keyboard/keyBoard/Class/AiTalk/VM/KBVoiceToTextManager.m

//
//  KBVoiceToTextManager.m
//  keyBoard
//
//  Created by Mac on 2026/1/26.
//

#import "KBVoiceToTextManager.h"
#import "DeepgramStreamingManager.h"
#import "KBVoiceInputBar.h"
#import "KBLocalizationManager.h"

@interface KBVoiceToTextManager () <KBVoiceInputBarDelegate,
                                    DeepgramStreamingManagerDelegate>

@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 kb_refreshDeepgramLanguage];
  [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 kb_refreshDeepgramLanguage];
  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:@""];
}

- (void)kb_refreshDeepgramLanguage {
  self.deepgramManager.language = [self kb_currentDeepgramLanguageCode];
}

- (NSString *)kb_currentDeepgramLanguageCode {
  NSString *languageCode = [KBLocalizationManager shared].currentLanguageCode ?: @"en";
  NSString *lc = languageCode.lowercaseString;
  if ([lc hasPrefix:@"es"]) { return @"es"; }
  if ([lc hasPrefix:@"id"]) { return @"id"; }
  if ([lc hasPrefix:@"pt"]) { return @"pt"; }
  if ([lc hasPrefix:@"zh-hant"] || [lc hasPrefix:@"zh_tw"] || [lc hasPrefix:@"zh-tw"] || [lc hasPrefix:@"zh-hk"]) {
    return @"zh-TW";
  }
  if ([lc hasPrefix:@"zh-hans"] || [lc hasPrefix:@"zh_cn"] || [lc hasPrefix:@"zh-cn"]) {
    return @"zh-CN";
  }
  return @"en";
}

#pragma mark - KBVoiceInputBarDelegate

- (void)voiceInputBarDidBeginRecording:(KBVoiceInputBar *)inputBar {
  [self resetTranscript];
  if (self.deepgramEnabled) {
    [self kb_refreshDeepgramLanguage];
    inputBar.statusText = KBLocalized(@"Voice Connecting...");
    [self.deepgramManager start];
  } else {
    inputBar.statusText = KBLocalized(@"Voice Recording...");
  }

  if ([self.delegate respondsToSelector:@selector
                     (voiceToTextManagerDidBeginRecording:)]) {
    [self.delegate voiceToTextManagerDidBeginRecording:self];
  }
}

- (void)voiceInputBarDidEndRecording:(KBVoiceInputBar *)inputBar {
  if (self.deepgramEnabled) {
    inputBar.statusText = KBLocalized(@"Voice Recognizing...");
    [self.deepgramManager stopAndFinalize];
  } else {
    inputBar.statusText = KBLocalized(@"Voice Recording Ended");
  }

  if ([self.delegate respondsToSelector:@selector
                     (voiceToTextManagerDidEndRecording:)]) {
    [self.delegate voiceToTextManagerDidEndRecording:self];
  }
}

- (void)voiceInputBarDidCancelRecording:(KBVoiceInputBar *)inputBar {
  inputBar.statusText = KBLocalized(@"Voice Cancelled");
  [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 = KBLocalized(@"Voice Listening...");
}

- (void)deepgramStreamingManagerDidDisconnect:(NSError *_Nullable)error {
  if (!self.deepgramEnabled) {
    return;
  }
  if (!error) {
    return;
  }

  self.inputBar.statusText = KBLocalized(@"Voice Recognition Failed");
  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 : KBLocalized(@"Voice Recognizing...");

  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 : KBLocalized(@"Voice Recognition Completed");

  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 = KBLocalized(@"Voice Recognition Failed");
  if ([self.delegate respondsToSelector:@selector
                     (voiceToTextManager:didFailWithError:)]) {
    [self.delegate voiceToTextManager:self didFailWithError:error];
  }
}

@end