deepvoice_client/lib/data/services/volcano_tts_service.dart

import 'dart:async';
import 'dart:convert';
import 'dart:typed_data';
import 'dart:io';
import 'dart:math' as math;
import 'package:get/get.dart';
import 'package:just_audio/just_audio.dart';
import 'package:web_socket_channel/web_socket_channel.dart';
import 'package:audio_session/audio_session.dart';
import 'package:flutter_dotenv/flutter_dotenv.dart';
import '../../core/utils/logger.dart';

/// 火山语音服务异常
class VolcanoTtsException implements Exception {
  final String message;
  VolcanoTtsException(this.message);
  
  @override
  String toString() => message;
}

/// 自定义音频源，用于从字节数组读取音频数据
class BytesAudioSource extends StreamAudioSource {
  final Uint8List _bytes;
  static const int _bufferSize = 4096;  // 4KB buffer size for better Android compatibility

  BytesAudioSource(this._bytes);

  @override
  Future<StreamAudioResponse> request([int? start, int? end]) async {
    start = start ?? 0;
    end = end ?? _bytes.length;

    try {
      final subData = _bytes.sublist(start, end);
      
      // 使用固定大小的块进行流式传输
      final chunks = <List<int>>[];
      var offset = 0;
      while (offset < subData.length) {
        final chunkSize = math.min(_bufferSize, subData.length - offset);
        chunks.add(subData.sublist(offset, offset + chunkSize));
        offset += chunkSize;
      }

      return StreamAudioResponse(
        sourceLength: _bytes.length,
        contentLength: subData.length,
        offset: start,
        stream: Stream.fromIterable(chunks),
        contentType: 'audio/mpeg',
      );
    } catch (e) {
      print('BytesAudioSource request error: $e');
      rethrow;
    }
  }

  @override
  Future<int> get length => Future.value(_bytes.length);
}

/// 火山语音合成服务
class VolcanoTtsService extends GetxService {
  static const String _host = 'openspeech.bytedance.com';
  static const String _apiUrl = 'wss://$_host/api/v1/tts/ws_binary';
  
  // 配置参数
  final String _appId;
  final String _token;
  final String _cluster;
  final String _voiceType;

  // 音频播放器
  late AudioPlayer _audioPlayer;
  final ConcatenatingAudioSource _playlist = ConcatenatingAudioSource(children: []);
  final isEnabled = true.obs;  // 默认启用
  final isPlaying = false.obs;
  StreamSubscription? _playbackEventSubscription;
  StreamSubscription? _playerStateSubscription;
  AudioSession? _audioSession;

  // WebSocket 相关
  WebSocketChannel? _channel;
  bool _isDisposed = false;

  // 句子管理
  String _pendingText = '';
  final List<String> _sentenceQueue = [];
  static final _sentenceBreaks = RegExp(r'[。！？.!?，,、]');  // 添加更多标点符号
  bool _isFetching = false;
  static const int _maxSegmentLength = 100;  // 最大分段长度
  static const int _optimalSegmentLength = 50;  // 最佳分段长度

  // WebSocket连接配置
  static final Map<String, String> _wsHeaders = {
    'User-Agent': 'DeepVoice/1.0',
    'Accept': '*/*',
    'Accept-Encoding': 'gzip, deflate, br',
    'Connection': 'Upgrade',
    'Upgrade': 'websocket',
    'Sec-WebSocket-Version': '13',
    'Sec-WebSocket-Extensions': 'permessage-deflate',
    'Origin': 'https://openspeech.bytedance.com',
  };

  // 二进制协议头部字段定义
  static const int PROTOCOL_VERSION = 0x1;  // 0b0001 - 版本1
  static const int HEADER_SIZE = 0x1;       // 0b0001 - 4字节
  static const int MESSAGE_TYPE = 0x1;      // 0b0001 - full client request
  static const int MESSAGE_FLAGS = 0x0;     // 0b0000 - 无特殊标记
  static const int SERIALIZATION = 0x1;     // 0b0001 - JSON
  static const int COMPRESSION = 0x0;       // 0b0000 - 无压缩
  static const int RESERVED = 0x0;          // 0b0000 - 保留字段

  // 构建二进制协议头
  static Uint8List _buildHeader() {
    final header = ByteData(4);  // 4字节的头部
    
    // 第一个字节: [协议版本(4位) | 报头大小(4位)]
    header.setUint8(0, (PROTOCOL_VERSION << 4) | HEADER_SIZE);
    
    // 第二个字节: [消息类型(4位) | 消息标记(4位)]
    header.setUint8(1, (MESSAGE_TYPE << 4) | MESSAGE_FLAGS);
    
    // 第三个字节: [序列化方法(4位) | 压缩方法(4位)]
    header.setUint8(2, (SERIALIZATION << 4) | COMPRESSION);
    
    // 第四个字节: [保留字段(8位)]
    header.setUint8(3, RESERVED);
    
    return header.buffer.asUint8List();
  }

  VolcanoTtsService() : 
    _appId = dotenv.env['VOLCANO_TTS_APP_ID'] ?? '',
    _token = dotenv.env['VOLCANO_TTS_APP_TOKEN'] ?? '',
    _cluster = dotenv.env['VOLCANO_TTS_CLUSTER'] ?? '',
    _voiceType = dotenv.env['VOLCANO_TTS_VOICE_TYPE'] ?? 'zh_male_M392_conversation_wvae_bigtts' {
    if (_appId.isEmpty || _token.isEmpty || _cluster.isEmpty) {
      throw Exception('火山语音配置信息不完整，请检查环境变量');
    }
  }

  @override
  Future<void> onInit() async {
    super.onInit();
    try {
      await _initAudioSession();
      await _initAudioPlayer();
      await useDefaultAudioDevice();
    } catch (e, stackTrace) {
      print('初始化TTS服务失败: $e');
      print('Stack trace: $stackTrace');
    }
  }

  Future<void> _initAudioSession() async {
    _audioSession = await AudioSession.instance;
    await _audioSession?.configure(AudioSessionConfiguration(
      avAudioSessionCategory: AVAudioSessionCategory.playback,
      androidAudioAttributes: const AndroidAudioAttributes(
        contentType: AndroidAudioContentType.speech,
        usage: AndroidAudioUsage.media,
        flags: AndroidAudioFlags.audibilityEnforced,
      ),
      androidAudioFocusGainType: AndroidAudioFocusGainType.gain,
      androidWillPauseWhenDucked: true,
    ));
    
    // 确保音频会话激活并设置正确的音频模式
    if (Platform.isAndroid) {
      await _audioSession?.setActive(true, avAudioSessionSetActiveOptions: AVAudioSessionSetActiveOptions.notifyOthersOnDeactivation);
    } else {
      await _audioSession?.setActive(true);
    }
  }

  Future<void> _initAudioPlayer() async {
    if (_isDisposed) return;
    
    _audioPlayer = AudioPlayer();
    
    // 配置Android特定的播放参数
    if (Platform.isAndroid) {
      await _audioPlayer.setAndroidAudioAttributes(const AndroidAudioAttributes(
        contentType: AndroidAudioContentType.speech,
        usage: AndroidAudioUsage.media,
        flags: AndroidAudioFlags.audibilityEnforced,
      ));
    }
    
    // 基本播放器配置
    await _audioPlayer.setVolume(1.0);
    await _audioPlayer.setLoopMode(LoopMode.off);
    await _audioPlayer.setSpeed(1.0);
    
    // 设置音频缓冲配置，启用预加载以减少延迟
    await _audioPlayer.setAudioSource(
      _playlist,
      initialPosition: Duration.zero,
      preload: true,
    );
    
    _setupEventListeners();
  }

  Future<void> speak(String text) async {
    if (!isEnabled.value || text.trim().isEmpty) return;

    try {
      _pendingText += text;
      _extractFullSentences();
      await _startPreloadIfNeeded();
    } catch (e, stack) {
      print('语音合成失败: $e');
      await _reinitializePlayer();
    }
  }

  void _extractFullSentences() {
    if (_pendingText.isEmpty) return;

    final matches = _sentenceBreaks.allMatches(_pendingText).toList();
    if (matches.isEmpty) {
      // 如果没有找到句子分隔符，按长度分割
      _splitByLength(_pendingText);
      _pendingText = '';
      return;
    }

    int lastPos = 0;
    String currentBatch = '';

    for (final match in matches) {
      final end = match.end;
      final sentence = _pendingText.substring(lastPos, end).trim();
      
      if (sentence.isEmpty) {
        lastPos = end;
        continue;
      }

      // 如果当前批次为空，直接添加句子
      if (currentBatch.isEmpty) {
        currentBatch = sentence;
      } 
      // 如果当前批次加上新句子不超过30字符，则合并
      else if ((currentBatch + sentence).length <= 30) {
        currentBatch += sentence;
      } 
      // 如果超过30字符，将当前批次加入队列，开始新批次
      else {
        if (currentBatch.isNotEmpty) {
          _sentenceQueue.add(currentBatch);
        }
        currentBatch = sentence;
      }
      
      lastPos = end;
    }

    // 处理最后一个批次
    if (currentBatch.isNotEmpty) {
      _sentenceQueue.add(currentBatch);
    }
    
    // 处理剩余的文本
    if (lastPos < _pendingText.length) {
      final remaining = _pendingText.substring(lastPos).trim();
      if (remaining.isNotEmpty) {
        if (remaining.length <= 30) {
          // 尝试与最后一个批次合并
          if (_sentenceQueue.isNotEmpty && 
              (_sentenceQueue.last + remaining).length <= 30) {
            _sentenceQueue[_sentenceQueue.length - 1] += remaining;
          } else {
            _sentenceQueue.add(remaining);
          }
        } else {
          _splitByLength(remaining);
        }
      }
    }

    _pendingText = '';
  }

  void _splitByLength(String text) {
    if (text.isEmpty) return;
    
    int start = 0;
    while (start < text.length) {
      int end = math.min(start + 30, text.length);
      
      // 尝试在标点或空格处分割
      if (end < text.length) {
        final spacePos = text.lastIndexOf(' ', end);
        final breakPos = _sentenceBreaks.firstMatch(
          text.substring(start, end)
        )?.end;
        
        if (breakPos != null && breakPos > start) {
          end = start + breakPos;
        } else if (spacePos > start) {
          end = spacePos;
        }
      }
      
      final segment = text.substring(start, end).trim();
      if (segment.isNotEmpty) {
        // 尝试与前一个批次合并
        if (_sentenceQueue.isNotEmpty && 
            (_sentenceQueue.last + segment).length <= 30) {
          _sentenceQueue[_sentenceQueue.length - 1] += segment;
        } else {
          _sentenceQueue.add(segment);
        }
      }
      start = end;
    }
  }

  Future<void> _startPreloadIfNeeded() async {
    if (_isFetching || _sentenceQueue.isEmpty) return;
    _isFetching = true;
    await _fetchNextSegment();
  }

  Future<void> _fetchNextSegment() async {
    if (_sentenceQueue.isEmpty) {
      _isFetching = false;
      return;
    }

    try {
      // 预加载多个片段以实现流畅播放
      final segments = <String>[];
      while (_sentenceQueue.isNotEmpty && segments.length < 2) {  // 减少预加载数量以降低内存压力
        segments.add(_sentenceQueue.removeAt(0));
      }

      for (final segment in segments) {
        print('获取音频片段: $segment');
        final audioData = await _synthesize(segment);
        
        if (audioData != null) {
          print('成功获取音频数据: ${audioData.length} bytes');
          try {
            final audioSource = BytesAudioSource(audioData);
            await _playlist.add(audioSource);
            print('音频片段已添加到播放列表');

            // 如果是第一个片段且播放器未在播放，开始播放
            if (!_audioPlayer.playing && _playlist.length == 1) {
              // 确保音频会话处于激活状态
              await _audioSession?.setActive(true);
              
              // 设置播放参数
              await _audioPlayer.setVolume(1.0);
              await _audioPlayer.seek(Duration.zero, index: 0);
              
              // 开始播放
              await _audioPlayer.play();
            }
          } catch (e) {
            print('添加音频源失败: $e');
          }
        }
      }

      // 如果队列还有内容，继续获取下一批
      if (_sentenceQueue.isNotEmpty) {
        await Future.delayed(const Duration(milliseconds: 100));  // 添加延迟以避免过快请求
        await _fetchNextSegment();
      } else {
        _isFetching = false;
        // 检查是否还有未处理的文本
        _extractFullSentences();
        if (_sentenceQueue.isNotEmpty) {
          await _startPreloadIfNeeded();
        }
      }
    } catch (e, stackTrace) {
      print('获取音频失败: $e');
      print('Stack trace: $stackTrace');
      _isFetching = false;
    }
  }

  Future<Uint8List?> _synthesize(String text) async {
    WebSocket? socket;
    StreamSubscription? subscription;
    
    try {
      final requestJson = {
        'app': {
          'appid': _appId,
          'token': _token,
          'cluster': _cluster,
        },
        'user': {
          'uid': '${DateTime.now().millisecondsSinceEpoch}',
        },
        'audio': {
          'voice_type': _voiceType,
          'encoding': 'mp3',
          'speed_ratio': 1.0,
          'volume_ratio': 1.0,
          'pitch_ratio': 1.0,
          'language': 'zh',
        },
        'request': {
          'reqid': '${DateTime.now().millisecondsSinceEpoch}',
          'text': text,
          'text_type': 'plain',
          'operation': 'query',
        },
      };
      
      final jsonBytes = utf8.encode(json.encode(requestJson));
      final header = _buildHeader();
      final fullRequest = Uint8List(header.length + 4 + jsonBytes.length);
      
      fullRequest.setAll(0, header);
      final lengthBytes = ByteData(4)..setInt32(0, jsonBytes.length, Endian.big);
      fullRequest.setAll(header.length, lengthBytes.buffer.asUint8List());
      fullRequest.setAll(header.length + 4, jsonBytes);
      
      final wsUrl = Uri.parse(_apiUrl);
      final completer = Completer<Uint8List?>();
      
      socket = await WebSocket.connect(
        wsUrl.toString(),
        headers: {
          ..._wsHeaders,
          'Authorization': 'Bearer;$_token',
        },
      ).timeout(
        const Duration(seconds: 5),
        onTimeout: () {
          throw TimeoutException('WebSocket连接超时');
        },
      );
      
      subscription = socket.listen(
        (response) {
          if (!completer.isCompleted && response is Uint8List) {
            if (response.length < 2) return;
            
            final messageType = response[1] >> 4;
            
            if (messageType == 0x0B) {
              final headerSize = response[0] & 0x0F;
              final payloadStart = headerSize * 4;
              
              if (response.length > payloadStart) {
                final audioData = response.sublist(payloadStart);
                completer.complete(audioData);
              }
            } else if (messageType == 0x0F) {
              try {
                final headerSize = response[0] & 0x0F;
                final payloadStart = headerSize * 4;
                
                if (response.length < payloadStart + 4) {
                  throw VolcanoTtsException('错误响应数据不完整');
                }

                final jsonData = response.sublist(payloadStart + 4);
                final jsonString = utf8.decode(jsonData, allowMalformed: true);
                
                if (jsonString.trim().startsWith('{')) {
                  final errorMap = json.decode(jsonString);
                  throw VolcanoTtsException(errorMap['error'] as String? ?? '未知错误');
                }
              } catch (e) {
                if (e is VolcanoTtsException) rethrow;
                throw VolcanoTtsException('服务返回错误: $e');
              }
            }
          }
        },
        onError: (error) {
          if (!completer.isCompleted) {
            completer.completeError(error);
          }
        },
        onDone: () {
          if (!completer.isCompleted) {
            completer.complete(null);
          }
        },
        cancelOnError: false,
      );

      socket.add(fullRequest);
      
      return await completer.future.timeout(
        const Duration(seconds: 10),
        onTimeout: () => null,
      );
    } catch (e) {
      print('合成请求失败: $e');
      return null;
    } finally {
      await subscription?.cancel();
      await socket?.close();
    }
  }

  Future<void> _playAudio(Uint8List audioData) async {
    if (_isDisposed) {
      print('TTS服务已销毁，跳过播放');
      return;
    }
    
    try {
      print('准备播放音频数据: ${audioData.length} bytes');
      final audioSource = BytesAudioSource(audioData);
      
      print('清除播放列表');
      await _playlist.clear();
      
      print('添加新的音频源到播放列表');
      await _playlist.add(audioSource);
      
      if (!_audioPlayer.playing) {
        print('设置音量和播放位置');
        await _audioPlayer.setVolume(1.0);
        await _audioPlayer.seek(Duration.zero, index: 0);
        
        // 确保音频会话处于激活状态
        await _audioSession?.setActive(true);
        
        print('等待50ms确保音频准备就绪');
        await Future.delayed(const Duration(milliseconds: 50));
        
        print('开始播放');
        await _audioPlayer.play();
        print('播放命令已发送');
      } else {
        print('播放器已在播放中');
      }
    } catch (e, stack) {
      print('播放音频失败: $e');
      print('错误堆栈: $stack');
      await _reinitializePlayer();
    }
  }

  Future<void> stop() async {
    try {
      print('停止播放');
      await _audioPlayer.stop();
      await _playlist.clear();
      _pendingText = '';
      _sentenceQueue.clear();
      _isFetching = false;
      isPlaying.value = false;
      
      // 确保清空所有待处理的音频数据
      try {
        // 取消所有正在进行的WebSocket连接
        await _channel?.sink.close();
        _channel = null;
        
        // 重置播放器状态
        await _audioPlayer.pause();
        await _audioPlayer.seek(Duration.zero);
        
        print('已清空所有待播放的文字和音频');
      } catch (e) {
        print('清空音频缓存时出错: $e');
      }
    } catch (e) {
      print('停止播放失败: $e');
    }
  }

  // 添加一个方法来检查音频播放器是否真正在播放
  bool isActuallyPlaying() {
    try {
      // 检查播放器状态
      final isPlayerPlaying = _audioPlayer.playing;
      
      // 检查播放列表是否为空
      final hasAudioSource = _playlist.length > 0;
      
      // 只有当播放器正在播放且播放列表不为空时，才认为真正在播放
      final actuallyPlaying = isPlayerPlaying && hasAudioSource;
      
      // 如果状态不一致，记录日志
      if (isPlaying.value != actuallyPlaying) {
        print('音频播放状态不一致: isPlaying.value=${isPlaying.value}, 实际播放状态=$actuallyPlaying (播放器状态=$isPlayerPlaying, 播放列表长度=${_playlist.length})');
      }
      
      return actuallyPlaying;
    } catch (e) {
      print('检查实际播放状态失败: $e');
      // 如果检查失败，返回当前的 isPlaying 值
      return isPlaying.value;
    }
  }

  Future<void> _reinitializePlayer() async {
    try {
      await stop();
      await _audioPlayer.dispose();
      
      // 重新初始化音频会话
      await _initAudioSession();
      
      _audioPlayer = AudioPlayer();
      await _audioPlayer.setVolume(1.0);
      await _audioPlayer.setLoopMode(LoopMode.off);
      
      // 重新配置Android特定的播放参数
      await _audioPlayer.setAndroidAudioAttributes(const AndroidAudioAttributes(
        contentType: AndroidAudioContentType.speech,
        usage: AndroidAudioUsage.media,
        flags: AndroidAudioFlags.audibilityEnforced,
      ));
      
      await _audioPlayer.setAudioSource(
        _playlist,
        initialPosition: Duration.zero,
        preload: true,
      );
      
      _setupEventListeners();
      print('播放器重新初始化成功');
    } catch (e) {
      print('播放器重新初始化失败: $e');
    }
  }

  void _setupEventListeners() {
    _playbackEventSubscription?.cancel();
    _playerStateSubscription?.cancel();
    
    _playbackEventSubscription = _audioPlayer.playbackEventStream.listen(
      (event) {
        if (_isDisposed) return;
        if (event.processingState == ProcessingState.completed) {
          _tryPlayNext();
        }
      },
      onError: (error) {
        if (_isDisposed) return;
        _reinitializePlayer();
      },
    );

    _playerStateSubscription = _audioPlayer.playerStateStream.listen(
      (state) {
        if (_isDisposed) return;
        isPlaying.value = state.playing;
        
        if (state.processingState == ProcessingState.completed) {
          _tryPlayNext();
        }
      },
      onError: (error) {
        if (_isDisposed) return;
        _reinitializePlayer();
      },
    );
  }

  Future<void> _tryPlayNext() async {
    if (_isDisposed || _playlist.length <= 0) return;
    
    try {
      if (!_audioPlayer.playing) {
        // 确保音频会话处于激活状态
        await _audioSession?.setActive(true);
        
        // 减少延迟时间
        await Future.delayed(const Duration(milliseconds: 30));
        
        if (!_isDisposed && !_audioPlayer.playing) {
          await _audioPlayer.seek(Duration.zero, index: 0);
          await _audioPlayer.play();
        }
      }
    } catch (e) {
      if (e is! StateError) {
        print('播放错误: $e');
        await _reinitializePlayer();
      }
    }
  }

  void toggleEnabled() {
    isEnabled.toggle();
    if (!isEnabled.value) {
      stop();
    }
  }

  Future<void> _cleanupResources() async {
    _isDisposed = true;
    await _playbackEventSubscription?.cancel();
    await _playerStateSubscription?.cancel();
    await _audioSession?.setActive(false);
    await _channel?.sink.close();
    await stop();
    await _audioPlayer.dispose();
  }

  @override
  void onClose() async {
    print('关闭TTS服务');
    await _cleanupResources();
    super.onClose();
  }

  /// 使用系统默认音频设备
  Future<bool> useDefaultAudioDevice() async {
    try {
      if (Platform.isAndroid) {
        // Android 平台 - 重置为正常模式，让系统自动选择
        final audioManager = AndroidAudioManager();
        await audioManager.setMode(AndroidAudioHardwareMode.normal);
        await audioManager.stopBluetoothSco();
        await audioManager.setBluetoothScoOn(false);
        // 不显式设置扬声器状态，让系统自动选择
        await audioManager.setSpeakerphoneOn(false);
        return true;
      } else if (Platform.isIOS) {
        // iOS 平台 - 清除任何输出端口覆盖
        final avAudioSession = AVAudioSession();
        await avAudioSession.overrideOutputAudioPort(AVAudioSessionPortOverride.none);
        return true;
      }
      return false;
    } catch (e) {
      Logger.error('设置默认音频设备失败: $e');
      return false;
    }
  }
}