deepvoice_client/lib/data/services/speech_impl/azure_asr_service.dart

import 'dart:async';
import 'dart:typed_data';
import '../../../data/models/appconfig.dart';
import 'package:flutter/services.dart';
import 'package:get_storage/get_storage.dart';
import '../../../core/utils/logger.dart';
import 'package:get/get.dart';
import '../asr_service.dart';

/// 音频源类型
enum AudioSourceType {
  microphone, // 使用设备麦克风
  external // 使用外部提供的音频数据
}

/// Azure 语音识别服务
///
/// 该服务提供了通过平台通道与原生 Microsoft Speech SDK 交互的接口
class AzureAsrService extends GetxService implements AsrService {
  static final AzureAsrService to = Get.put(AzureAsrService());
  static const MethodChannel _channel = MethodChannel('azure_speech/asr');
  static const EventChannel _eventChannel =
      EventChannel('azure_speech/asr_events');
  // 音频数据事件通道
  static const EventChannel _audioDataEventChannel =
      EventChannel('azure_speech/audio_data_events');

  // final GetStorage _storage = GetStorage();
  bool _isInitialized = false;
  late final String _subscriptionKey;
  late final String _serviceRegion;
  late final String _xunfeiAppId;
  late final String _xunfeiAccessKeyId;
  late final String _xunfeiAccessKeySecret;

  final List<String> _defaultSupportedLanguages = ['zh-CN', 'en-US'];
  @override
  List<String> get supportedLanguages => _defaultSupportedLanguages;

  // 连续识别相关
  bool _isContinuousRecognitionActive = false;
  StreamController<RecognitionEvent>? _eventStreamController;
  StreamSubscription? _eventSubscription;

  // 最新的识别结果
  String _latestRecognizedText = '';
  String get latestRecognizedText => _latestRecognizedText;

  // 最新检测到的语言
  String _latestDetectedLanguage = '';
  String get latestDetectedLanguage => _latestDetectedLanguage;

  String _latestAsrProvider = '';
  String get latestAsrProvider => _latestAsrProvider;

// 音频数据事件订阅
  StreamSubscription? _audioDataEventSubscription;
  // 音频数据流控制器
  StreamController<Uint8List>? _audioDataStreamController;

  // 当前音频源类型
  AudioSourceType _audioSourceType = AudioSourceType.microphone;

  // 防抖时间戳与间隔（实例级）
  DateTime? _lastStartAt;
  DateTime? _lastStopAt;
  static const Duration _debounce = Duration(milliseconds: 600);

  AzureAsrService() {
    _loadConfig();
  }

  /// 从环境变量加载配置
  void _loadConfig() {
    // final _env = _storage.read("ENV") as Map<String, String>;
    _subscriptionKey = AppConfig.env('AZURE_SPEECH_KEY') ?? '';
    _serviceRegion = AppConfig.env('AZURE_SPEECH_REGION') ?? '';
    _xunfeiAppId = AppConfig.env('XUNFEI_ASR_APP_ID') ?? '';
    _xunfeiAccessKeyId = AppConfig.env('XUNFEI_ASR_ACCESS_KEY_ID') ?? '';
    _xunfeiAccessKeySecret =
        AppConfig.env('XUNFEI_ASR_ACCESS_KEY_SECRET') ?? '';

    if (_subscriptionKey.isEmpty || _serviceRegion.isEmpty) {
      throw Exception(
          '未找到 Azure 语音服务配置。请在 .env 文件中设置 AZURE_SPEECH_KEY 和 AZURE_SPEECH_REGION');
    }
  }

  /// 设置事件通道
  void _setupEventChannel() {
    _eventSubscription?.cancel();
    _eventSubscription = _eventChannel.receiveBroadcastStream().listen((event) {
      if (event is Map) {
        _handleRecognitionEvent(event);
      }
    }, onError: _handleRecognitionError);
  }

  /// 设置音频数据事件通道
  void _setupAudioDataEventChannel() {
    _audioDataEventSubscription?.cancel();
    _audioDataEventSubscription =
        _audioDataEventChannel.receiveBroadcastStream().listen((event) {
      if (event is Map) {
        _handleAudioDataEvent(event);
      }
    }, onError: (error) {
      Logger.error('音频数据事件流错误: ${error.toString()}');
    });
  }

  @override
  Future<bool> initialize({
    List<String>? supportedLanguages,
    bool useExternalAudio = false,
    bool useEchoCancellation = false,
  }) async {
    try {
      final List<String> languages =
          supportedLanguages ?? _defaultSupportedLanguages;
//底层会初始化前释放
      // // 检查是否需要重新初始化
      // if (_isInitialized) {
      //   await dispose();
      // }

      // 设置音频源类型
      _audioSourceType = useExternalAudio
          ? AudioSourceType.external
          : AudioSourceType.microphone;

      final bool result = await _channel.invokeMethod('initialize', {
        'subscriptionKey': _subscriptionKey,
        'region': _serviceRegion,
        'supportedLanguages': languages,
        'audioSourceType': _audioSourceType.toString().split('.').last,
        'useEchoCancellation': useEchoCancellation,
        'xunfeiAppId': _xunfeiAppId,
        'xunfeiAccessKeyId': _xunfeiAccessKeyId,
        'xunfeiAccessKeySecret': _xunfeiAccessKeySecret,
      });

      _setupEventChannel();

      _isInitialized = result;
      Logger.info('Azure 语音识别服务初始化${result ? '成功' : '失败'}');
      return result;
    } catch (e) {
      Logger.error('Azure 语音识别服务初始化失败: ${e.toString()}');
      _isInitialized = false;
      rethrow;
    }
  }

  @override
  void onClose() async {
    // 页面或服务被删除时，确保释放底层资源
    await dispose();
    super.onClose();
  }

  @override
  Future<RecognitionEvent> recognizeOnce() async {
    if (!_isInitialized) {
      await initialize();
    }

    try {
      // 调用同步的recognizeOnce方法
      final result = await _channel.invokeMethod('recognizeOnce');

      if (result is Map) {
        // 处理新格式返回
        final String text = result['text'] as String? ?? '';
        final String detectedLanguage =
            result['detectedLanguage'] as String? ?? '';

        // 更新最新识别结果
        _latestRecognizedText = text;
        _latestDetectedLanguage = detectedLanguage;

        return RecognitionEvent.finalResult(
          text: text,
          detectedLanguage: detectedLanguage,
        );
      } else if (result is String) {
        // 兼容旧版本
        _latestRecognizedText = result;
        return RecognitionEvent.finalResult(text: result);
      }

      throw Exception('无效的识别结果格式');
    } catch (e) {
      Logger.error('语音识别失败: ${e.toString()}');
      return RecognitionEvent.error(e.toString());
    }
  }

  @override

  /// 开始连续语音识别（带防抖）
  ///
  /// 行为说明：
  /// - 在 600ms 的冷却时间内再次调用会被防抖拦截并返回 false；
  /// - 首次或冷却期外调用会继续启动连续识别；
  /// - 若当前已处于连续识别中，会先调用停止后再启动。
  Future<bool> startContinuousRecognition(
    bool audioSourceType, {
    bool isRemoveFirstPunctuation = true,
    String mode = "normal",
  }) async {
    // 防抖判断：短时间内重复调用直接拦截
    final now = DateTime.now();
    if (_lastStartAt != null && now.difference(_lastStartAt!) < _debounce) {
      return false;
    }
    _lastStartAt = now;

    if (!_isInitialized) {
      await initialize();
    }

    if (_isContinuousRecognitionActive) {
      await stopContinuousRecognition();
    }

    try {
      // 开始连续识别
      final bool result =
          await _channel.invokeMethod('startContinuousRecognition', {
        'audioSourceType': audioSourceType,
        'isRemoveFirstPunctuation': isRemoveFirstPunctuation,
        'mode': mode,
      });

      if (!result) {
        throw Exception('启动连续识别失败');
      }

      _isContinuousRecognitionActive = true;
      return result;
    } catch (e) {
      Logger.error('开始连续语音识别失败: ${e.toString()}');
      rethrow;
    }
  }

  @override

  /// 停止连续语音识别（带防抖）
  ///
  /// 行为说明：
  /// - 在 600ms 的冷却时间内再次调用会被防抖拦截并返回 false；
  /// - 若当前不在识别中，直接返回 true；
  /// - 调用成功后立即将内部状态置为 false（同时事件通道回调仍会再次校正）。
  Future<bool> stopContinuousRecognition() async {
    // 移除防抖判断，确保停止操作总是被执行
    // 这与 AgentService 的逻辑保持一致，防止“松开还在录”的问题
    _lastStopAt = DateTime.now();

    if (!_isInitialized || !_isContinuousRecognitionActive) {
      return true;
    }

    try {
      final bool result =
          await _channel.invokeMethod('stopContinuousRecognition');
      if (result) {
        _isContinuousRecognitionActive = false;
      }
      return result;
    } catch (e) {
      Logger.error('停止连续语音识别失败: ${e.toString()}');
      rethrow;
    }
  }

  @override
  Future<Stream<RecognitionEvent>> recognizeCallback() async {
    if (!_isInitialized) {
      await initialize();
    }
    try {
      _eventStreamController = StreamController<RecognitionEvent>.broadcast();

      // 开始连续识别
      final bool result = await _channel.invokeMethod('recognizeCallback');
      if (!result) {
        _cleanupEventStream();
      }

      return _eventStreamController!.stream;
    } catch (e) {
      Logger.error('开始连续语音识别失败: ${e.toString()}');
      rethrow;
    }
  }

  @override
  bool isContinuousRecognitionActive() {
    return _isContinuousRecognitionActive;
  }

  /// 处理音频数据事件
  void _handleAudioDataEvent(dynamic event) {
    if (event is! Map) return;

    final Map<dynamic, dynamic> eventMap = event;
    final String eventType = eventMap['type'] as String? ?? '';

    switch (eventType) {
      case 'audioData':
        final Uint8List data = eventMap['data'] as Uint8List? ?? Uint8List(0);
        final double timestamp =
            (eventMap['timestamp'] as num?)?.toDouble() ?? 0.0;
        final int size = eventMap['size'] as int? ?? 0;
        print('接收到音频数据: 大小=${size}字节, 时间戳=${timestamp}');
        Logger.debug('接收到音频数据: 大小=${size}字节, 时间戳=${timestamp}');

        // 发送到音频数据流
        _audioDataStreamController?.add(data);

        break;
    }
  }

  /// 获取音频数据流
  Stream<Uint8List>? getAudioDataStream() {
    print("getAudioDataStream");
    _audioDataStreamController ??= StreamController<Uint8List>.broadcast();
    return _audioDataStreamController?.stream;
  }

  /// 处理来自原生端的识别事件
  void _handleRecognitionEvent(dynamic event) {
    if (event is! Map || _eventStreamController == null) return;

    final Map<dynamic, dynamic> eventMap = event;
    final String eventType = eventMap['type'] as String? ?? '';
    final String provider = eventMap['provider'] as String? ?? '';
    if (provider.isNotEmpty) {
      if (provider != _latestAsrProvider) {
        _latestAsrProvider = provider;
        Logger.info('ASR provider: $_latestAsrProvider');
      }
    } else {
      if (eventType == 'sessionStarted') {
        Logger.warning('ASR provider missing in event');
      }
    }

    // 添加日志帮助调试

    switch (eventType) {
      case 'result':
        final String text = eventMap['text'] as String? ?? '';
        final String detectedLanguage =
            eventMap['detectedLanguage'] as String? ?? '';
        _latestRecognizedText = text;
        _latestDetectedLanguage = detectedLanguage;
        _eventStreamController?.add(RecognitionEvent(
          type: RecognitionEventType.finalResult,
          text: text,
          detectedLanguage: detectedLanguage,
          provider: _latestAsrProvider,
        ));
        break;
      case 'result1':
        final String text = eventMap['text'] as String? ?? '';
        final String detectedLanguage =
            eventMap['detectedLanguage'] as String? ?? '';
        _latestRecognizedText = text;
        _latestDetectedLanguage = detectedLanguage;
        _eventStreamController?.add(RecognitionEvent(
          type: RecognitionEventType.finalResult1,
          text: text,
          detectedLanguage: detectedLanguage,
          provider: _latestAsrProvider,
        ));
        break;

      case 'recognizing':
        final String text = eventMap['text'] as String? ?? '';
        final String detectedLanguage =
            eventMap['detectedLanguage'] as String? ?? '';
        _eventStreamController?.add(RecognitionEvent(
          type: RecognitionEventType.intermediateResult,
          text: text,
          detectedLanguage: detectedLanguage,
          provider: _latestAsrProvider,
        ));
        break;
      case 'sessionStarted':
        _eventStreamController?.add(RecognitionEvent(
          type: RecognitionEventType.sessionStarted,
          provider: _latestAsrProvider,
        ));
        break;

      case 'sessionStopped':
        _isContinuousRecognitionActive = false;
        _eventStreamController?.add(RecognitionEvent(
          type: RecognitionEventType.sessionStopped,
          provider: _latestAsrProvider,
        ));
        break;

      case 'canceled':
        _isContinuousRecognitionActive = false;
        final String reason = eventMap['reason'] as String? ?? '';
        final String errorDetails = eventMap['errorDetails'] as String? ?? '';

        if (reason.isNotEmpty || errorDetails.isNotEmpty) {
          Logger.error('识别取消: $reason - ${errorDetails.toString()}');
        }

        _eventStreamController?.add(RecognitionEvent(
          type: RecognitionEventType.canceled,
          error: '$reason: $errorDetails',
          provider: _latestAsrProvider,
        ));
        break;

      case 'error':
        final String error = eventMap['message'] as String? ?? '';
        Logger.error('识别错误: ${error.toString()}');
        _eventStreamController?.add(RecognitionEvent(
          type: RecognitionEventType.error,
          error: error,
          provider: _latestAsrProvider,
        ));
        break;
    }
  }

  /// 处理识别事件流错误
  void _handleRecognitionError(Object error) {
    Logger.error('识别事件流错误: ${error.toString()}');
    _eventStreamController?.addError(error);
    _cleanupEventStream();
  }

  /// 清理事件流资源
  void _cleanupEventStream() {
    _eventStreamController?.close();
    _eventStreamController = null;
    _isContinuousRecognitionActive = false;
  }

  @override
  Future<void> dispose() async {
    try {
      if (_isContinuousRecognitionActive) {
        await stopContinuousRecognition();
      }

      // 取消事件订阅
      await _eventSubscription?.cancel();
      _eventSubscription = null;

      // 取消音频数据事件订阅
      await _audioDataEventSubscription?.cancel();
      _audioDataEventSubscription = null;

      // 关闭音频数据流
      await _audioDataStreamController?.close();
      _audioDataStreamController = null;

      // 通知原生端释放资源
      await _channel.invokeMethod('dispose');
      _isInitialized = false;

      Logger.info('Azure 语音识别资源已释放');
    } catch (e) {
      Logger.error('释放语音识别资源失败: ${e.toString()}');
      _cleanupEventStream();
      _isInitialized = false;
    }
  }

  @override
  Future<bool> pushAudioData(Uint8List data) async {
    if (!_isInitialized) {
      await initialize(useExternalAudio: true);
    }

    if (_audioSourceType != AudioSourceType.external) {
      Logger.error('当前非外部音频模式，不能推送音频数据');
      return false;
    }

    try {
      final bool result = await _channel.invokeMethod('pushAudioData', {
        'audioData': data // Flutter 会自动将 Uint8List 转换为 ByteBuffer
      });

      if (!result) {
        throw Exception('推送音频数据失败');
      }

      return true;
    } catch (e) {
      Logger.error('推送音频数据失败: ${e.toString()}');
      return false;
    }
  }

  @override
  Future<bool> enableRecord(bool audioSourceType, String filePath,
      {bool acceptAudioData = false}) async {
    try {
      final bool result = await _channel.invokeMethod('enableRecord', {
        'audioSourceType': audioSourceType,
        'filePath': filePath,
        'acceptAudioData': acceptAudioData, // 新增参数
      });
      if (acceptAudioData) {
        _setupAudioDataEventChannel();
      }

      return result;
    } catch (e) {
      Logger.error('开始录音: ${e.toString()}');
      rethrow;
    }
  }

  @override
  Future<bool> stopRecord(bool isSave) async {
    try {
      final bool result = await _channel.invokeMethod('stopRecord', {
        'isSave': isSave,
      });

      return result;
    } catch (e) {
      Logger.error('停止录音: ${e.toString()}');
      rethrow;
    }
  }

  @override
  Future<bool> pauseRecord() async {
    try {
      final bool result = await _channel.invokeMethod('pauseRecord');

      return result;
    } catch (e) {
      Logger.error('暂停录音: ${e.toString()}');
      rethrow;
    }
  }

  /// 设置音频配置
  ///
  /// [sampleRate] 采样率，默认16000
  /// [channels] 声道数，默认1（单声道）
  Future<bool> setAudioConfig({
    int sampleRate = 16000,
    int channels = 1,
  }) async {
    try {
      final bool result = await _channel.invokeMethod('setAudioConfig', {
        'sampleRate': sampleRate,
        'channels': channels,
      });

      Logger.info(
          '音频配置设置${result ? '成功' : '失败'}: 采样率=$sampleRate, 声道数=$channels');
      return result;
    } catch (e) {
      Logger.error('设置音频配置失败: ${e.toString()}');
      rethrow;
    }
  }

  @override
  Future<bool> moveFile(String sourcePath, String destPath) async {
    // TODO: implement moveFile
    try {
      final bool result = await _channel.invokeMethod('moveFile', {
        'sourcePath': sourcePath,
        'destPath': destPath,
      });

      return result;
    } catch (e) {
      Logger.error('开始录音: ${e.toString()}');
      rethrow;
    }
  }

  @override
  Future<bool> renameFile(String filePath, String newName) async {
    // TODO: implement renameFile
    try {
      final bool result = await _channel.invokeMethod('renameFile', {
        'filePath': filePath,
        'newName': newName,
      });

      return result;
    } catch (e) {
      Logger.error('开始录音: ${e.toString()}');
      rethrow;
    }
  }

  @override
  Future<void> disableBluetoothAudio() async {
    // TODO: implement disableBluetoothAudio
    try {
      final bool result = await _channel.invokeMethod('disableBluetoothAudio');
      return;
    } catch (e) {
      Logger.error('开始录音: ${e.toString()}');
      rethrow;
    }
  }

  @override
  Future<void> restoreOriginalAudioState() async {
    // TODO: implement restoreOriginalAudioState
    try {
      final bool result =
          await _channel.invokeMethod('restoreOriginalAudioState');
      return;
    } catch (e) {
      Logger.error('开始录音: ${e.toString()}');
      rethrow;
    }
  }

  @override
  Future<bool> resumeRecord() async {
    try {
      final bool result = await _channel.invokeMethod('resumeRecord');

      return result;
    } catch (e) {
      Logger.error('继续录音: ${e.toString()}');
      rethrow;
    }
  }
}