feat: add AI voice transcription fallback for domestic Android ROMs

This commit is contained in:
shen
2026-09-15 22:02:26 +08:00
parent 670260f8e2
commit 35b493e652
8 changed files with 683 additions and 90 deletions
+122
View File
@@ -1,3 +1,4 @@
import 'dart:io';
import 'dart:convert';
import 'package:flutter_secure_storage/flutter_secure_storage.dart';
@@ -84,6 +85,127 @@ class AiService {
return Uri.tryParse('$base/v1/chat/completions');
}
static Uri? resolveChatCompletionsUri({
required AiProviderType provider,
required String endpoint,
required String model,
}) {
final base = endpoint.trim().replaceFirst(RegExp(r'/+$'), '');
if (base.isEmpty) return null;
if (provider == AiProviderType.gemini) {
if (base.contains(':generateContent')) {
return Uri.tryParse(base);
}
return Uri.tryParse('$base/models/$model:generateContent');
}
if (base.endsWith('/responses')) {
final root = base.substring(0, base.length - '/responses'.length);
return Uri.tryParse('$root/chat/completions');
}
if (base.endsWith('/chat/completions')) {
return Uri.tryParse(base);
}
if (base.endsWith('/v1')) {
return Uri.tryParse('$base/chat/completions');
}
return Uri.tryParse('$base/v1/chat/completions');
}
/// Transcribes spoken audio file to English text using the configured AI multimodal model.
Future<String?> transcribeAudio({
required String filePath,
required AiProviderType provider,
required String endpoint,
required String model,
}) async {
if (provider == AiProviderType.mock) return null;
final file = File(filePath);
if (!await file.exists()) return null;
final bytes = await file.readAsBytes();
if (bytes.isEmpty) return null;
final key = await resolveApiKey();
final uri = resolveChatCompletionsUri(
provider: provider,
endpoint: endpoint,
model: model,
);
if (key == null || key.isEmpty || uri == null) return null;
final ext = filePath.split('.').last.toLowerCase();
final format = (ext == 'wav' || ext == 'mp3' || ext == 'm4a' || ext == 'aac') ? ext : 'm4a';
final base64Data = base64Encode(bytes);
try {
if (provider == AiProviderType.gemini) {
final mimeType = format == 'wav' ? 'audio/wav' : (format == 'mp3' ? 'audio/mp3' : 'audio/mp4');
final response = await http.post(
uri,
headers: {'x-goog-api-key': key, 'Content-Type': 'application/json'},
body: jsonEncode({
'contents': [
{
'parts': [
{
'text': 'Transcribe the spoken English speech in this audio file accurately. Return ONLY the transcribed English words. If silence or unintelligible, output nothing.',
},
{
'inline_data': {
'mime_type': mimeType,
'data': base64Data,
}
}
]
}
]
}),
).timeout(const Duration(seconds: 25));
if (response.statusCode < 200 || response.statusCode >= 300) return null;
return _extractResponseContent(provider, response.body);
} else {
final response = await http.post(
uri,
headers: {
'Authorization': 'Bearer $key',
'Content-Type': 'application/json',
},
body: jsonEncode({
'model': model,
'messages': [
{
'role': 'user',
'content': [
{
'type': 'text',
'text': 'Transcribe the spoken English speech in this audio file accurately. Output ONLY the raw transcribed English words without quotes, punctuation tags, or commentary. If silence or noise, return nothing.',
},
{
'type': 'input_audio',
'input_audio': {
'data': base64Data,
'format': format,
},
}
],
}
],
'temperature': 0.1,
}),
).timeout(const Duration(seconds: 25));
if (response.statusCode < 200 || response.statusCode >= 300) return null;
final raw = _extractResponseContent(provider, response.body);
if (raw == null) return null;
var text = raw.trim();
if (text.startsWith('"') && text.endsWith('"') && text.length >= 2) {
text = text.substring(1, text.length - 1).trim();
}
return text;
}
} catch (_) {
return null;
}
}
static Map<String, dynamic> _buildOpenAiPayload({
required Uri uri,
required String model,
+6
View File
@@ -57,6 +57,12 @@ class AppState extends ChangeNotifier {
AiProviderType aiProvider = AiProviderType.mock;
String aiEndpoint = '';
String aiModel = '';
AiConfigFile get aiConfig => AiConfigFile(
provider: aiProvider,
endpoint: aiEndpoint,
model: aiModel,
);
String? cachedAdaptiveLessonRaw;
DateTime? cachedAdaptiveLessonAuditedAt;
String? cachedAdaptiveLessonAuditor;
+64 -16
View File
@@ -19,6 +19,9 @@ class VoiceService {
bool _speechReady = false;
bool _ttsInitialized = false;
void Function(String status)? _statusListener;
void Function(String error)? _errorListener;
Future<void> _initTts() async {
if (_ttsInitialized) return;
try {
@@ -65,6 +68,8 @@ class VoiceService {
} catch (_) {}
}
Future<bool> hasRecordPermission() => _recorder.hasPermission();
Future<bool> startRecording() async {
if (!await _recorder.hasPermission()) return false;
final directory = await getApplicationDocumentsDirectory();
@@ -125,27 +130,70 @@ class VoiceService {
await _player.dispose();
}
Future<bool> initializeSpeech() async {
_speechReady = await _stt.initialize();
return _speechReady;
Future<bool> initializeSpeech({
void Function(String status)? onStatus,
void Function(String error)? onError,
}) async {
_statusListener = onStatus;
_errorListener = onError;
try {
_speechReady = await _stt.initialize(
onError: (val) {
_errorListener?.call(val.errorMsg);
},
onStatus: (val) {
_statusListener?.call(val);
},
debugLogging: false,
);
return _speechReady;
} catch (_) {
_speechReady = false;
return false;
}
}
Future<bool> startListening(
void Function(String text, bool finalResult) onResult,
) async {
if (!_speechReady && !await initializeSpeech()) {
void Function(String text, bool finalResult) onResult, {
void Function(String status)? onStatus,
void Function(String error)? onError,
}) async {
_statusListener = onStatus;
_errorListener = onError;
try {
if (!_speechReady || !_stt.isAvailable) {
final ready = await initializeSpeech(onStatus: onStatus, onError: onError);
if (!ready) return false;
}
String? targetLocaleId = 'en_US';
try {
final locales = await _stt.locales();
if (locales.isNotEmpty) {
final enLocale = locales.firstWhere(
(l) => l.localeId.toLowerCase().startsWith('en'),
orElse: () => locales.first,
);
targetLocaleId = enLocale.localeId;
}
} catch (_) {}
await _stt.listen(
onResult: (result) =>
onResult(result.recognizedWords, result.finalResult),
listenOptions: SpeechListenOptions(
localeId: targetLocaleId,
listenFor: const Duration(seconds: 30),
pauseFor: const Duration(seconds: 4),
partialResults: true,
cancelOnError: false,
),
);
return _stt.isListening;
} catch (e) {
if (onError != null) onError(e.toString());
return false;
}
await _stt.listen(
onResult: (result) =>
onResult(result.recognizedWords, result.finalResult),
listenOptions: SpeechListenOptions(
localeId: 'en_US',
listenFor: const Duration(seconds: 30),
pauseFor: const Duration(seconds: 4),
),
);
return true;
}
Future<void> stopListening() => _stt.stop();