🚀 New Features
- Cloud and self-hosted image & video models on the Images and Video pages — pick provider-hosted media models alongside local checkpoints, browse curated provider catalogs, and generate through the core so results land in the same gallery.
- Bundle the upstream llama.cpp release the pinned core would choose — the installer now picks the newest approved upstream build that the pinned atomic-chat-core supports, ensuring the bundled backend matches the core's expectations.
🔧 Improvements & Fixes
- Pin atomic-chat-core 0.15.0 so the app uses the released core that reads upstream llama.cpp releases.
- AppImage runs natively on Wayland (no XWayland) to avoid blurry scaling on Wayland compositors; the build now strips the GDK_BACKEND=x11 hook.
- Out-of-memory error details are now accessible: the engine's own diagnostic sentence and logs appear under Show details for clearer troubleshooting.
- 10+ engine, installation and UI fixes — better activation of installs, cleaner version lists and update flows, more robust engine-update dialogs, and various test and stability fixes around engine installs and switches.
⚠️ Breaking Changes
- None in this release.
🙏 Contributors
Thanks to @Vect0rM, @Ooooze for their contributions to this release!