[ajaydsouza]

The installed app now keeps the newest unread articles, with their stored images, on each device, so they can be read without a connection.

How it works

• GET /offline/articles lists what to keep: the newest unread articles that have a readable copy (Settings → Articles kept for offline reading, 30 by default, 0 for none). Each entry has the reader URL, a version that changes whenever the page would, and the stored images the article uses.
• public/sw.js syncs whenever a logged-in page opens online (at most every 30 seconds). It fetches new or changed reader pages, their stored images and the built CSS/JS they load, and deletes anything no longer listed. Archiving or deleting an article removes its copy at the next sync.
• Offline: a kept article opens as normal, with a notice that it's the saved copy. Any other page falls back to offline.html, which now lists the saved articles (via offline.js, a plain file so it runs under any CSP).
• Reading progress made offline is held in localStorage and sent when the connection is back. Online progress also clears anything queued for that article, so an older value never overwrites a newer one.
• Online behaviour is unchanged: navigations always go to the network first.

Logging out

Nothing stays on the device after logout:

• the logout form clears the copies before it submits;
• the logout response sends Clear-Site-Data: "cache", "storage";
• a sync that finds the session gone (401 or a redirect to login) deletes them.

Testing

• OfflineReadingTest: auth, selection and ordering, the limit and 0, image extraction, versioning, the setting on the page, the sync marker on logged-in pages only, the logout header, and that sw.js asks for the route the app serves.
• Full suite, Pint and PHPStan pass locally.
• End to end in Chromium against php artisan serve, with the browser offline and the server stopped (Playwright's offline mode doesn't reach service-worker requests):
  • all 3 unread articles synced and the archived one left out;
  • the article, its image, the CSS and JS all load offline, and the offline notice shows;
  • / and /archive fall back to the offline list;
  • progress queued offline reaches the database once back online;
  • archiving drops the copy at the next sync;
  • logging out leaves no article cache.

---

Generated by Claude Code