Implement screen reader detection service with reactive stream
epic-screen-reader-support-foundation-task-007 — Implement the ScreenReaderDetectionService that detects VoiceOver (iOS) and TalkBack (Android) activation state at app startup using MediaQuery.accessibleNavigation or AccessibilityFeatures. Expose a Stream<bool> isScreenReaderActive that re-evaluates on app foreground via WidgetsBindingObserver. Register as a Riverpod StreamProvider.
Acceptance Criteria
Technical Requirements
Execution Context
Tier 1 - 540 tasks
Can start after Tier 0 completes
Implementation Notes
Use a StreamController
Note MediaQuery.accessibleNavigation requires a BuildContext and cannot be used outside of widget tree — prefer WidgetsBinding.instance.accessibilityFeatures for service-layer detection. Integrate with AccessibilitySettingsRepository to read the screenReaderOverride flag — if override is true, emit true unconditionally. Deduplicate stream emissions to avoid causing redundant UI rebuilds: compare new value with last emitted before adding to the controller.
Testing Requirements
Write unit tests using flutter_test: (1) stream emits false by default (no screen reader active in test environment), (2) manual override forces stream to emit true, (3) didChangeAccessibilityFeatures triggers re-evaluation, (4) didChangeAppLifecycleState(resumed) triggers re-evaluation, (5) dispose removes observer and closes stream without error, (6) stream does not emit duplicate consecutive equal values. Use TestWidgetsFlutterBinding and mockMediaQuery to simulate accessibility state. Integration test on a device/simulator should verify real VoiceOver/TalkBack detection — tag as @Tags(['integration']). Minimum 80% branch coverage on the service.
Flutter's SemanticsService behaves differently between iOS (VoiceOver) and Android (TalkBack) in edge cases — e.g., announcement queuing, focus-gain timing, and attribute support. If the facade does not correctly abstract these differences, announcements may be silent or misfired on one platform, causing regression on the other platform to go unnoticed until device testing.
Mitigation & Contingency
Mitigation: Write platform-divergence unit tests early using SemanticsServiceFacade mocks. Validate announcement delivery on a physical iPhone (VoiceOver) and Android device (TalkBack) at the end of each sprint. Document known platform differences in the facade's inline API comments.
Contingency: If a platform difference cannot be abstracted cleanly, expose a platform-specific override path in the facade and implement targeted workarounds per platform, accepting the added complexity in exchange for correct behaviour.
Accessibility preferences stored in local storage may need new fields as higher-tier epics are implemented (e.g., announcement verbosity, sensitive-field guard toggle). Schema changes to an already-persisted store risk data migration failures or silent defaults on existing installs, breaking user preferences.
Mitigation & Contingency
Mitigation: Design the AccessibilitySettingsRepository with a versioned JSON schema from the start, using merge-with-defaults on read so new fields fall back gracefully. Define the full expected field list upfront based on all downstream epic requirements before writing the first record.
Contingency: If migration fails on a live install, fall back to full reset-to-defaults with a one-time in-app notification informing the user that accessibility preferences have been reset and inviting them to reconfigure.