feat(callkit): Add CallKit service for native call UI and permissions

android/app/proguard-rules.pro

@@ -18,4 +18,7 @@
 
 # Keep WebSocket functionality
 -keep class org.java_websocket.** { *; }
--dontwarn org.java_websocket.**
+-dontwarn org.java_websocket.**
+
+# Keep Flutter CallKit Incoming classes
+-keep class com.hiennv.flutter_callkit_incoming.** { *; }

android/app/src/main/AndroidManifest.xml

@@ -47,7 +47,7 @@
         <activity
             android:name=".MainActivity"
             android:exported="true"
-            android:launchMode="singleTop"
+            android:launchMode="singleInstance"
             android:theme="@style/LaunchTheme"
             android:configChanges="orientation|keyboardHidden|keyboard|screenSize|smallestScreenSize|locale|layoutDirection|fontScale|screenLayout|density|uiMode"
             android:hardwareAccelerated="true"

ios/Podfile.lock

@@ -4,6 +4,7 @@ PODS:
     - FlutterMacOS
   - connectivity_plus (0.0.1):
     - Flutter
+  - CryptoSwift (1.8.4)
   - DKImagePickerController/Core (4.3.9):
     - DKImagePickerController/ImageDataManager
     - DKImagePickerController/Resource
@@ -39,6 +40,9 @@ PODS:
     - DKImagePickerController/PhotoGallery
     - Flutter
   - Flutter (1.0.0)
+  - flutter_callkit_incoming (0.0.1):
+    - CryptoSwift
+    - Flutter
   - flutter_local_notifications (0.0.1):
     - Flutter
   - flutter_native_splash (2.4.3):
@@ -98,6 +102,7 @@ DEPENDENCIES:
   - connectivity_plus (from `.symlinks/plugins/connectivity_plus/ios`)
   - file_picker (from `.symlinks/plugins/file_picker/ios`)
   - Flutter (from `Flutter`)
+  - flutter_callkit_incoming (from `.symlinks/plugins/flutter_callkit_incoming/ios`)
   - flutter_local_notifications (from `.symlinks/plugins/flutter_local_notifications/ios`)
   - flutter_native_splash (from `.symlinks/plugins/flutter_native_splash/ios`)
   - flutter_secure_storage (from `.symlinks/plugins/flutter_secure_storage/ios`)
@@ -119,6 +124,7 @@ DEPENDENCIES:
 
 SPEC REPOS:
   trunk:
+    - CryptoSwift
     - DKImagePickerController
     - DKPhotoGallery
     - onnxruntime-c
@@ -135,6 +141,8 @@ EXTERNAL SOURCES:
     :path: ".symlinks/plugins/file_picker/ios"
   Flutter:
     :path: Flutter
+  flutter_callkit_incoming:
+    :path: ".symlinks/plugins/flutter_callkit_incoming/ios"
   flutter_local_notifications:
     :path: ".symlinks/plugins/flutter_local_notifications/ios"
   flutter_native_splash:
@@ -175,10 +183,12 @@ EXTERNAL SOURCES:
 SPEC CHECKSUMS:
   audioplayers_darwin: 4f9ca89d92d3d21cec7ec580e78ca888e5fb68bd
   connectivity_plus: cb623214f4e1f6ef8fe7403d580fdad517d2f7dd
+  CryptoSwift: e64e11850ede528a02a0f3e768cec8e9d92ecb90
   DKImagePickerController: 946cec48c7873164274ecc4624d19e3da4c1ef3c
   DKPhotoGallery: b3834fecb755ee09a593d7c9e389d8b5d6deed60
   file_picker: a0560bc09d61de87f12d246fc47d2119e6ef37be
   Flutter: cabc95a1d2626b1b06e7179b784ebcf0c0cde467
+  flutter_callkit_incoming: cb8138af67cda6dd981f7101a5d709003af21502
   flutter_local_notifications: a5a732f069baa862e728d839dd2ebb904737effb
   flutter_native_splash: c32d145d68aeda5502d5f543ee38c192065986cf
   flutter_secure_storage: 1ed9476fba7e7a782b22888f956cce43e2c62f13

ios/Runner/Info.plist

@@ -87,11 +87,13 @@
 			</dict>
 		</array>
 		<!-- Background Modes -->
-		<key>UIBackgroundModes</key>
+	<key>UIBackgroundModes</key>
-		<array>
+	<array>
-			<string>audio</string>
+		<string>audio</string>
-			<string>processing</string>
+		<string>remote-notification</string>
-		</array>
+		<string>processing</string>
+		<string>voip</string>
+	</array>
 		<!-- Background Task Identifiers -->
 		<key>BGTaskSchedulerPermittedIdentifiers</key>
 		<array>

lib/core/providers/storage_providers.dart

@@ -33,3 +33,4 @@ final optimizedStorageServiceProvider = Provider<OptimizedStorageService>((
     workerManager: ref.watch(workerManagerProvider),
   );
 });
+

lib/core/services/callkit_service.dart

@@ -0,0 +1,184 @@
+import 'dart:developer' as developer;
+
+import 'package:flutter_callkit_incoming/entities/entities.dart';
+import 'package:flutter_callkit_incoming/flutter_callkit_incoming.dart';
+import 'package:riverpod_annotation/riverpod_annotation.dart';
+import 'package:uuid/uuid.dart';
+
+part 'callkit_service.g.dart';
+
+/// Thin wrapper around `flutter_callkit_incoming` for voice calls.
+class CallKitService {
+  CallKitService({Uuid? uuid}) : _uuid = uuid ?? const Uuid();
+
+  final Uuid _uuid;
+  static const int _defaultCallDurationMs = 2 * 60 * 60 * 1000; // 2 hours
+
+  /// Requests the notification/full-screen intent permissions needed on Android.
+  Future<void> requestPermissions() async {
+    await _safe(
+      () => FlutterCallkitIncoming.requestNotificationPermission(
+        <String, dynamic>{
+          'title': 'Notification permission',
+          'rationaleMessagePermission':
+              'Call alerts need notification permission.',
+          'postNotificationMessageRequired':
+              'Allow notifications to show incoming calls.',
+        },
+      ),
+    );
+
+    await _safe(() => FlutterCallkitIncoming.requestFullIntentPermission());
+  }
+
+  /// Starts an outgoing call with the native UI and returns the call id.
+  Future<String?> startOutgoingVoiceCall({
+    required String calleeName,
+    required String handle,
+    String? avatar,
+    int? durationMs,
+  }) async {
+    final id = _uuid.v4();
+    final params = _buildParams(
+      id: id,
+      callerName: calleeName,
+      handle: handle,
+      avatar: avatar,
+      durationMs: durationMs ?? _defaultCallDurationMs,
+    );
+    try {
+      await FlutterCallkitIncoming.startCall(params);
+      return id;
+    } catch (error, stackTrace) {
+      developer.log(
+        'CallKit startCall failed: $error',
+        name: 'callkit',
+        error: error,
+        stackTrace: stackTrace,
+      );
+      return null;
+    }
+  }
+
+  /// Marks the current call as connected so iOS shows an incrementing timer.
+  Future<void> markCallConnected(String id) async {
+    try {
+      await FlutterCallkitIncoming.setCallConnected(id);
+    } catch (error, stackTrace) {
+      developer.log(
+        'CallKit setCallConnected failed: $error',
+        name: 'callkit',
+        error: error,
+        stackTrace: stackTrace,
+      );
+    }
+  }
+
+  /// Ends a specific call id.
+  Future<void> endCall(String id) async {
+    await _safe(() => FlutterCallkitIncoming.endCall(id));
+  }
+
+  /// Clears all ongoing/missed calls.
+  Future<void> endAllCalls() async {
+    await _safe(FlutterCallkitIncoming.endAllCalls);
+  }
+
+  /// Returns the platform VOIP token (iOS PushKit) when available.
+  Future<String?> getVoipToken() async {
+    final token = await _safe<dynamic>(
+      () => FlutterCallkitIncoming.getDevicePushTokenVoIP(),
+    );
+    if (token == null) return null;
+    if (token is String) return token;
+    return token.toString();
+  }
+
+  /// Returns the raw active call list from the plugin.
+  Future<List<Map<String, dynamic>>> activeCalls() async {
+    final calls = await _safe<dynamic>(FlutterCallkitIncoming.activeCalls);
+    if (calls is List) {
+      return calls
+          .whereType<Map<dynamic, dynamic>>()
+          .map(Map<String, dynamic>.from)
+          .toList();
+    }
+    return <Map<String, dynamic>>[];
+  }
+
+  /// Stream of CallKit events from the native layer.
+  Stream<CallEvent> get events =>
+      FlutterCallkitIncoming.onEvent.where((event) => event != null).cast();
+
+  CallKitParams _buildParams({
+    required String id,
+    required String callerName,
+    required String handle,
+    String? avatar,
+    int durationMs = _defaultCallDurationMs,
+  }) {
+    return CallKitParams(
+      id: id,
+      nameCaller: callerName,
+      appName: 'Conduit',
+      avatar: avatar,
+      handle: handle,
+      type: 0, // 0 = audio call
+      duration: durationMs,
+      textAccept: 'Accept',
+      textDecline: 'Decline',
+      missedCallNotification: const NotificationParams(
+        showNotification: true,
+        isShowCallback: true,
+        subtitle: 'Missed call',
+        callbackText: 'Call back',
+      ),
+      callingNotification: const NotificationParams(
+        showNotification: true,
+        isShowCallback: true,
+        subtitle: 'Calling...',
+        callbackText: 'Hang up',
+      ),
+      extra: const <String, dynamic>{'transport': 'voice'},
+      android: const AndroidParams(
+        isCustomNotification: true,
+        isShowLogo: true,
+        ringtonePath: 'system_ringtone_default',
+        backgroundColor: '#0D1726',
+        actionColor: '#4CAF50',
+        incomingCallNotificationChannelName: 'Incoming Call',
+        missedCallNotificationChannelName: 'Missed Call',
+      ),
+      ios: const IOSParams(
+        handleType: 'generic',
+        supportsVideo: false,
+        audioSessionMode: 'default',
+        audioSessionActive: true,
+        audioSessionPreferredSampleRate: 44100.0,
+        audioSessionPreferredIOBufferDuration: 0.005,
+        supportsDTMF: true,
+        supportsHolding: true,
+        supportsGrouping: false,
+        supportsUngrouping: false,
+        ringtonePath: 'system_ringtone_default',
+      ),
+    );
+  }
+
+  Future<T?> _safe<T>(Future<T> Function() action) async {
+    try {
+      return await action();
+    } catch (error, stackTrace) {
+      developer.log(
+        'CallKit error: $error',
+        name: 'callkit',
+        error: error,
+        stackTrace: stackTrace,
+      );
+      return null;
+    }
+  }
+}
+
+@Riverpod(keepAlive: true)
+CallKitService callKitService(Ref ref) => CallKitService();

lib/features/chat/services/text_to_speech_service.dart

@@ -69,6 +69,22 @@ class TextToSpeechService {
           break;
       }
     });
+
+    if (!kIsWeb && Platform.isIOS) {
+      final context = AudioContext(
+        iOS: AudioContextIOS(
+          category: AVAudioSessionCategory.playAndRecord,
+          options: const {
+            AVAudioSessionOptions.defaultToSpeaker,
+            AVAudioSessionOptions.mixWithOthers,
+            AVAudioSessionOptions.allowBluetooth,
+            AVAudioSessionOptions.allowBluetoothA2DP,
+          },
+        ),
+        android: const AudioContextAndroid(),
+      );
+      _player.setAudioContext(context);
+    }
   }
 
   Future<void> _configureDeviceEngine({
@@ -87,12 +103,13 @@ class TextToSpeechService {
 
       if (!kIsWeb && Platform.isIOS) {
         await _tts.setSharedInstance(true);
-        await _tts.setIosAudioCategory(IosTextToSpeechAudioCategory.playback, [
+        await _tts
-          IosTextToSpeechAudioCategoryOptions.mixWithOthers,
+            .setIosAudioCategory(IosTextToSpeechAudioCategory.playAndRecord, [
-          IosTextToSpeechAudioCategoryOptions.defaultToSpeaker,
+              IosTextToSpeechAudioCategoryOptions.mixWithOthers,
-          IosTextToSpeechAudioCategoryOptions.allowBluetooth,
+              IosTextToSpeechAudioCategoryOptions.defaultToSpeaker,
-          IosTextToSpeechAudioCategoryOptions.allowBluetoothA2DP,
+              IosTextToSpeechAudioCategoryOptions.allowBluetooth,
-        ]);
+              IosTextToSpeechAudioCategoryOptions.allowBluetoothA2DP,
+            ]);
       }
 
       if (_engine != TtsEngine.server) {

lib/features/chat/services/voice_call_service.dart

@@ -1,12 +1,15 @@
 import 'dart:async';
 import 'dart:collection';
+import 'dart:developer' as developer;
 
 import 'package:audioplayers/audioplayers.dart';
+import 'package:flutter_callkit_incoming/entities/call_event.dart';
 import 'package:riverpod_annotation/riverpod_annotation.dart';
 import 'package:wakelock_plus/wakelock_plus.dart';
 
 import '../../../core/providers/app_providers.dart';
 import '../../../core/services/background_streaming_handler.dart';
+import '../../../core/services/callkit_service.dart';
 import '../../../core/services/socket_service.dart';
 import '../../../core/utils/markdown_to_text.dart';
 import '../providers/chat_providers.dart';
@@ -38,6 +41,7 @@ class VoiceCallService {
   final TextToSpeechService _tts;
   final SocketService _socketService;
   final Ref _ref;
+  final CallKitService _callKitService;
   final VoiceCallNotificationService _notificationService =
       VoiceCallNotificationService();
 
@@ -64,6 +68,9 @@ class VoiceCallService {
   bool _serverPipelineActive = false;
   int _nextServerChunkId = 0;
   int _nextServerPlaybackId = 0;
+  bool _callKitPermissionsRequested = false;
+  String? _callKitCallId;
+  bool _callKitConnectedReported = false;
 
   final StreamController<VoiceCallState> _stateController =
       StreamController<VoiceCallState>.broadcast();
@@ -73,15 +80,18 @@ class VoiceCallService {
       StreamController<String>.broadcast();
   final StreamController<int> _intensityController =
       StreamController<int>.broadcast();
+  StreamSubscription<CallEvent>? _callKitEventSubscription;
 
   VoiceCallService({
     required VoiceInputService voiceInput,
     required TextToSpeechService tts,
     required SocketService socketService,
+    required CallKitService callKitService,
     required Ref ref,
   }) : _voiceInput = voiceInput,
        _tts = tts,
        _socketService = socketService,
+       _callKitService = callKitService,
        _ref = ref {
     _tts.bindHandlers(
       onStart: _handleTtsStart,
@@ -161,10 +171,124 @@ class VoiceCallService {
     );
   }
 
+  Future<void> _ensureCallKitPermissions() async {
+    if (_callKitPermissionsRequested) return;
+    _callKitPermissionsRequested = true;
+    await _callKitService.requestPermissions();
+  }
+
+  Future<void> _startCallKitSession({required String modelName}) async {
+    try {
+      await _ensureCallKitPermissions();
+      final callId = await _callKitService.startOutgoingVoiceCall(
+        calleeName: modelName,
+        handle: 'Conduit AI',
+      );
+      _callKitCallId = callId;
+      _callKitConnectedReported = false;
+    } catch (error, stackTrace) {
+      developer.log(
+        'CallKit outgoing setup failed: $error',
+        name: 'voice_call',
+        error: error,
+        stackTrace: stackTrace,
+      );
+    }
+  }
+
+  Future<void> _endCallKitSession() async {
+    if (_callKitCallId == null) {
+      return;
+    }
+    final callId = _callKitCallId;
+    _callKitCallId = null;
+    _callKitConnectedReported = false;
+    try {
+      await _callKitService.endCall(callId!);
+    } catch (error, stackTrace) {
+      developer.log(
+        'CallKit endCall failed: $error',
+        name: 'voice_call',
+        error: error,
+        stackTrace: stackTrace,
+      );
+      await _callKitService.endAllCalls();
+    }
+  }
+
+  void _listenForCallKitEvents() {
+    _callKitEventSubscription?.cancel();
+    if (_callKitCallId == null) return;
+
+    _callKitEventSubscription = _callKitService.events.listen((callEvent) {
+      final eventId = _extractCallId(callEvent.body);
+      if (_callKitCallId != null &&
+          eventId != null &&
+          eventId != _callKitCallId) {
+        return;
+      }
+
+      switch (callEvent.event) {
+        case Event.actionCallEnded:
+        case Event.actionCallDecline:
+        case Event.actionCallTimeout:
+          if (_state != VoiceCallState.disconnected) {
+            unawaited(stopCall());
+          }
+          break;
+        case Event.actionCallToggleMute:
+          _handleCallKitMute(callEvent.body);
+          break;
+        case Event.actionCallToggleHold:
+          _handleCallKitHold(callEvent.body);
+          break;
+        case Event.actionCallConnected:
+          unawaited(_markCallKitConnected());
+          break;
+        default:
+          break;
+      }
+    });
+  }
+
+  void _handleCallKitMute(dynamic body) {
+    final isMuted = body is Map ? body['isMuted'] == true : body == true;
+    if (_isMuted != isMuted) {
+      _toggleMute();
+    }
+  }
+
+  void _handleCallKitHold(dynamic body) {
+    final onHold = body is Map ? body['isOnHold'] == true : body == true;
+    if (onHold) {
+      unawaited(pauseListening(reason: VoiceCallPauseReason.system));
+    } else {
+      unawaited(resumeListening(reason: VoiceCallPauseReason.system));
+    }
+  }
+
+  Future<void> _markCallKitConnected() async {
+    if (_callKitCallId == null || _callKitConnectedReported) return;
+    await _callKitService.markCallConnected(_callKitCallId!);
+    _callKitConnectedReported = true;
+  }
+
+  String? _extractCallId(dynamic body) {
+    if (body is Map) {
+      final id = body['id'];
+      return id?.toString();
+    }
+    return null;
+  }
+
   Future<void> startCall(String? conversationId) async {
     if (_isDisposed) return;
 
     try {
+      final modelName = _ref.read(selectedModelProvider)?.name ?? 'Assistant';
+      await _startCallKitSession(modelName: modelName);
+      _listenForCallKitEvents();
+
       // Update state (this will trigger notification)
       _updateState(VoiceCallState.connecting);
 
@@ -200,15 +324,19 @@ class VoiceCallService {
 
       // Start listening for user voice input
       await _startListening();
+      await _markCallKitConnected();
     } catch (e) {
       _updateState(VoiceCallState.error);
       _keepAliveTimer?.cancel();
       _keepAliveTimer = null;
+      await _callKitEventSubscription?.cancel();
+      _callKitEventSubscription = null;
       await WakelockPlus.disable();
       await _notificationService.cancelNotification();
       await BackgroundStreamingHandler.instance.stopBackgroundExecution(const [
         _voiceCallStreamId,
       ]);
+      await _endCallKitSession();
       rethrow;
     }
   }
@@ -551,10 +679,12 @@ class VoiceCallService {
       return;
     }
     _isSpeaking = false;
+    _listeningSuspendedForSpeech = false;
     if (_serverAudioBuffer.containsKey(_nextServerPlaybackId)) {
       _maybeStartServerAudio();
       return;
     }
+    _responseCompleted = true;
     _maybeResumeListeningAfterSpeech();
   }
 
@@ -603,6 +733,7 @@ class VoiceCallService {
       return;
     }
 
+    _listeningSuspendedForSpeech = false;
     unawaited(_startListening());
   }
 
@@ -618,6 +749,8 @@ class VoiceCallService {
       unawaited(_startNextSpeechChunk());
       return;
     }
+    _responseCompleted = true;
+    _listeningSuspendedForSpeech = false;
     _maybeResumeListeningAfterSpeech();
   }
 
@@ -641,6 +774,8 @@ class VoiceCallService {
 
     await _transcriptSubscription?.cancel();
     await _intensitySubscription?.cancel();
+    await _callKitEventSubscription?.cancel();
+    _callKitEventSubscription = null;
     _socketSubscription?.dispose();
 
     await _voiceInput.stopListening();
@@ -653,6 +788,7 @@ class VoiceCallService {
 
     // Cancel notification
     await _notificationService.cancelNotification();
+    await _endCallKitSession();
 
     // Disable wake lock when call ends
     await WakelockPlus.disable();
@@ -734,6 +870,10 @@ class VoiceCallService {
   }
 
   Future<void> _updateNotification() async {
+    // When CallKit is active, rely on native UI instead of the ongoing
+    // notification to avoid duplicate surfaces.
+    if (_callKitCallId != null) return;
+
     // Skip notification for idle, error, and disconnected states
     if (_state == VoiceCallState.idle ||
         _state == VoiceCallState.error ||
@@ -801,6 +941,8 @@ class VoiceCallService {
 
     await _transcriptSubscription?.cancel();
     await _intensitySubscription?.cancel();
+    await _callKitEventSubscription?.cancel();
+    _callKitEventSubscription = null;
     _socketSubscription?.dispose();
 
     _voiceInput.dispose();
@@ -809,6 +951,7 @@ class VoiceCallService {
 
     // Cancel notification
     await _notificationService.cancelNotification();
+    await _endCallKitSession();
 
     // Ensure wake lock is disabled on dispose
     await WakelockPlus.disable();
@@ -830,6 +973,7 @@ VoiceCallService voiceCallService(Ref ref) {
   final api = ref.watch(apiServiceProvider);
   final tts = TextToSpeechService(api: api);
   final socketService = ref.watch(socketServiceProvider);
+  final callKit = ref.watch(callKitServiceProvider);
 
   if (socketService == null) {
     throw Exception('Socket service not available');
@@ -839,6 +983,7 @@ VoiceCallService voiceCallService(Ref ref) {
     voiceInput: voiceInput,
     tts: tts,
     socketService: socketService,
+    callKitService: callKit,
     ref: ref,
   );
 

lib/features/chat/services/voice_input_service.dart

@@ -5,13 +5,15 @@ import 'dart:typed_data';
 
 import 'package:flutter/widgets.dart';
 import 'package:flutter_riverpod/flutter_riverpod.dart';
-import 'package:record/record.dart';
+import 'package:record/record.dart'
+    hide IosAudioCategory, IosAudioCategoryOptions;
 import 'package:riverpod_annotation/riverpod_annotation.dart';
 import 'package:stts/stts.dart';
 import 'package:vad/vad.dart';
 
 import '../../../core/providers/app_providers.dart';
 import '../../../core/services/api_service.dart';
+import '../../../core/services/background_streaming_handler.dart';
 import '../../../core/services/settings_service.dart';
 
 part 'voice_input_service.g.dart';
@@ -27,6 +29,7 @@ class VoiceInputService {
   static const int _vadSampleRate = 16000;
   static const int _vadFrameSamples = 1536;
   static const Duration _localeFetchTimeout = Duration(seconds: 2);
+  static const String _backgroundSttStreamId = 'voice-input-stt';
 
   final VadHandler _vadHandler = VadHandler.create();
   final Stt _speech = Stt();
@@ -36,11 +39,13 @@ class VoiceInputService {
   bool _isInitialized = false;
   bool _isListening = false;
   bool _localSttAvailable = false;
+  bool _localSttActive = false;
   SttPreference _preference = SttPreference.deviceOnly;
   bool _usingServerStt = false;
   String? _selectedLocaleId;
   List<LocaleName> _locales = const [];
   bool _usingFallbackLocales = false;
+  Future<void>? _startingLocalStt;
   StreamController<String>? _textStreamController;
   String _currentText = '';
   StreamController<int>? _intensityController;
@@ -49,6 +54,7 @@ class VoiceInputService {
   int _lastIntensity = 0;
   Timer? _intensityDecayTimer;
   List<double>? _vadPendingSamples;
+  bool _backgroundMicPinned = false;
 
   Stream<String> get textStream =>
       _textStreamController?.stream ?? const Stream<String>.empty();
@@ -266,22 +272,38 @@ class VoiceInputService {
   Future<void> _startLocalRecognition({
     required bool allowOnlineFallback,
   }) async {
+    if (_startingLocalStt != null) {
+      await _startingLocalStt;
+    }
+    final completer = Completer<void>();
+    _startingLocalStt = completer.future;
+    _localSttActive = false;
+
+    await _ensureLocalSttReset();
+    await _configureIosAudioSession();
+
     if (_selectedLocaleId != null) {
       await _speech.setLanguage(_selectedLocaleId!);
     }
 
-    Future<void> attempt(bool offline) => _speech.start(
+    Future<void> attempt(bool offline) async {
-      SttRecognitionOptions(punctuation: true, offline: offline),
+      await _speech.start(
-    );
+        SttRecognitionOptions(punctuation: true, offline: offline),
+      );
+      _localSttActive = true;
+    }
 
     try {
       await attempt(true);
     } catch (error) {
+      _localSttActive = false;
+      await _ensureLocalSttReset();
       if (Platform.isIOS && allowOnlineFallback) {
         try {
           await attempt(false);
           return;
         } catch (secondary) {
+          await _ensureLocalSttReset();
           throw Exception(
             'On-device speech failed ($error); '
             'online fallback failed ($secondary).',
@@ -289,16 +311,25 @@ class VoiceInputService {
         }
       }
       rethrow;
+    } finally {
+      completer.complete();
+      _startingLocalStt = null;
     }
   }
 
-  Stream<String> startListening() {
+  Future<Stream<String>> startListening() async {
     if (!_isInitialized) {
       throw Exception('Voice input not initialized');
     }
 
+    if (_startingLocalStt != null) {
+      try {
+        await _startingLocalStt;
+      } catch (_) {}
+    }
+
     if (_isListening) {
-      unawaited(stopListening());
+      await stopListening();
     }
 
     _textStreamController = StreamController<String>.broadcast();
@@ -316,62 +347,81 @@ class VoiceInputService {
         canUseLocal && _preference != SttPreference.serverOnly;
     final bool shouldUseServer =
         serverAvailable &&
-        (_preference == SttPreference.serverOnly || !shouldUseLocal);
+        (_preference == SttPreference.serverOnly ||
+            (!shouldUseLocal && _preference != SttPreference.deviceOnly));
 
     if (shouldUseLocal) {
+      await _pinBackgroundMicrophone();
       _autoStopTimer?.cancel();
       _autoStopTimer = Timer(const Duration(seconds: 60), () {
         if (_isListening) {
           unawaited(_stopListening());
         }
       });
+      try {
+        final isStillAvailable = await _speech.isSupported();
+        if (!isStillAvailable && _isListening) {
+          _localSttAvailable = false;
+          _textStreamController?.addError(
+            Exception('On-device speech recognition unavailable'),
+          );
+          await _stopListening();
+          return _textStreamController!.stream;
+        }
+      } catch (_) {
+        // ignore availability check errors
+      }
 
-      Future.microtask(() async {
+      _sttResultSub = _speech.onResultChanged.listen(
-        try {
+        (SttRecognition result) {
-          final isStillAvailable = await _speech.isSupported();
+          if (!_isListening) return;
-          if (!isStillAvailable && _isListening) {
+          final prevLen = _currentText.length;
-            _localSttAvailable = false;
+          _currentText = result.text;
-            _textStreamController?.addError(
+          _textStreamController?.add(_currentText);
-              Exception('On-device speech recognition unavailable'),
+          final delta = (_currentText.length - prevLen).clamp(0, 50);
-            );
+          final mapped = (delta / 5.0).ceil();
+          _lastIntensity = mapped.clamp(0, 10);
+          try {
+            _intensityController?.add(_lastIntensity);
+          } catch (_) {}
+          if (result.isFinal) {
             unawaited(_stopListening());
           }
-        } catch (_) {
+        },
-          // ignore availability check errors
+        onError: (error) {
-        }
+          debugPrint('Local STT Error: $error');
-      });
+          _handleLocalRecognizerError(error);
-
+        },
-      _sttResultSub = _speech.onResultChanged.listen((SttRecognition result) {
-        if (!_isListening) return;
-        final prevLen = _currentText.length;
-        _currentText = result.text;
-        _textStreamController?.add(_currentText);
-        final delta = (_currentText.length - prevLen).clamp(0, 50);
-        final mapped = (delta / 5.0).ceil();
-        _lastIntensity = mapped.clamp(0, 10);
-        try {
-          _intensityController?.add(_lastIntensity);
-        } catch (_) {}
-        if (result.isFinal) {
-          unawaited(_stopListening());
-        }
-      }, onError: _handleLocalRecognizerError);
-
-      _sttStateSub = _speech.onStateChanged.listen(
-        (_) {},
-        onError: _handleLocalRecognizerError,
       );
 
-      Future(() async {
+      _sttStateSub = _speech.onStateChanged.listen(
-        try {
+        (state) {
-          await _startLocalRecognition(allowOnlineFallback: !prefersDeviceOnly);
+          debugPrint('Local STT State: $state');
-        } catch (error) {
+          if (state == SttState.start) {
-          _localSttAvailable = false;
+            _localSttActive = true;
-          if (!_isListening) return;
+          } else if (state == SttState.stop) {
-          _textStreamController?.addError(error);
+            _localSttActive = false;
-          await _stopListening();
+          }
+        },
+        onError: (error) {
+          debugPrint('Local STT State Error: $error');
+          _handleLocalRecognizerError(error);
+        },
+      );
+
+      try {
+        debugPrint('Starting local recognition...');
+        await _startLocalRecognition(allowOnlineFallback: !prefersDeviceOnly);
+        debugPrint('Local recognition started');
+      } catch (error) {
+        debugPrint('Failed to start local recognition: $error');
+        _localSttAvailable = false;
+        if (!_isListening) {
+          return _textStreamController!.stream;
         }
-      });
+        _textStreamController?.addError(error);
+        await _stopListening();
+      }
     } else if (shouldUseServer) {
       _usingServerStt = true;
       _autoStopTimer?.cancel();
@@ -406,7 +456,7 @@ class VoiceInputService {
       });
     }
 
-    return _textStreamController!.stream;
+    return _textStreamController?.stream ?? const Stream<String>.empty();
   }
 
   /// Centralized entry point to begin voice recognition.
@@ -417,7 +467,7 @@ class VoiceInputService {
     if (!hasMic) {
       throw Exception('Microphone permission not granted');
     }
-    return startListening();
+    return await startListening();
   }
 
   Future<void> stopListening() async {
@@ -453,9 +503,16 @@ class VoiceInputService {
     await _closeControllers();
 
     _usingServerStt = false;
+    await _releaseBackgroundMicrophone();
   }
 
   Future<void> _stopLocalStt() async {
+    final pendingStart = _startingLocalStt;
+    if (pendingStart != null) {
+      try {
+        await pendingStart;
+      } catch (_) {}
+    }
     if (_sttResultSub != null) {
       try {
         await _sttResultSub?.cancel();
@@ -469,11 +526,67 @@ class VoiceInputService {
       _sttStateSub = null;
     }
 
-    if (_localSttAvailable) {
+    final shouldStopStt = _localSttActive && _localSttAvailable;
+    _localSttActive = false;
+    if (shouldStopStt) {
       try {
         await _speech.stop();
       } catch (_) {}
     }
+    if (Platform.isIOS) {
+      try {
+        await _speech.ios?.setAudioSessionActive(false);
+      } catch (_) {}
+    }
+  }
+
+  Future<void> _pinBackgroundMicrophone() async {
+    if (!Platform.isIOS || _backgroundMicPinned) return;
+    try {
+      await BackgroundStreamingHandler.instance.startBackgroundExecution(const [
+        _backgroundSttStreamId,
+      ], requiresMicrophone: true);
+      _backgroundMicPinned = true;
+    } catch (_) {}
+  }
+
+  Future<void> _releaseBackgroundMicrophone() async {
+    if (!Platform.isIOS || !_backgroundMicPinned) return;
+    _backgroundMicPinned = false;
+    try {
+      await BackgroundStreamingHandler.instance.stopBackgroundExecution(const [
+        _backgroundSttStreamId,
+      ]);
+    } catch (_) {}
+  }
+
+  Future<void> _ensureLocalSttReset() async {
+    try {
+      await _speech.stop();
+    } catch (_) {}
+    if (Platform.isIOS) {
+      try {
+        await _speech.ios?.setAudioSessionActive(false);
+      } catch (_) {}
+    }
+  }
+
+  Future<void> _configureIosAudioSession() async {
+    if (!Platform.isIOS) return;
+    final ios = _speech.ios;
+    if (ios == null) return;
+    try {
+      await ios.setAudioSessionCategory(
+        category: IosAudioCategory.playAndRecord,
+        options: [
+          IosAudioCategoryOptions.allowBluetooth,
+          IosAudioCategoryOptions.allowBluetoothA2DP,
+          IosAudioCategoryOptions.defaultToSpeaker,
+          IosAudioCategoryOptions.duckOthers,
+        ],
+      );
+      await ios.setAudioSessionActive(true);
+    } catch (_) {}
   }
 
   Future<void> _startServerRecording() async {
@@ -548,7 +661,9 @@ class VoiceInputService {
 
   Future<void> _stopVadRecording() async {
     try {
-      await _vadHandler.stopListening();
+      if (_isListening) {
+        await _vadHandler.stopListening();
+      }
     } catch (_) {}
     await _vadSpeechEndSub?.cancel();
     _vadSpeechEndSub = null;

pubspec.lock

@@ -486,6 +486,14 @@ packages:
       url: "https://pub.dev"
     source: hosted
     version: "3.4.1"
+  flutter_callkit_incoming:
+    dependency: "direct main"
+    description:
+      name: flutter_callkit_incoming
+      sha256: "3589deb8b71e43f2d520a9c8a5240243f611062a8b246cdca4b1fda01fbbf9b8"
+      url: "https://pub.dev"
+    source: hosted
+    version: "3.0.0"
   flutter_lints:
     dependency: "direct dev"
     description:

pubspec.yaml

@@ -73,6 +73,7 @@ dependencies:
   connectivity_plus: ^7.0.0
   flutter_cache_manager: ^3.4.1
   http: ^1.5.0
+  flutter_callkit_incoming: ^3.0.0
   
   # Clipboard functionality is available through flutter/services (part of Flutter SDK)