_SimpleNew_t PyArray_SimpleNew =
-                reinterpret_cast<PyArray_SimpleNew_t>(numpyApi_[93]);
-        std::cout << "    函数地址: " << (void*)PyArray_SimpleNew << std::endl;
-
-        std::cout << "[5.1] 验证函数指针..." << std::endl;
-        void* func_ptr = numpyApi_[93];
-        if (reinterpret_cast<uintptr_t>(func_ptr) < 0x1000) { // 检查是否为合法地址
-            std::cerr << "非法函数指针: " << func_ptr << std::endl;
-            throw std::runtime_error("Invalid PyArray_SimpleNew pointer");
-        }
-        // 6. 创建NumPy数组
-        std::cout << "[6] 创建NumPy数组..." << std::endl;
-        PyObject* pyArray = PyArray_SimpleNew(1, dims, NPY_INT16);
-        std::cout << "    数组地址: " << pyArray << std::endl;
-
-        if (!pyArray) {
-            throw std::bad_alloc();
-        }
-
-        // 7. 检查内存对齐
-        std::cout << "[7] 检查内存对齐..." << std::endl;
-        std::cout << "    音频数据地址: " << (void*)audioFrame.data
-                  << " 对齐要求: " << alignof(int16_t) << std::endl;
-
-        if (reinterpret_cast<uintptr_t>(audioFrame.data) % alignof(int16_t) != 0) {
-            Py_DECREF(pyArray);
-            throw std::runtime_error("Unaligned audio data pointer");
-        }
-
-        // 8. 拷贝数据
-        std::cout << "[8] 拷贝音频数据..." << std::endl;
-        std::cout << "    目标地址: " << PyArray_DATA((PyArrayObject*)pyArray)
-                  << " 字节数: " << audioFrame.dataCount * sizeof(int16_t) << std::endl;
-
-        std::memcpy(PyArray_DATA((PyArrayObject*)pyArray),
-                    audioFrame.data,
-                    audioFrame.dataCount * sizeof(int16_t));
-
-        // 9. 执行回调
-        if (!pyCallback_.is_none()) {
-            std::cout << "[9] 准备执行Python回调..." << std::endl;
-            try {
-                pyCallback_(
-                        py::handle<>(pyArray),
-                        audioFrame.dataCount,
-                        audioFrame.sampleRate,
-                        audioFrame.numChannels,
-                        audioFrame.channelIndex
-                );
-                std::cout << "    回调执行成功" << std::endl;
-            } catch (...) {
-                std::cout << "[ERROR] 回调执行失败" << std::endl;
-                Py_DECREF(pyArray);
-                throw;
-            }
-        } else {
-            std::cout << "[9] 无回调函数设置" << std::endl;
-        }
-
-        // 10. 释放资源
-        std::cout << "[10] 释放资源..." << std::endl;
-        Py_DECREF(pyArray);
-        std::cout << "[11] 释放GIL..." << std::endl;
-        PyGILState_Release(gstate);
-        std::cout << "=== 音频处理完成 ===" << std::endl;
-
-    } catch (const std::exception& e) {
-        std::cout << "[EXCEPTION] 异常捕获: " << e.what() << std::endl;
-        PyGILState_Release(gstate);
-        PyErr_Print();
-        std::cerr << "Audio process error: " << e.what() << std::endl;
-    }
-}
- */
 void printTimestamp() {
     // 获取系统当前时间点
     auto now = std::chrono::system_clock::now();
@@ -310,173 +75,7 @@ void RTCContext::onAudioProcess(const char* roomId, const char* peerId,
                                 mrtc::MRTCAudioFrame& audioFrame,
                                 mrtc::MRTCAudioSourceType audioSourceType)
 {
-    //namespace py = boost::python;
-    //std::cout << "=== 开始音频处理(共享内存版) ===" << std::endl;
-    //std::cout << "audioFrame:" << audioFrame.dataCount << "," << audioFrame.sampleRate << "," <<
-    //           audioFrame.numChannels << "," << audioFrame.channelIndex << std::endl;
-
-    //printTimestamp();
     setData(audioFrame);
-    // 1. 获取GIL
-    //std::cout << "[1] 获取GIL锁..." << std::endl;
-    ////PyGILState_STATE gstate = PyGILState_Ensure();
-
-    //try {
-    //    // 2. 输入参数校验
-    //    std::cout << "[2] 检查输入参数..." << std::endl;
-    //    std::cout << "    dataCount: " << audioFrame.dataCount
-    //              << " (max: " << std::numeric_limits<npy_intp>::max() << ")" << std::endl;
-
-    //    if (!audioFrame.data || audioFrame.dataCount <= 0) {
-    //        std::cout << "[ERROR] 无效音频数据指针或长度" << std::endl;
-    //        throw std::invalid_argument("Invalid audio frame data");
-    //    }
-
-    //    const size_t data_size = audioFrame.dataCount * sizeof(int16_t);
-
-    //    // 3. 创建共享内存
-    //    std::cout << "[3] 创建共享内存..." << std::endl;
-    //    char shm_name[32];
-    //    //snprintf(shm_name, sizeof(shm_name), "/audio_shm_%d", getpid());
-    //    snprintf(shm_name, sizeof(shm_name), "/audio_shm_test");
-
-    //    int fd = shm_open(shm_name, O_CREAT | O_RDWR, 0666);
-    //    if (fd == -1) {
-    //        std::cout << "[ERROR] shm_open失败: " << strerror(errno) << std::endl;
-    //        throw std::runtime_error("Failed to create shared memory");
-    //    }
-    //    std::cout << "    共享内存fd: " << fd << " 名称: " << shm_name << std::endl;
-
-    //    // 4. 设置共享内存大小
-    //    std::cout << "[4] 设置共享内存大小..." << std::endl;
-    //    if (ftruncate(fd, data_size) == -1) {
-    //        close(fd);
-    //        std::cout << "[ERROR] ftruncate失败: " << strerror(errno) << std::endl;
-    //        throw std::runtime_error("Failed to resize shared memory");
-    //    }
-    //    std::cout << "    内存大小: " << data_size << " bytes" << std::endl;
-
-    //    // 5. 内存映射
-    //    std::cout << "[5] 内存映射..." << std::endl;
-    //    void* ptr = mmap(NULL, data_size, PROT_WRITE, MAP_SHARED, fd, 0);
-    //    if (ptr == MAP_FAILED) {
-    //        close(fd);
-    //        std::cout << "[ERROR] mmap失败: " << strerror(errno) << std::endl;
-    //        throw std::runtime_error("Failed to map shared memory");
-    //    }
-    //    std::cout << "    映射地址: " << ptr << std::endl;
-
-    //    namespace py = boost::python;
-    //    namespace np = boost::python::numpy;
-    //    // 6. 拷贝数据到共享内存
-    //    std::cout << "[6] 拷贝音频数据到共享内存..." << std::endl;
-    //    memcpy(ptr, audioFrame.data, data_size);
-    //    std::cout << "step1" << std::endl;
-    //    /*
-    //    npy_intp shape[1] = { static_cast<npy_intp>(audioFrame.dataCount) };
-    //    std::cout << "step2" << std::endl;
-    //    np::dtype dtype = np::dtype::get_builtin<int16_t>();
-    //    std::cout << "step3" << std::endl;
-    //    np::ndarray audioArray = np::from_data(
-    //            audioFrame.data,                // 数据指针
-    //            dtype,                          // 数据类型 (int16)
-    //            py::make_tuple(shape[0]),       // 形状 (1D)
-    //            py::make_tuple(sizeof(int16_t)), // 步长
-    //            py::object()                    // 所有者（Python管理）
-    //    );
-    //     */
-    //    std::cout << "    数据拷贝完成" << std::endl;
-
-    //    // 7. 执行回调
-    //    //if (!pyCallback_.is_none()) {
-    //    //    std::cout << "[7] 准备执行Python回调..." << std::endl;
-    //    //    // 增加引用计数防止提前释放
-    //    //    Py_INCREF(pyCallback_.ptr());
-    //    //    try {
-    //    //        std::cout << "    pyCallback_ type: " << Py_TYPE(pyCallback_.ptr())->tp_name << std::endl;
-    //    //        PyObject* repr = PyObject_Repr(pyCallback_.ptr());
-    //    //        if (repr) {
-    //    //            std::cout << "    pyCallback_ repr: " << PyUnicode_AsUTF8(repr) << std::endl;
-    //    //            Py_DECREF(repr);  // 必须手动释放
-    //    //        }
-    //    //        // 传递共享内存信息
-    //    //        pyCallback_(
-    //    //                py::str(shm_name),  // 共享内存名称
-    //    //                data_size,          // 数据大小
-    //    //                audioFrame.dataCount,
-    //    //                audioFrame.sampleRate,
-    //    //                audioFrame.numChannels,
-    //    //                audioFrame.channelIndex
-    //    //        );
-    //    //        /*
-    //    //        pyCallback_(
-    //    //                audioArray,                     // numpy 数组
-    //    //                data_size,          // 数据大小
-    //    //                audioFrame.dataCount,
-    //    //                audioFrame.sampleRate,
-    //    //                audioFrame.numChannels,
-    //    //                audioFrame.channelIndex
-    //    //        );
-    //    //         */
-    //    //        std::cout << "    after callback" << std::endl;
-    //    //        if (PyErr_Occurred()) {
-    //    //            PyObject *type, *value, *traceback;
-    //    //            PyErr_Fetch(&type, &value, &traceback);
-    //    //            if (value) {
-    //    //                PyObject* str = PyObject_Str(value);
-    //    //                if (str) {
-    //    //                    std::cerr << "Python Error: " << PyUnicode_AsUTF8(str) << std::endl;
-    //    //                    Py_DECREF(str);
-    //    //                }
-    //    //            }
-    //    //            Py_XDECREF(type);
-    //    //            Py_XDECREF(value);
-    //    //            Py_XDECREF(traceback);
-    //    //            //PyErr_Print();
-    //    //            throw std::runtime_error("Python callback error");
-    //    //        }
-    //    //        std::cout << "    回调执行成功" << std::endl;
-
-    //    //    } catch (const py::error_already_set& e) {
-    //    //        std::cerr << "[PYTHON ERROR] ";
-    //    //        PyErr_Print();  // 自动打印到stderr
-    //    //        // 可选：获取更详细的错误信息
-    //    //        if (PyErr_Occurred()) {
-    //    //            PyObject *type, *value, *traceback;
-    //    //            PyErr_Fetch(&type, &value, &traceback);
-    //    //            std::cerr << "Details: "
-    //    //                      << PyUnicode_AsUTF8(PyObject_Str(value)) << std::endl;
-    //    //            PyErr_Restore(type, value, traceback);
-    //    //        }
-    //    //        Py_DECREF(pyCallback_.ptr());
-    //    //    } catch (...) {
-    //    //        std::cout << "[ERROR] 回调执行失败" << std::endl;
-    //    //        munmap(ptr, data_size);
-    //    //        close(fd);
-    //    //        shm_unlink(shm_name);
-    //    //        Py_DECREF(pyCallback_.ptr());
-    //    //        throw;
-    //    //    }
-    //    //    Py_DECREF(pyCallback_.ptr());
-    //    //} else {
-    //    //    std::cout << "[7] 无回调函数设置" << std::endl;
-    //    //}
-
-    //    // 8. 释放资源
-    //    std::cout << "[8] 释放共享内存资源..." << std::endl;
-    //    munmap(ptr, data_size);
-    //    close(fd);
-    //    shm_unlink(shm_name);
-
-    //    std::cout << "[9] 释放GIL..." << std::endl;
-    //    //PyGILState_Release(gstate);
-    //    std::cout << "=== 音频处理完成 ===" << std::endl;
-
-    //} catch (const std::exception& e) {
-    //    std::cout << "[EXCEPTION] 异常捕获: " << e.what() << std::endl;
-    //    //PyGILState_Release(gstate);
-    //    std::cerr << "Audio process error: " << e.what() << std::endl;
-    //}
 }
 
 void RTCContext::onProducer(uint32_t msgId, mrtc::MRTCProducerInfo& info)
@@ -486,12 +85,6 @@ void RTCContext::onProducer(uint32_t msgId, mrtc::MRTCProducerInfo& info)
 }
 bool RTCContext::init(const char* selfUserId, const char* selfDisplayName, const char* selfRoomId)
 {
-    std::cout << "init, numpyApi_:" << numpyApi_[93] << std::endl;
-    if (!numpyApi_ || !numpyApi_[93]) { // 93是PyArray_SimpleNew的偏移量
-        std::cout << "numpyApi_ is null in init" << std::endl;
-    } else {
-        std::cout << "numpyApi_ is not null in init" << std::endl;
-    }
     mrtc::IMRTCEngineFactory * rtcFactory = mrtc::getMRTCEngineFactory();
     if (!rtcFactory)
     {
@@ -537,46 +130,16 @@ bool RTCContext::init(const char* selfUserId, const char* selfDisplayName, const
         std::cout << "RTCContext::instance().registerListener() failed" << std::endl;
         return false;
     }
-    namespace py = boost::python;
-    namespace np = boost::python::numpy;
-    Py_Initialize();          // 初始化 Python
-    np::initialize();
 
     return true;
 }
 bool RTCContext::initRecv(const char* destRoomId, const char* srcUserId, const int16_t destChannelIndex)
 {
-    std::cout << "initRecv, numpyApi_:" << numpyApi_[93] << std::endl;
-    if (!numpyApi_ || !numpyApi_[93]) { // 93是PyArray_SimpleNew的偏移量
-        std::cout << "numpyApi_ is null in initRecv" << std::endl;
-    } else {
-        std::cout << "numpyApi_ is not null in initRecv" << std::endl;
-    }
     while (!isOnConsumer_)
     {
         std::cout << "wait for OnConsumer" << std::endl;
         sleep(3);
     }
-    /*
-    std::cout << "registerSoundLevelListener" << std::endl;
-    int16_t ret1 = rtcEngine_->registerSoundLevelListener(mrtc::TYPE_AUDIO_SOURCE_CUSTOM, destRoomId,
-                                                          srcUserId, destChannelIndex, this);
-    if (0 != ret1)
-    {
-        std::cout << "RTCContext::instance().registerSoundLevelListener() inUser failed, ret:" << ret1;
-        return false;
-    }
-
-    std::cout << "muteAudio" << std::endl;
-    int16_t ret2 = rtcEngine_->muteAudio(destRoomId, srcUserId, mrtc::TYPE_AUDIO_SOURCE_CUSTOM, false, destChannelIndex);
-    if (0 != ret2)
-    {
-        std::cout << "RTCContext::instance().muteAudio() failed, ret:" << ret2;
-        return false;
-    }
-
-    std::cout << "init recv succ" << std::endl;
-     */
     return true;
 
 }
@@ -661,15 +224,11 @@ void RTCContext::setpData(void* pData)
     std::lock_guard<std::mutex> lock(mutex_);
     pData_ = pData;
 }
-void RTCContext::setPyCallback(boost::python::object callback) {
+void RTCContext::setPyCallback(py::object callback) {
     std::lock_guard<std::mutex> lock(mutex_);
     pyCallback_ = callback;
 }
-void RTCContext::setNumpyApi(void **numpyApi) {
-    std::lock_guard<std::mutex> lock(mutex_);
-    numpyApi_ = numpyApi;
-    std::cout << "setNupyApi, numpyApi_:" << numpyApi_[93] << std::endl;
-}
+
 void RTCContext::setData(const mrtc::MRTCAudioFrame& frame) {
     std::lock_guard<std::mutex> lock(dataMutex_);
     if (dataSize_ == totalSize_) {
@@ -699,8 +258,6 @@ RetAudioFrame RTCContext::getData() {
     }
     return {}; // 返回空对象
 }
-namespace bp = boost::python;
-namespace np = boost::python::numpy;
 np::ndarray RTCContext::getNumpyData() {
     std::cout << "step1" << std::endl;
     std::lock_guard<std::mutex> lock(dataMutex_);
@@ -710,37 +267,25 @@ np::ndarray RTCContext::getNumpyData() {
     std::cout << "step3" << std::endl;
     size_t length = frame.dataCount;    // 数据长度
     std::cout << "step4" << std::endl;
-
-    PyGILState_STATE gstate = PyGILState_Ensure();
-    np::ndarray result = np::empty(bp::make_tuple(length), np::dtype::get_builtin<int16_t>());
-    try {
     if (!dataPtr || length == 0) {
-            result = np::zeros(bp::make_tuple(0), np::dtype::get_builtin<int16_t>());
-        } else {
-            result = np::empty(bp::make_tuple(length), np::dtype::get_builtin<int16_t>());
-            std::memcpy(result.get_data(), dataPtr, length * sizeof(int16_t));
+        return py::array_t<int16_t>({0});  // 返回空数组
     }
-    } catch (...) {
-        PyGILState_Release(gstate);  // 异常时释放GIL
-        throw;
-    }
-    PyGILState_Release(gstate);
-    return result;
 
+    // 直接构造 pybind11 的 NumPy 数组（自动管理内存）
+    py::array_t<int16_t> result({static_cast<py::ssize_t>(length)});
+    auto buf = result.mutable_unchecked();
+    for (size_t i = 0; i < length; i++) {
+        buf[i] = dataPtr[i];
     }
-bp::list RTCContext::getListData() {
-    std::cout << "step1" << std::endl;
+    return result;
+}
+py::list RTCContext::getListData() {
     std::lock_guard<std::mutex> lock(dataMutex_);
     RetAudioFrame frame = getData();
-    std::cout << "step2" << std::endl;
-    int16_t* dataPtr = frame.data.get();  // 你的数据指针
-    std::cout << "step3" << std::endl;
-    size_t length = frame.dataCount;    // 数据长度
-    std::cout << "step4" << std::endl;
-    bp::list result;
-    if (dataPtr && length > 0) {
-        for (size_t i = 0; i < length; ++i) {
-            result.append(dataPtr[i]);  // 逐个元素添加（值传递）
+    py::list result;
+    if (frame.data) {
+        for (int i = 0; i < frame.dataCount; i++) {
+            result.append(frame.data.get()[i]);
         }
     }
     return result;

util/RTCContext.h

@@ -1,9 +1,6 @@
 // RTCContext.h
 #pragma once
 
-//#include "numpyConfig.h"
-#include "numpyStub.h"
-
 #include "IMRTCEngine.hpp"
 #include "MRTCEngineDefine.hpp"
 #include "IMRTCEngineFactory.hpp"
@@ -18,27 +15,18 @@
 #include <sys/stat.h>
 
 #include <Python.h>
-#include <boost/python.hpp>
-#include <boost/python/detail/wrap_python.hpp>
-#include <boost/python/numpy.hpp>
-#include <boost/python/detail/prefix.hpp>
-#include <boost/python/module.hpp>
-#include <boost/python/def.hpp>
-#include <numpy/ndarrayobject.h>
+// pybind11 头文件
+#include <pybind11/pybind11.h>
+#include <pybind11/numpy.h>
+#include <pybind11/stl.h>
+namespace py = pybind11;
 #include <numpy/arrayobject.h>
 
-
-// 必须声明外部变量（关键！）
-//#define NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION
-
-namespace bp = boost::python;
-namespace fs = std::filesystem;
-namespace np = boost::python::numpy;
 #define ENV_PRODUCT
 //#define SEND_MODE
 
 // 音频数据帧
-struct RetAudioFrame final
+struct RetAudioFrame
 {
     std::unique_ptr<int16_t[]> data;
     int dataCount = 0;
@@ -100,14 +88,13 @@ public:
     int16_t getSize();
     void setData(const mrtc::MRTCAudioFrame& frame);
     RetAudioFrame getData();
-    np::ndarray getNumpyData();
-    bp::list getListData();
+    py::array_t<int16_t> getNumpyData();
+    py::list getListData();
     int16_t getDataCount();
 
     void* getpData() const;
     void setpData(void* pData);
-    void setPyCallback(boost::python::object callback);
-    void setNumpyApi(void** numpyApi);
+    void setPyCallback(py::object callback);
 
     int16_t sendAudioData(uint8_t channelIndex = 0, const void* pData = nullptr, int32_t nSampleRate = 48000, uint64_t nNumberOfChannels = 2, uint64_t dataLength = 0);
     int16_t sendCustomAudioData(const int16_t channelIndex, void* customData, int32_t sampleRate,
@@ -126,8 +113,7 @@ private:
     bool isOnConsumer_ = false;
     bool isJoinMultiRoom_ = false;
     bool isMultiRoom_ = false;
-    boost::python::object pyCallback_;
-    void ** numpyApi_;
+    py::object pyCallback_;
     std::vector<RetAudioFrame> data_;
     mutable std::mutex dataMutex_;
     const int16_t totalSize_ = 100;

util/numpyStub.h

@@ -1,12 +0,0 @@
-// numpy_interface.h （新建）
-#pragma once
-
-#ifdef IMPLEMENT_NUMPY_API
-// 主模块实现路径
-    #define PY_ARRAY_UNIQUE_SYMBOL RTC_PLUGINS_ARRAY_API
-    #define NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION
-    #include <numpy/arrayobject.h>
-#else
-// 用户头文件路径
-extern void* RTC_PLUGINS_ARRAY_API[];  // 严格匹配NumPy类型
-#endif