Changed #

• Removed flutter_bloc dependency — replaced FaceDetectionBloc with FaceDetectionController (extends ChangeNotifier). No external state management dependency required.
• Removed bloc_test dev dependency — tests now use standard test() with stream collection instead of blocTest<>.
• Removed event pattern — replaced add(Event) dispatch with direct methods: initializeCamera(), pause(), resume(), close().
• Renamed FaceDetectionBlocProvider → FaceDetectionProvider — a lightweight InheritedWidget for dependency injection.
• Renamed isBlocInitializeAbove → isControllerProvidedAbove.
• Widget internals — BlocConsumer replaced with ListenableBuilder.
• State file is now a standalone import (no longer a part file).

Migration (from 0.2.x) #

• Replace FaceDetectionBloc with FaceDetectionController.
• Replace FaceDetectionBlocProvider(bloc: ...) with FaceDetectionProvider(controller: ...).
• Replace bloc.add(const InitializeCam()) with controller.initializeCamera().
• Replace bloc.add(const PauseDetection()) with controller.pause().
• Replace bloc.add(const ResumeDetection()) with controller.resume().
• Replace isBlocInitializeAbove with isControllerProvidedAbove.
• Remove flutter_bloc from your app's pubspec.yaml.

Added #

• Face verification — optional verificationProvider parameter enables "only enrolled faces unlock" mode.
• FaceVerificationProvider — abstract interface for pluggable verification backends (local, cloud, or custom).
• LocalFaceVerificationProvider — on-device verification using ML Kit contour-based feature extraction and cosine similarity matching.
• FaceGateCloudProvider — REST client implementing FaceVerificationProvider with configurable baseUrl, apiKey, timeout, retryCount, and basePath.
• FallbackVerificationProvider — wraps a primary and fallback provider for resilient verification (e.g. cloud-first with local fallback).
• FaceFeatureExtractor — normalized contour feature extraction with cosine similarity and feature averaging utilities.
• LivenessChecker — basic anti-spoofing via eye open probability and head euler angle checks.
• FaceTemplate / FaceVerificationResult / LivenessResult domain models with JSON serialization.
• FaceTemplateStore — abstract template storage with InMemoryTemplateStore implementation.
• FaceDetectionUnverified state — emitted when a face is detected but does not match any enrolled template.
• Frame throttling — configurable detectionInterval (default 300ms) skips redundant frames for battery efficiency.
• Debounced lock/unlock — lockDelay and unlockDelay prevent UI flicker from momentary detection gaps.
• Smooth transitions — AnimatedSwitcher with configurable transitionDuration between lock/unlock states.
• Haptic feedback — optional enableHapticFeedback on state transitions.
• Lifecycle awareness — automatically pauses detection when the app is backgrounded and resumes on foreground.
• Pause / Resume methods and FaceDetectionPaused state with previous-state restoration.
• Configurable camDirection, resolution, and minFaceSize parameters.
• Unit tests (15 test cases) and widget tests (17 test cases).
• GitHub Actions CI pipeline (flutter analyze + flutter test).
• Example app with three demo modes: Basic, Verification, and Advanced.
• Comprehensive dartdoc comments on all public APIs.
• MIGRATION.md — migration guide from v0.0.1.

Fixed #

• Memory leak — internal CameraController is now properly disposed.
• Image stream never stopped on dispose.
• Swapped noFace / noCameraOnDevice widget mapping.
• Removed dead Completer code.

Changed #

• Migrated from deprecated google_ml_vision to google_mlkit_face_detection.
• Upgraded to Dart 3 / Flutter 3.x SDK constraints.
• Added FaceDetectionPermissionDenied and FaceDetectionInitializationFailed error states.
• Added try-catch around camera init and face detection.
• Added configurable camera initialization timeout (initTimeout).
• Added explicit camera permission handling.

Dependencies #

• google_mlkit_face_detection: ^0.12.0 (replaces google_ml_vision)
• http: ^1.2.0

• Initial release.
• Widget locks screen when no face is detected.
• Face snapshot accessible via callback.