AI-assisted six-screen hero based on demo captures, not untouched screenshots. Originals and presentation notes.

Offline-first native Android study companion built around subjects, optional chapters, and lessons. Flashcards help retain lessons over time rather than acting as the primary organization system.

Download the latest public preview for Android 7.0 or newer. Preview APKs are debug-signed; export a full backup before updating and read the release's installation notes.

Recall's application implementation was developed 100% with AI using OpenAI Codex, under the project owner's direction and supervision. The project owner defines the product, requests and evaluates changes, and gives final approval. This credit describes Recall's own implementation, not the authorship of Android, Jetpack, fonts, or other third-party dependencies, which retain their own licenses.

AI-assisted development does not mean the app requires an AI service to run. Core studying works offline. AI card generation is an optional, external workflow: copy Recall's prompt to your preferred AI, then validate and import its JSON.

See the complete feature guide for each feature, how to use it, screenshots, and current limitations. See latest changes for the localization, calendar and reliability improvements included in this source version.

Lesson-based library	Subject resources
Focused review	Arabic & science
Study activity	Review calendar

These are AI-assisted feature panels based on real Preview 4 Android captures, not proposed UI designs or untouched screenshots. AI rendering can slightly alter small visual details; click a panel to inspect its unedited screenshot. All content and review history are synthetic demonstration data; no personal library is shown. Dates, counts and intervals are examples, not promises about your workload or retention. The running app uses your actual data. See the capture inventory and presentation provenance and prompts.

• Kotlin, Jetpack Compose, Material 3, Room, DataStore, and WorkManager.
• FSRS-6 review scheduling with rating previews and persisted review history.
• Manual cards, provider-independent AI JSON import, and science notation support.
• Subject resources for notes, PDFs, and photos; portable full-data backups.
• Quiet times, study windows, reminder pauses, and notification diagnostics.
• Review activity heatmap and Skip for now without changing card schedules.
• Review calendar in Today and Insights: current next due dates, daily counts and types, read-only card previews, and lesson links. Overdue cards remain under Today; suspended/archived content is excluded. Local midnight boundaries handle timezone/DST changes. No scheduling or database-schema change is required.
• Subtle short navigation/dock/calendar transitions, saved tab scroll positions, bounded calendar detail loading, and fresh due counts while the app is open.
• Offline English, Arabic, Spanish, French and German interfaces—not just layout direction—with localized settings, review controls, reminders, errors, plurals and dates. Study content and pronunciation languages stay independent of the interface.
• A localized guide explaining the study workflow and scientific foundations, shown once at startup and available again from Settings.
• Light/dark appearance and bidirectional educational text rendering.

Use Android Studio, JDK 17, and Android SDK 36. Configure your local SDK location in local.properties (not committed).

./gradlew assembleDebug
./gradlew assemblePreview
./gradlew testDebugUnitTest lintDebug
./gradlew connectedDebugAndroidTest

The debug APK is generated under app/build/outputs/apk/debug/. Instrumented tests require a connected emulator or device. The application name is defined in Android string resources.

For everyday testing, app/build/outputs/apk/preview/app-preview.apk is an optimized, resource-shrunk, non-debuggable build. It uses the local debug key to update existing preview installations without uninstalling or resetting data. It is not a store release: public/store releases require a managed release signing key. Keep the debug build for instrumentation and debugging.

See Phase 3 reliability, optional small sessions, measured performance and device checklist. New review dates are anchored coherently to rating submission; existing history is preserved, not reconstructed.

Learning Insights reports event-based observed recall, estimated memory maturity, and optional attention suggestions. It shows sample sizes, excludes unreliable legacy audit records, and distinguishes memory estimates from comprehension. Optional expandable current predicted recall reuses the FSRS-6 curve with explicit active-card coverage. Historical prediction and calibration remain deferred; no analytics migration or scheduling change is included.

Recall supports retrieval practice and spaced practice, with FSRS-6 memory-state equations. Predictions are estimates, not guarantees of learning outcomes. This implementation uses default FSRS weights rather than personalized parameter training; Again includes a product-level 10-minute relearning step.

• Retrieval practice research
• Distributed practice research
• FSRS algorithm documentation
• Recall scheduler notes

Core study data is local and no account is required. External AI providers and research links are optional and governed by their own policies. Android battery restrictions and force-stop can delay notifications. Export a full backup to preserve attached file contents; data-only JSON does not include them.

Build artifacts, machine-local configuration, validation captures, credentials, and personal backups are excluded from this repository. The chemistry JSON in recall-imports/ is an educational regression fixture used by instrumented tests. Third-party licenses and attribution are documented in THIRD_PARTY_NOTICES.md.

Contributions and reports in English or Arabic are welcome. Read the contribution guide, code of conduct and support guide. Use the issue forms for bugs/features or Discussions for questions. Report vulnerabilities privately according to SECURITY.md.

Recall is licensed under the MIT License. Third-party components retain their own licenses and attribution; see THIRD_PARTY_NOTICES.md.