|
|
@ -46,31 +46,95 @@ class BackgroundAgentService extends GetxService { |
|
|
// 添加一个标志,表示系统是否正在播放 TTS |
|
|
// 添加一个标志,表示系统是否正在播放 TTS |
|
|
bool _isSpeaking = false; |
|
|
bool _isSpeaking = false; |
|
|
|
|
|
|
|
|
// 添加一个订阅,用于监听 TTS 状态变化 |
|
|
|
|
|
StreamSubscription? _ttsSpeakingSubscription; |
|
|
|
|
|
|
|
|
|
|
|
// 添加一个变量来跟踪当前的AI响应流订阅 |
|
|
// 添加一个变量来跟踪当前的AI响应流订阅 |
|
|
StreamSubscription? _aiResponseSubscription; |
|
|
StreamSubscription? _aiResponseSubscription; |
|
|
|
|
|
|
|
|
// 添加一个标志,表示是否应该取消当前的AI响应 |
|
|
// 添加一个标志,表示是否应该取消当前的AI响应 |
|
|
bool _shouldCancelAiResponse = false; |
|
|
bool _shouldCancelAiResponse = false; |
|
|
|
|
|
|
|
|
|
|
|
// 添加一个变量来跟踪最后一次 TTS 状态变化的时间 |
|
|
|
|
|
DateTime? _lastTtsStateChangeTime; |
|
|
|
|
|
|
|
|
BackgroundAgentService() |
|
|
BackgroundAgentService() |
|
|
: _aiService = VolcanoAIService(), |
|
|
: _aiService = VolcanoAIService(), |
|
|
_ttsService = Get.find<VolcanoTtsService>(), |
|
|
_ttsService = Get.find<VolcanoTtsService>(), |
|
|
_voiceRecognitionService = Get.find<VoiceRecognitionService>() { |
|
|
_voiceRecognitionService = Get.find<VoiceRecognitionService>() { |
|
|
// 监听 TTS 播放状态 |
|
|
// 不再监听 TTS 播放状态,而是使用定期检查 |
|
|
_ttsSpeakingSubscription = _ttsService.isPlaying.listen((speaking) { |
|
|
// 初始化 TTS 状态 |
|
|
_isSpeaking = speaking; |
|
|
_isSpeaking = _ttsService.isActuallyPlaying(); |
|
|
print('TTS 播放状态变化: $_isSpeaking'); |
|
|
_lastTtsStateChangeTime = DateTime.now(); |
|
|
|
|
|
|
|
|
|
|
|
// 添加定期检查实际播放状态的计时器 |
|
|
|
|
|
Timer.periodic(const Duration(seconds: 1), (timer) { |
|
|
|
|
|
// 检查实际播放状态 |
|
|
|
|
|
final actuallyPlaying = _ttsService.isActuallyPlaying(); |
|
|
|
|
|
|
|
|
// 如果 TTS 停止播放,且正在进行语音识别,重新启动无语音超时计时器 |
|
|
// 如果状态发生变化,更新状态并记录时间 |
|
|
if (!speaking && _isListening && _hasRecognizedSpeech && _noSpeechTimer == null) { |
|
|
if (_isSpeaking != actuallyPlaying) { |
|
|
_startNoSpeechTimer(); |
|
|
print('TTS 播放状态变化: $_isSpeaking -> $actuallyPlaying'); |
|
|
|
|
|
_isSpeaking = actuallyPlaying; |
|
|
|
|
|
_lastTtsStateChangeTime = DateTime.now(); |
|
|
|
|
|
|
|
|
|
|
|
// 如果 TTS 服务的 isPlaying 与实际状态不一致,尝试修正 |
|
|
|
|
|
if (actuallyPlaying != _ttsService.isPlaying.value) { |
|
|
|
|
|
try { |
|
|
|
|
|
_ttsService.isPlaying.value = actuallyPlaying; |
|
|
|
|
|
print('已修正 TTS 服务的播放状态'); |
|
|
|
|
|
} catch (e) { |
|
|
|
|
|
print('修正 TTS 播放状态失败: $e'); |
|
|
|
|
|
} |
|
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
// 如果 TTS 停止播放,且正在进行语音识别,重新启动无语音超时计时器 |
|
|
|
|
|
if (!actuallyPlaying && _isListening && _hasRecognizedSpeech && _noSpeechTimer == null) { |
|
|
|
|
|
_startNoSpeechTimer(); |
|
|
|
|
|
} |
|
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
// 如果正在处理交互,进行更全面的检查 |
|
|
|
|
|
if (_isProcessing) { |
|
|
|
|
|
_checkAndUpdateTtsState(); |
|
|
} |
|
|
} |
|
|
}); |
|
|
}); |
|
|
} |
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
// 添加一个方法来检查和更新 TTS 状态 |
|
|
|
|
|
void _checkAndUpdateTtsState() { |
|
|
|
|
|
try { |
|
|
|
|
|
// 获取实际播放状态 |
|
|
|
|
|
final actuallyPlaying = _ttsService.isActuallyPlaying(); |
|
|
|
|
|
|
|
|
|
|
|
// 如果状态显示正在播放,但已经很长时间没有状态变化,强制重置 |
|
|
|
|
|
if (actuallyPlaying && _lastTtsStateChangeTime != null) { |
|
|
|
|
|
final now = DateTime.now(); |
|
|
|
|
|
final difference = now.difference(_lastTtsStateChangeTime!); |
|
|
|
|
|
|
|
|
|
|
|
// 如果超过15秒没有状态变化,强制重置 |
|
|
|
|
|
if (difference.inSeconds > 15) { |
|
|
|
|
|
print('强制重置 TTS 状态:已经 ${difference.inSeconds} 秒没有状态变化'); |
|
|
|
|
|
|
|
|
|
|
|
try { |
|
|
|
|
|
// 停止 TTS 播放 |
|
|
|
|
|
_ttsService.stop(); |
|
|
|
|
|
|
|
|
|
|
|
// 强制更新状态 |
|
|
|
|
|
_isSpeaking = false; |
|
|
|
|
|
_ttsService.isPlaying.value = false; |
|
|
|
|
|
|
|
|
|
|
|
// 更新最后状态变化时间 |
|
|
|
|
|
_lastTtsStateChangeTime = DateTime.now(); |
|
|
|
|
|
|
|
|
|
|
|
print('TTS 状态已强制重置'); |
|
|
|
|
|
} catch (e) { |
|
|
|
|
|
print('强制重置 TTS 状态失败: $e'); |
|
|
|
|
|
} |
|
|
|
|
|
} |
|
|
|
|
|
} |
|
|
|
|
|
} catch (e) { |
|
|
|
|
|
print('检查 TTS 播放状态失败: $e'); |
|
|
|
|
|
} |
|
|
|
|
|
} |
|
|
|
|
|
|
|
|
// 处理蓝牙耳机按钮触发的交互 |
|
|
// 处理蓝牙耳机按钮触发的交互 |
|
|
Future<void> handleAgentInteraction(String systemPrompt) async { |
|
|
Future<void> handleAgentInteraction(String systemPrompt) async { |
|
|
if (_isProcessing) { |
|
|
if (_isProcessing) { |
|
|
@ -110,9 +174,26 @@ class BackgroundAgentService extends GetxService { |
|
|
|
|
|
|
|
|
// 启动无语音超时计时器 |
|
|
// 启动无语音超时计时器 |
|
|
void _startNoSpeechTimer() { |
|
|
void _startNoSpeechTimer() { |
|
|
|
|
|
// 检查 TTS 播放状态,使用 isActuallyPlaying 获取实际状态 |
|
|
|
|
|
_isSpeaking = _ttsService.isActuallyPlaying(); |
|
|
|
|
|
|
|
|
// 如果系统正在播放 TTS,不启动计时器 |
|
|
// 如果系统正在播放 TTS,不启动计时器 |
|
|
if (_isSpeaking) { |
|
|
if (_isSpeaking) { |
|
|
print('系统正在播放 TTS,不启动无语音超时计时器'); |
|
|
print('系统正在播放 TTS,不启动无语音超时计时器 (isPlaying.value: ${_ttsService.isPlaying.value}, 实际播放状态: $_isSpeaking)'); |
|
|
|
|
|
|
|
|
|
|
|
// 即使TTS正在播放,也设置一个更长的安全超时,防止系统永远卡在循环中 |
|
|
|
|
|
_noSpeechTimer?.cancel(); |
|
|
|
|
|
_noSpeechTimer = Timer(const Duration(seconds: 30), () { |
|
|
|
|
|
// 再次检查 TTS 状态,使用 isActuallyPlaying |
|
|
|
|
|
_isSpeaking = _ttsService.isActuallyPlaying(); |
|
|
|
|
|
|
|
|
|
|
|
print('安全超时触发:即使TTS正在播放,30秒内没有新的语音输入,强制退出循环交互 (实际播放状态: $_isSpeaking)'); |
|
|
|
|
|
if (_recognitionCompleter != null && !_recognitionCompleter!.isCompleted) { |
|
|
|
|
|
_recognitionCompleter!.complete(''); |
|
|
|
|
|
} |
|
|
|
|
|
_shouldContinueInteraction = false; |
|
|
|
|
|
}); |
|
|
|
|
|
|
|
|
return; |
|
|
return; |
|
|
} |
|
|
} |
|
|
|
|
|
|
|
|
@ -121,10 +202,17 @@ class BackgroundAgentService extends GetxService { |
|
|
|
|
|
|
|
|
// 设置新的计时器,如果 10 秒内没有新的识别结果,且系统没有在播放 TTS,则认为用户已经停止交互 |
|
|
// 设置新的计时器,如果 10 秒内没有新的识别结果,且系统没有在播放 TTS,则认为用户已经停止交互 |
|
|
_noSpeechTimer = Timer(const Duration(seconds: 10), () { |
|
|
_noSpeechTimer = Timer(const Duration(seconds: 10), () { |
|
|
// 再次检查是否正在播放 TTS |
|
|
// 再次检查是否正在播放 TTS,使用 isActuallyPlaying |
|
|
|
|
|
_isSpeaking = _ttsService.isActuallyPlaying(); |
|
|
|
|
|
|
|
|
|
|
|
print('无语音超时计时器触发,当前 TTS 播放状态: $_isSpeaking (isPlaying.value: ${_ttsService.isPlaying.value}, 实际播放状态: $_isSpeaking)'); |
|
|
|
|
|
|
|
|
if (!_isSpeaking && _isListening && _recognitionCompleter != null && !_recognitionCompleter!.isCompleted) { |
|
|
if (!_isSpeaking && _isListening && _recognitionCompleter != null && !_recognitionCompleter!.isCompleted) { |
|
|
print('10秒内没有新的语音输入,且系统没有在播放 TTS,退出循环交互'); |
|
|
print('10秒内没有新的语音输入,且系统没有在播放 TTS,退出循环交互'); |
|
|
_recognitionCompleter!.complete(''); |
|
|
_recognitionCompleter!.complete(''); |
|
|
|
|
|
|
|
|
|
|
|
// 直接设置标志,停止循环交互 |
|
|
|
|
|
_shouldContinueInteraction = false; |
|
|
} |
|
|
} |
|
|
}); |
|
|
}); |
|
|
|
|
|
|
|
|
@ -138,6 +226,9 @@ class BackgroundAgentService extends GetxService { |
|
|
_hasFinalResult = false; |
|
|
_hasFinalResult = false; |
|
|
|
|
|
|
|
|
try { |
|
|
try { |
|
|
|
|
|
// 强制重置 TTS 状态,确保不会因为错误的状态导致系统卡住 |
|
|
|
|
|
_resetTtsStateIfNeeded(); |
|
|
|
|
|
|
|
|
// 先播放一个简短的提示音或提示语,表示开始监听 |
|
|
// 先播放一个简短的提示音或提示语,表示开始监听 |
|
|
try { |
|
|
try { |
|
|
await _ttsService.speak("我在听"); |
|
|
await _ttsService.speak("我在听"); |
|
|
@ -175,8 +266,8 @@ class BackgroundAgentService extends GetxService { |
|
|
_noSpeechTimer = null; |
|
|
_noSpeechTimer = null; |
|
|
|
|
|
|
|
|
// 如果没有识别到语音,则停止循环交互 |
|
|
// 如果没有识别到语音,则停止循环交互 |
|
|
if (!_hasRecognizedSpeech) { |
|
|
if (!_hasRecognizedSpeech || userInput.trim().isEmpty) { |
|
|
print('没有识别到用户语音,退出循环交互'); |
|
|
print('没有识别到用户语音或识别结果为空,退出循环交互'); |
|
|
|
|
|
|
|
|
// 停止语音识别 |
|
|
// 停止语音识别 |
|
|
if (_isListening) { |
|
|
if (_isListening) { |
|
|
@ -199,17 +290,7 @@ class BackgroundAgentService extends GetxService { |
|
|
// 只有在退出交互时才停止语音识别 |
|
|
// 只有在退出交互时才停止语音识别 |
|
|
} catch (e) { |
|
|
} catch (e) { |
|
|
print('语音识别过程出错: $e'); |
|
|
print('语音识别过程出错: $e'); |
|
|
// 如果语音识别失败,尝试使用默认问候语 |
|
|
|
|
|
userInput = '你好,请帮我回答一个问题'; |
|
|
|
|
|
|
|
|
|
|
|
// 尝试重置语音识别服务 |
|
|
|
|
|
try { |
|
|
|
|
|
await stopVoiceRecognition(); |
|
|
|
|
|
await Future.delayed(const Duration(milliseconds: 500)); |
|
|
|
|
|
await _voiceRecognitionService.initialize(); |
|
|
|
|
|
} catch (resetError) { |
|
|
|
|
|
print('重置语音识别服务失败: $resetError'); |
|
|
|
|
|
} |
|
|
|
|
|
} finally { |
|
|
} finally { |
|
|
// 清理资源,但保持语音识别状态 |
|
|
// 清理资源,但保持语音识别状态 |
|
|
_recognitionCompleter = null; |
|
|
_recognitionCompleter = null; |
|
|
@ -405,6 +486,18 @@ class BackgroundAgentService extends GetxService { |
|
|
print('收到最终识别结果,立即处理: ${event.text}'); |
|
|
print('收到最终识别结果,立即处理: ${event.text}'); |
|
|
_recognitionCompleter!.complete(event.text); |
|
|
_recognitionCompleter!.complete(event.text); |
|
|
} |
|
|
} |
|
|
|
|
|
} else { |
|
|
|
|
|
// 如果最终结果为空,但之前有中间结果,仍然标记为有语音 |
|
|
|
|
|
if (_hasRecognizedSpeech) { |
|
|
|
|
|
print('最终识别结果为空,但之前有中间结果,使用最后的中间结果'); |
|
|
|
|
|
if (_recognitionCompleter != null && !_recognitionCompleter!.isCompleted) { |
|
|
|
|
|
_recognitionCompleter!.complete(recognizedText.value); |
|
|
|
|
|
} |
|
|
|
|
|
} else { |
|
|
|
|
|
// 如果最终结果为空,且没有中间结果,重新启动无语音超时计时器 |
|
|
|
|
|
print('最终识别结果为空,且没有中间结果,重新启动无语音超时计时器'); |
|
|
|
|
|
_startNoSpeechTimer(); |
|
|
|
|
|
} |
|
|
} |
|
|
} |
|
|
print('最终识别结果: ${event.text}'); |
|
|
print('最终识别结果: ${event.text}'); |
|
|
} else if (event.type == RecognitionEventType.intermediateResult) { |
|
|
} else if (event.type == RecognitionEventType.intermediateResult) { |
|
|
@ -537,9 +630,42 @@ class BackgroundAgentService extends GetxService { |
|
|
_recognitionSubscription?.cancel(); |
|
|
_recognitionSubscription?.cancel(); |
|
|
_silenceTimer?.cancel(); |
|
|
_silenceTimer?.cancel(); |
|
|
_noSpeechTimer?.cancel(); |
|
|
_noSpeechTimer?.cancel(); |
|
|
_ttsSpeakingSubscription?.cancel(); |
|
|
|
|
|
_aiResponseSubscription?.cancel(); |
|
|
_aiResponseSubscription?.cancel(); |
|
|
_shouldContinueInteraction = false; |
|
|
_shouldContinueInteraction = false; |
|
|
super.onClose(); |
|
|
super.onClose(); |
|
|
} |
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
// 添加一个方法来强制重置 TTS 状态 |
|
|
|
|
|
void _resetTtsStateIfNeeded() { |
|
|
|
|
|
// 检查 TTS 状态,使用 isActuallyPlaying 获取实际状态 |
|
|
|
|
|
bool isCurrentlyPlaying = false; |
|
|
|
|
|
try { |
|
|
|
|
|
isCurrentlyPlaying = _ttsService.isActuallyPlaying(); |
|
|
|
|
|
} catch (e) { |
|
|
|
|
|
print('获取实际播放状态失败: $e'); |
|
|
|
|
|
isCurrentlyPlaying = _ttsService.isPlaying.value; |
|
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
// 如果实际状态与当前状态不一致,更新它 |
|
|
|
|
|
if (_isSpeaking != isCurrentlyPlaying) { |
|
|
|
|
|
print('重置前检测到 TTS 播放状态不一致,修正: _isSpeaking=$_isSpeaking, 实际=${isCurrentlyPlaying}'); |
|
|
|
|
|
_isSpeaking = isCurrentlyPlaying; |
|
|
|
|
|
|
|
|
|
|
|
// 如果实际状态与 TTS 服务的 isPlaying 不一致,尝试修正 TTS 服务的状态 |
|
|
|
|
|
if (isCurrentlyPlaying != _ttsService.isPlaying.value) { |
|
|
|
|
|
try { |
|
|
|
|
|
_ttsService.isPlaying.value = isCurrentlyPlaying; |
|
|
|
|
|
print('已修正 TTS 服务的播放状态'); |
|
|
|
|
|
} catch (e) { |
|
|
|
|
|
print('修正 TTS 播放状态失败: $e'); |
|
|
|
|
|
} |
|
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
// 更新最后一次状态变化的时间 |
|
|
|
|
|
_lastTtsStateChangeTime = DateTime.now(); |
|
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
// 检查是否需要强制重置 |
|
|
|
|
|
_checkAndUpdateTtsState(); |
|
|
|
|
|
} |
|
|
} |
|
|
} |