// // KBVoiceToTextManager.m // keyBoard // // Created by Mac on 2026/1/26. // #import "KBVoiceToTextManager.h" #import "DeepgramStreamingManager.h" #import "KBVoiceInputBar.h" @interface KBVoiceToTextManager () @property(nonatomic, strong) DeepgramStreamingManager *deepgramManager; @property(nonatomic, weak) KBVoiceInputBar *inputBar; @property(nonatomic, strong) NSMutableString *fullText; @end @implementation KBVoiceToTextManager - (void)setDeepgramEnabled:(BOOL)deepgramEnabled { if (_deepgramEnabled == deepgramEnabled) { return; } _deepgramEnabled = deepgramEnabled; if (!deepgramEnabled) { [self.deepgramManager cancel]; [self resetTranscript]; } else { [self.deepgramManager prepareConnection]; } } - (instancetype)initWithInputBar:(KBVoiceInputBar *)inputBar { self = [super init]; if (self) { _inputBar = inputBar; _inputBar.delegate = self; _fullText = [[NSMutableString alloc] init]; _deepgramEnabled = YES; [self setupDeepgram]; } return self; } - (void)dealloc { [self.deepgramManager disconnect]; } #pragma mark - Public Methods - (void)prepareConnection { if (!self.deepgramEnabled) { return; } [self.deepgramManager prepareConnection]; } - (void)disconnect { if (!self.deepgramEnabled) { return; } [self.deepgramManager disconnect]; } #pragma mark - Private Methods - (void)setupDeepgram { self.deepgramManager = [[DeepgramStreamingManager alloc] init]; self.deepgramManager.delegate = self; self.deepgramManager.serverURL = @"wss://api.deepgram.com/v1/listen"; self.deepgramManager.apiKey = @"9c792eb63a65d644cbc95785155754cd1e84f8cf"; self.deepgramManager.language = @"en"; self.deepgramManager.model = @"nova-3"; self.deepgramManager.punctuate = YES; self.deepgramManager.smartFormat = YES; self.deepgramManager.interimResults = YES; self.deepgramManager.encoding = @"linear16"; self.deepgramManager.sampleRate = 16000.0; self.deepgramManager.channels = 1; } - (void)resetTranscript { [self.fullText setString:@""]; } #pragma mark - KBVoiceInputBarDelegate - (void)voiceInputBarDidBeginRecording:(KBVoiceInputBar *)inputBar { [self resetTranscript]; if (self.deepgramEnabled) { inputBar.statusText = @"正在连接..."; [self.deepgramManager start]; } else { inputBar.statusText = @"正在录音..."; } if ([self.delegate respondsToSelector:@selector (voiceToTextManagerDidBeginRecording:)]) { [self.delegate voiceToTextManagerDidBeginRecording:self]; } } - (void)voiceInputBarDidEndRecording:(KBVoiceInputBar *)inputBar { if (self.deepgramEnabled) { inputBar.statusText = @"正在识别..."; [self.deepgramManager stopAndFinalize]; } else { inputBar.statusText = @"录音结束"; } if ([self.delegate respondsToSelector:@selector (voiceToTextManagerDidEndRecording:)]) { [self.delegate voiceToTextManagerDidEndRecording:self]; } } - (void)voiceInputBarDidCancelRecording:(KBVoiceInputBar *)inputBar { inputBar.statusText = @"已取消"; [self resetTranscript]; if (self.deepgramEnabled) { [self.deepgramManager cancel]; } if ([self.delegate respondsToSelector:@selector (voiceToTextManagerDidCancelRecording:)]) { [self.delegate voiceToTextManagerDidCancelRecording:self]; } } #pragma mark - DeepgramStreamingManagerDelegate - (void)deepgramStreamingManagerDidConnect { if (!self.deepgramEnabled) { return; } self.inputBar.statusText = @"正在聆听..."; } - (void)deepgramStreamingManagerDidDisconnect:(NSError *_Nullable)error { if (!self.deepgramEnabled) { return; } if (!error) { return; } self.inputBar.statusText = @"识别失败"; if ([self.delegate respondsToSelector:@selector (voiceToTextManager:didFailWithError:)]) { [self.delegate voiceToTextManager:self didFailWithError:error]; } } - (void)deepgramStreamingManagerDidUpdateRMS:(float)rms { if (!self.deepgramEnabled) { return; } [self.inputBar updateVolumeRMS:rms]; } - (void)deepgramStreamingManagerDidReceiveInterimTranscript:(NSString *)text { if (!self.deepgramEnabled) { return; } NSString *displayText = text ?: @""; if (self.fullText.length > 0 && displayText.length > 0) { displayText = [NSString stringWithFormat:@"%@ %@", self.fullText, displayText]; } else if (self.fullText.length > 0) { displayText = [self.fullText copy]; } self.inputBar.statusText = displayText.length > 0 ? displayText : @"正在识别..."; if ([self.delegate respondsToSelector:@selector (voiceToTextManager:didUpdateInterimText:)]) { [self.delegate voiceToTextManager:self didUpdateInterimText:displayText]; } } - (void)deepgramStreamingManagerDidReceiveFinalTranscript:(NSString *)text { if (!self.deepgramEnabled) { return; } if (text.length > 0) { if (self.fullText.length > 0) { [self.fullText appendString:@" "]; } [self.fullText appendString:text]; } NSString *finalText = [self.fullText copy]; self.inputBar.statusText = finalText.length > 0 ? finalText : @"识别完成"; if (finalText.length > 0 && [self.delegate respondsToSelector:@selector (voiceToTextManager:didReceiveFinalText:)]) { [self.delegate voiceToTextManager:self didReceiveFinalText:finalText]; } } - (void)deepgramStreamingManagerDidFail:(NSError *)error { if (!self.deepgramEnabled) { return; } self.inputBar.statusText = @"识别失败"; if ([self.delegate respondsToSelector:@selector (voiceToTextManager:didFailWithError:)]) { [self.delegate voiceToTextManager:self didFailWithError:error]; } } @end