Refactor `main.py`: remove redundant code, modularize functionality, and set multiprocessing to `spawn` for PyInstaller compatibility.

main.py

@@ -1,311 +1,56 @@
-import argparse
+import multiprocessing
+import os
 import sys
-import time
 import typing
 from pathlib import Path
-import multiprocessing
 
-import uvicorn
+if __name__ == '__main__':
+    if hasattr(sys, '_voice_dialogue_started'):
+        sys.exit(0)
+    sys._voice_dialogue_started = True
+
+    # 设置multiprocessing启动方法为spawn，避免fork问题
+    if hasattr(multiprocessing, 'set_start_method'):
+        try:
+            multiprocessing.set_start_method('spawn', force=True)
+        except RuntimeError:
+            pass
+
+    # Pyinstaller 多进程支持
+    multiprocessing.freeze_support()
+
+    # 禁用各种可能导致多进程问题的并行处理
+    os.environ.update({
+        "TOKENIZERS_PARALLELISM": "false",
+        # "OMP_NUM_THREADS": "1",
+        # "MKL_NUM_THREADS": "1", 
+        # "NUMEXPR_NUM_THREADS": "1",
+        # "OPENBLAS_NUM_THREADS": "1",
+        # "VECLIB_MAXIMUM_THREADS": "1",
+        # "BLIS_NUM_THREADS": "1",
+        # # 禁用huggingface的多进程
+        # "HF_HUB_DISABLE_PROGRESS_BARS": "1",
+        # "TRANSFORMERS_NO_ADVISORY_WARNINGS": "1",
+        # # 禁用torch的多进程
+        # "TORCH_NUM_THREADS": "1",
+        # "PYTORCH_JIT": "0",
+        # # 禁用joblib的loky后端，使用threading
+        # "JOBLIB_START_METHOD": "threading",
+        # "SKLEARN_JOBLIB_START_METHOD": "threading",
+    })
 
 HERE = Path(__file__).parent
 lib_path = HERE / "src"
 if lib_path.exists() and lib_path.as_posix() not in sys.path:
     sys.path.insert(0, lib_path.as_posix())
 
-from voice_dialogue.core.constants import (
-    audio_frames_queue,
-    user_voice_queue,
-    transcribed_text_queue,
-    text_input_queue,
-    audio_output_queue
-)
-from voice_dialogue.services.audio.capture import EchoCancellingAudioCapture
-from voice_dialogue.services.audio.generator import TTSAudioGenerator
-from voice_dialogue.services.audio.generators.models import tts_config_registry
-from voice_dialogue.services.audio.player import AudioStreamPlayer
-from voice_dialogue.services.speech.monitor import SpeechStateMonitor
-from voice_dialogue.services.speech.recognizer import ASRWorker
-from voice_dialogue.services.text.generator import LLMResponseGenerator
+from voice_dialogue.core.launcher import launch_system
+from voice_dialogue.cli.args import create_argument_parser
+from voice_dialogue.api.server import launch_api_server
 
 language: typing.Literal['zh', 'en'] = 'en'
 
 
-def launch_system(
-        user_language: str,
-        speaker: str
-) -> None:
-    """
-    启动完整的语音对话系统
-
-    该函数负责启动并协调语音对话系统的所有组件，包括音频采集、语音识别、
-    文本生成、语音合成和音频播放等功能模块。系统采用多线程架构，各组件
-    通过队列进行数据传递和通信。
-
-    系统工作流程：
-    1. 音频采集：EchoCancellingAudioCapture 采集用户语音并进行回声消除
-    2. 语音监测：SpeechStateMonitor 检测用户是否在说话
-    3. 语音识别：ASRWorker 将用户语音转换为文本
-    4. 文本生成：LLMResponseGenerator 基于用户问题生成AI回答
-    5. 语音合成：TTSAudioGenerator 将AI回答转换为语音
-    6. 音频播放：AudioStreamPlayer 播放生成的语音
-
-    Args:
-        user_language (str): 用户语言，支持 'zh'（中文）和 'en'（英文）
-        speaker (str): 语音合成使用的说话人，支持：
-                      '罗翔', '马保国', '沈逸', '杨幂', '周杰伦', '马云'
-
-    Raises:
-        ValueError: 当指定的说话人不在支持列表中时抛出异常
-
-    Returns:
-        None: 函数会一直运行直到所有线程结束
-
-    Note:
-        该函数会阻塞运行，直到系统被外部停止或发生异常
-    """
-
-    threads = []
-    #
-    audio_frame_probe = EchoCancellingAudioCapture(audio_frames_queue=audio_frames_queue)
-    audio_frame_probe.start()
-    threads.append(audio_frame_probe)
-
-    #
-    user_voice_checker = SpeechStateMonitor(
-        audio_frame_queue=audio_frames_queue,
-        user_voice_queue=user_voice_queue,
-    )
-    user_voice_checker.start()
-    threads.append(user_voice_checker)
-
-    #
-    whisper_worker = ASRWorker(
-        user_voice_queue=user_voice_queue, transcribed_text_queue=transcribed_text_queue,
-        language=user_language
-    )
-    whisper_worker.start()
-    threads.append(whisper_worker)
-
-    answer_generator_worker = LLMResponseGenerator(
-        user_question_queue=transcribed_text_queue,
-        generated_answer_queue=text_input_queue
-    )
-    answer_generator_worker.start()
-    threads.append(answer_generator_worker)
-
-    # 动态获取TTS配置，而不是使用固定映射
-    tts_speaker_config = _get_tts_config_by_speaker_name(speaker)
-    if tts_speaker_config is None:
-        # 如果找不到指定说话人，列出所有可用说话人并抛出异常
-        available_speakers = _get_available_speaker_names()
-        raise ValueError(f"不支持的TTS说话人: {speaker}。可用说话人: {', '.join(available_speakers)}")
-
-    audio_generator_worker = TTSAudioGenerator(
-        text_input_queue=text_input_queue,
-        audio_output_queue=audio_output_queue,
-        tts_config=tts_speaker_config
-    )
-    audio_generator_worker.start()
-    threads.append(audio_generator_worker)
-
-    audio_playing_worker = AudioStreamPlayer(audio_playing_queue=audio_output_queue)
-    audio_playing_worker.start()
-    threads.append(audio_playing_worker)
-
-    while not all([thread.is_ready for thread in threads]):
-        time.sleep(0.1)
-
-    # audio_frame_probe.start_record()
-    print(f'{"=" * 80}\n服务启动成功\n{"=" * 80}')
-    for thread in threads:
-        thread.join()
-
-
-def _get_tts_config_by_speaker_name(speaker_name: str):
-    """
-    根据说话人名称获取TTS配置
-
-    支持中文名称和英文名称，优先匹配中文名称映射，
-    如果找不到则直接使用英文名称搜索
-
-    Args:
-        speaker_name (str): 说话人名称
-
-    Returns:
-        BaseTTSConfig: TTS配置，如果找不到则返回None
-    """
-    # 中文名称到英文名称的映射（保持向后兼容）
-    chinese_to_english_mapping = {
-        '罗翔': 'Luo Xiang',
-        '马保国': 'Ma Baoguo',
-        '沈逸': 'Shen Yi',
-        '杨幂': 'Yang Mi',
-        '周杰伦': 'Zhou Jielun',
-        '马云': 'Ma Yun',
-    }
-
-    # 首先尝试中文名称映射
-    english_name = chinese_to_english_mapping.get(speaker_name, speaker_name)
-
-    # 获取所有可用配置
-    all_configs = tts_config_registry.get_all_configs()
-
-    # 搜索匹配的配置
-    for config in all_configs:
-        if config.character_name == english_name:
-            return config
-
-    # 如果通过映射找不到，尝试直接匹配输入的名称
-    if speaker_name != english_name:
-        for config in all_configs:
-            if config.character_name == speaker_name:
-                return config
-
-    return None
-
-
-def _get_available_speaker_names():
-    """
-    获取所有可用的说话人名称列表
-
-    Returns:
-        list[str]: 包含中文显示名称和英文原始名称的列表
-    """
-    # 中文显示名称映射
-    english_to_chinese_mapping = {
-        'Luo Xiang': '罗翔',
-        'Ma Baoguo': '马保国',
-        'Shen Yi': '沈逸',
-        'Yang Mi': '杨幂',
-        'Zhou Jielun': '周杰伦',
-        'Ma Yun': '马云',
-    }
-
-    all_configs = tts_config_registry.get_all_configs()
-    speaker_names = []
-
-    for config in all_configs:
-        # 优先显示中文名称
-        chinese_name = english_to_chinese_mapping.get(config.character_name)
-        if chinese_name:
-            speaker_names.append(chinese_name)
-        else:
-            # 如果没有中文映射，使用英文原名
-            speaker_names.append(config.character_name)
-
-    return sorted(speaker_names)
-
-
-def _update_argument_parser_speaker_choices():
-    """
-    动态更新命令行参数解析器中的说话人选项
-
-    Returns:
-        list[str]: 可用的说话人选择列表
-    """
-    return _get_available_speaker_names()
-
-
-def create_argument_parser():
-    """创建命令行参数解析器"""
-    # 动态获取可用说话人列表
-    available_speakers = _update_argument_parser_speaker_choices()
-
-    parser = argparse.ArgumentParser(
-        description="VoiceDialogue - 语音对话系统",
-        formatter_class=argparse.RawDescriptionHelpFormatter,
-        epilog=f"""
-示例用法:
-  # 启动命令行模式（默认）
-  python main.py
-
-  # 启动命令行模式并指定参数
-  python main.py --mode cli --language zh --speaker 沈逸
-
-  # 启动API服务器
-  python main.py --mode api
-
-  # 启动API服务器并指定端口
-  python main.py --mode api --port 9000
-
-  # 启动API服务器并启用热重载（开发模式）
-  python main.py --mode api --port 8000 --reload
-
-支持的说话人:
-  {', '.join(available_speakers)}
-        """
-    )
-
-    # 运行模式选择
-    parser.add_argument(
-        '--mode', '-m',
-        choices=['cli', 'api'],
-        default='cli',
-        help='运行模式: cli=命令行模式, api=API服务器模式 (默认: cli)'
-    )
-
-    # 命令行模式参数
-    cli_group = parser.add_argument_group('命令行模式参数')
-    cli_group.add_argument(
-        '--language', '-l',
-        choices=['zh', 'en'],
-        default='zh',
-        help='用户语言: zh=中文, en=英文 (默认: zh)'
-    )
-    cli_group.add_argument(
-        '--speaker', '-s',
-        choices=available_speakers,
-        default='沈逸' if '沈逸' in available_speakers else (available_speakers[0] if available_speakers else '沈逸'),
-        help='TTS说话人 (默认: 沈逸)'
-    )
-
-    # API服务器模式参数
-    api_group = parser.add_argument_group('API服务器模式参数')
-    api_group.add_argument(
-        '--host',
-        default='0.0.0.0',
-        help='服务器主机地址 (默认: 0.0.0.0)'
-    )
-    api_group.add_argument(
-        '--port', '-p',
-        type=int,
-        default=8000,
-        help='服务器端口 (默认: 8000)'
-    )
-    api_group.add_argument(
-        '--reload',
-        action='store_true',
-        help='启用热重载（开发模式）'
-    )
-
-    return parser
-
-
-def launch_api_server(host: str = "0.0.0.0", port: int = 8000, reload: bool = False):
-    """
-    启动API服务器
-
-    Args:
-        host (str): 服务器主机地址，默认为 "0.0.0.0"
-        port (int): 服务器端口，默认为 8000
-        reload (bool): 是否启用热重载，默认为 False
-    """
-    print(f'{"=" * 80}\n正在启动API服务器...\n{"=" * 80}')
-    print(f"服务器地址: http://{host}:{port}")
-    print(f"API文档: http://{host}:{port}/docs")
-    print(f"热重载: {'启用' if reload else '禁用'}")
-    print(f'{"=" * 80}')
-
-    # 导入并启动FastAPI应用
-    uvicorn.run(
-        "voice_dialogue.api.app:app",
-        host=host,
-        port=port,
-        reload=reload,
-        log_level="info"
-    )
-
-
 def main():
     """
     主程序入口函数
@@ -347,5 +92,4 @@ VoiceDialogue - 语音对话系统
 
 
 if __name__ == '__main__':
-    multiprocessing.freeze_support()
     main()