| 1 | import 'dart:async'; |
| 2 | |
| 3 | import 'package:flutter/foundation.dart'; |
| 4 | import 'package:flutter/services.dart'; |
| 5 | import 'package:flutter_riverpod/flutter_riverpod.dart'; |
| 6 | import 'package:speech_to_text/speech_to_text.dart'; |
| 7 | |
| 8 | import '../../../core/logger/app_logger.dart'; |
| 9 | import '../utils/stt_error_messages.dart'; |
| 10 | |
| 11 | /// Double in dB-ish units (iOS) or 0..1-ish (Android) reported by the |
| 12 | /// speech_to_text plugin via `onSoundLevelChange`. Higher = louder. |
| 13 | typedef SoundLevelChange = void Function(double level); |
| 14 | |
| 15 | class SttState { |
| 16 | final bool isListening; |
| 17 | final bool isAvailable; |
| 18 | final String recognizedWords; |
| 19 | final String? error; |
| 20 | |
| 21 | const SttState({ |
| 22 | this.isListening = false, |
| 23 | this.isAvailable = false, |
| 24 | this.recognizedWords = '', |
| 25 | this.error, |
| 26 | }); |
| 27 | |
| 28 | SttState copyWith({ |
| 29 | bool? isListening, |
| 30 | bool? isAvailable, |
| 31 | String? recognizedWords, |
| 32 | String? error, |
| 33 | bool clearError = false, |
| 34 | }) { |
| 35 | return SttState( |
| 36 | isListening: isListening ?? this.isListening, |
| 37 | isAvailable: isAvailable ?? this.isAvailable, |
| 38 | recognizedWords: recognizedWords ?? this.recognizedWords, |
| 39 | error: clearError ? null : (error ?? this.error), |
| 40 | ); |
| 41 | } |
| 42 | } |
| 43 | |
| 44 | class SttNotifier extends Notifier<SttState> { |
| 45 | SttNotifier({SpeechToText Function()? speechFactory}) |
| 46 | : _speechFactory = speechFactory ?? SpeechToText.new; |
| 47 | |
| 48 | /// Delay before re-listening after an `error_client`, giving the platform |
| 49 | /// recognizer time to release the previous session. |
| 50 | static const clientErrorRetryDelay = Duration(milliseconds: 300); |
| 51 | |
| 52 | final SpeechToText Function() _speechFactory; |
| 53 | late SpeechToText _speech; |
| 54 | bool _isInit = false; |
| 55 | Future<bool>? _initialization; |
| 56 | // Monotonically increasing counter; status callbacks from a previous |
| 57 | // session cannot clobber state set by a newer one. |
| 58 | int _session = 0; |
| 59 | // Snapshot of the most recently issued session id. Set to -1 when no |
| 60 | // listen() is in flight. |
| 61 | int _activeSession = -1; |
| 62 | // Re-issues the current listen() with the same callbacks. Null when no |
| 63 | // listen is in flight. |
| 64 | Future<void> Function()? _relisten; |
| 65 | bool _retriedClientError = false; |
| 66 | // Android: use the on-device recognizer instead of the default service. |
| 67 | // Flipped after an `error_client`, which is what devices without Google |
| 68 | // speech services (e.g. GrapheneOS) report for the default one (#100). |
| 69 | bool _useOnDeviceRecognizer = false; |
| 70 | bool _switchedRecognizerForRetry = false; |
| 71 | |
| 72 | @override |
| 73 | SttState build() { |
| 74 | _speech = _speechFactory(); |
| 75 | ref.onDispose(() { |
| 76 | try { |
| 77 | _speech.cancel(); |
| 78 | } catch (e) { |
| 79 | Log.error('STT dispose/cancel error: $e'); |
| 80 | } |
| 81 | }); |
| 82 | return const SttState(); |
| 83 | } |
| 84 | |
| 85 | Future<bool> initSpeech() { |
| 86 | if (_isInit) return Future.value(state.isAvailable); |
| 87 | return _initialization ??= _initializeSpeech(); |
| 88 | } |
| 89 | |
| 90 | Future<bool> _initializeSpeech() async { |
| 91 | try { |
| 92 | final available = await _speech.initialize( |
| 93 | onError: (val) { |
| 94 | if (!ref.mounted) return; |
| 95 | Log.error('STT error: ${val.errorMsg} - permanent: ${val.permanent}'); |
| 96 | if (val.errorMsg == 'error_client' && _retryAfterClientError()) { |
| 97 | return; |
| 98 | } |
| 99 | if (_switchedRecognizerForRetry) { |
| 100 | // The other recognizer didn't help either; go back to the |
| 101 | // original one for the next attempt. |
| 102 | _useOnDeviceRecognizer = !_useOnDeviceRecognizer; |
| 103 | _switchedRecognizerForRetry = false; |
| 104 | } |
| 105 | state = state.copyWith(error: val.errorMsg, isListening: false); |
| 106 | }, |
| 107 | onStatus: (val) { |
| 108 | if (!ref.mounted) return; |
| 109 | Log.debug('STT status: $val'); |
| 110 | // Only honour status updates from the active session; stale |
| 111 | // callbacks from a previous listen() (or callbacks fired while |
| 112 | // we've already stopped the current one) must not clobber state. |
| 113 | if (_activeSession != _session) return; |
| 114 | if (val == 'listening') { |
| 115 | state = state.copyWith(isListening: true); |
| 116 | } else if (val == 'notListening' || val == 'done') { |
| 117 | state = state.copyWith(isListening: false); |
| 118 | } |
| 119 | }, |
| 120 | options: [SpeechToText.androidNoBluetooth], |
| 121 | ); |
| 122 | if (!ref.mounted) return false; |
| 123 | // A denied permission must remain retryable after the user changes it |
| 124 | // in Android Settings. Do not cache an unsuccessful initialization. |
| 125 | _isInit = available; |
| 126 | state = state.copyWith(isAvailable: available, clearError: available); |
| 127 | return available; |
| 128 | } catch (e) { |
| 129 | if (!ref.mounted) return false; |
| 130 | Log.error('STT initialization failed: $e'); |
| 131 | final noRecognizer = |
| 132 | e is PlatformException && e.code == 'recognizerNotAvailable'; |
| 133 | state = state.copyWith( |
| 134 | isAvailable: false, |
| 135 | error: noRecognizer ? sttUnavailableCode : e.toString(), |
| 136 | ); |
| 137 | return false; |
| 138 | } finally { |
| 139 | _initialization = null; |
| 140 | } |
| 141 | } |
| 142 | |
| 143 | Future<void> startListening({ |
| 144 | required void Function(String) onResult, |
| 145 | void Function(String)? onFinal, |
| 146 | SoundLevelChange? onSoundLevelChange, |
| 147 | }) async { |
| 148 | final available = await initSpeech(); |
| 149 | if (!ref.mounted) return; |
| 150 | if (!available) { |
| 151 | state = state.copyWith( |
| 152 | error: 'Speech recognition not available or permission denied', |
| 153 | ); |
| 154 | return; |
| 155 | } |
| 156 | |
| 157 | state = state.copyWith( |
| 158 | isListening: true, |
| 159 | recognizedWords: '', |
| 160 | clearError: true, |
| 161 | ); |
| 162 | _retriedClientError = false; |
| 163 | _switchedRecognizerForRetry = false; |
| 164 | _relisten = () => _listen( |
| 165 | onResult: onResult, |
| 166 | onFinal: onFinal, |
| 167 | onSoundLevelChange: onSoundLevelChange, |
| 168 | ); |
| 169 | await _relisten!(); |
| 170 | } |
| 171 | |
| 172 | Future<void> _listen({ |
| 173 | required void Function(String) onResult, |
| 174 | void Function(String)? onFinal, |
| 175 | SoundLevelChange? onSoundLevelChange, |
| 176 | }) async { |
| 177 | _session++; |
| 178 | _activeSession = _session; |
| 179 | |
| 180 | try { |
| 181 | await _speech.listen( |
| 182 | onResult: (result) { |
| 183 | if (!ref.mounted) return; |
| 184 | state = state.copyWith(recognizedWords: result.recognizedWords); |
| 185 | onResult(result.recognizedWords); |
| 186 | if (result.finalResult && onFinal != null) { |
| 187 | onFinal(result.recognizedWords); |
| 188 | } |
| 189 | }, |
| 190 | onSoundLevelChange: (level) { |
| 191 | // iOS reports negative dB (-2 quiet, 10 loud), Android ~0..1. |
| 192 | final normalised = level <= 0 |
| 193 | ? ((level + 2) / 12).clamp(0.0, 1.0) |
| 194 | : level.clamp(0.0, 1.0); |
| 195 | onSoundLevelChange?.call(normalised); |
| 196 | }, |
| 197 | listenOptions: SpeechListenOptions( |
| 198 | listenFor: const Duration(seconds: 30), |
| 199 | pauseFor: const Duration(seconds: 5), |
| 200 | partialResults: true, |
| 201 | cancelOnError: true, |
| 202 | onDevice: _useOnDeviceRecognizer, |
| 203 | ), |
| 204 | ); |
| 205 | } catch (e) { |
| 206 | if (!ref.mounted) return; |
| 207 | Log.error('STT listen error: $e'); |
| 208 | state = state.copyWith(isListening: false, error: e.toString()); |
| 209 | } |
| 210 | } |
| 211 | |
| 212 | /// Retries the current listen once after an `error_client`. On Android the |
| 213 | /// retry switches between the default and the on-device recognizer. |
| 214 | /// Returns false when the error should be surfaced instead. |
| 215 | bool _retryAfterClientError() { |
| 216 | final relisten = _relisten; |
| 217 | if (relisten == null || _retriedClientError) return false; |
| 218 | _retriedClientError = true; |
| 219 | if (defaultTargetPlatform == TargetPlatform.android) { |
| 220 | _useOnDeviceRecognizer = !_useOnDeviceRecognizer; |
| 221 | _switchedRecognizerForRetry = true; |
| 222 | } |
| 223 | Log.warning( |
| 224 | 'STT error_client, retrying once ' |
| 225 | '(onDevice: $_useOnDeviceRecognizer)', |
| 226 | ); |
| 227 | final session = _session; |
| 228 | unawaited(() async { |
| 229 | try { |
| 230 | await _speech.cancel(); |
| 231 | } catch (e) { |
| 232 | Log.error('STT cancel before retry failed: $e'); |
| 233 | } |
| 234 | await Future<void>.delayed(clientErrorRetryDelay); |
| 235 | // Bail if the caller stopped or restarted listening meanwhile. |
| 236 | if (!ref.mounted || _session != session || _relisten != relisten) { |
| 237 | return; |
| 238 | } |
| 239 | await relisten(); |
| 240 | }()); |
| 241 | return true; |
| 242 | } |
| 243 | |
| 244 | Future<void> stopListening() async { |
| 245 | // Invalidate the current session so any in-flight status callbacks |
| 246 | // are ignored. |
| 247 | _relisten = null; |
| 248 | _session++; |
| 249 | _activeSession = -1; |
| 250 | try { |
| 251 | await _speech.stop(); |
| 252 | if (!ref.mounted) return; |
| 253 | state = state.copyWith(isListening: false); |
| 254 | } catch (e) { |
| 255 | Log.error('STT stop error: $e'); |
| 256 | } |
| 257 | } |
| 258 | |
| 259 | Future<void> cancelListening() async { |
| 260 | _relisten = null; |
| 261 | _session++; |
| 262 | _activeSession = -1; |
| 263 | try { |
| 264 | await _speech.cancel(); |
| 265 | if (!ref.mounted) return; |
| 266 | state = state.copyWith(isListening: false); |
| 267 | } catch (e) { |
| 268 | Log.error('STT cancel error: $e'); |
| 269 | } |
| 270 | } |
| 271 | } |
| 272 | |
| 273 | final sttProvider = NotifierProvider<SttNotifier, SttState>(() { |
| 274 | return SttNotifier(); |
| 275 | }); |