Rattel

The reader lays out all 604 pages line by line from a bundled dataset in the Madani layout, rendered with the King Fahd Complex Hafs font. Pages turn right to left, bookmarks and the last page read persist, and the reader reports each page read to the khatma tracker through a callback wired in the router, so the reader never imports the khatma feature.

The home screen widgets are written natively: nine for Android in Kotlin on a shared base provider, and twelve for iOS in SwiftUI and WidgetKit. All of them read prayer times and Quran data the Flutter side shares with them.

The in-app browser blurs inappropriate images on the device. Three TensorFlow Lite models (an NSFW classifier, face detection and gender classification) run in a long-lived background isolate so scrolling stays smooth.

Group khatmas need a server, so I wrote one in TypeScript with Fastify and MongoDB. Claiming a juz is atomic: two people can’t take the same part, and the loser gets a 409. Khatmas expire after seven days through a TTL index. It has rate limiting, integration tests against an in-memory MongoDB, and a Docker image for Cloud Run.

Features never import each other. Shared models and state move down into common/; when that doesn’t fit, features are wired together in the composition root. A Dart script in CI rejects any import that points the wrong way. Twenty-eight test files include integration tests that drive the real app, one of which produces the store screenshots.