🚀 New Features
- Qwen-Image-2.1-Turbo support — you can now create, reference, and edit images with Qwen-Image-2.1-Turbo (same 7B family as Qwen-Image-2.1) with a 1024px default and its distilled sigma schedule included.
🔧 Improvements & Fixes
- Core upgrade to atomic-chat-core 0.11.6→0.11.7 — fixes include a corrected sigma schedule closure so distilled runs finish without residual noise, img_gen now sends schedules required by Qwen-Image-2.1-Turbo, more resilient llama.cpp build selection, faster video decode on Apple Silicon (M1), and a WSL model-download fix.
- retrieve and docs retrieval improved — file_ids now match by id or file name (case-insensitive and path-end), the system prompt tells models when documents are attached, and retrieve falls back to searching all documents when no exact match is found.
- sends wait for web search activation — messages sent while Web Search is "Connecting" now wait up to 12s for the search tool to become available and warn if the message proceeds without search.
- context indicator token counting fixed — the header now correctly counts tokens when only assistant turns are present and avoids empty requests that caused server errors.
- toast dismissal bug fixed — overlapping "Starting Model" toasts no longer get stuck after a refused load.
- CLI removed from Settings and startup — the app no longer offers or auto-installs the Atomic Chat CLI from Settings, while existing installs are still refreshed after updates.
- app update UX improved — the Update button now shows download progress (and installation status), the downloads panel no longer lists the app-update row, and the installer flow displays accurate progress and messaging while preventing cancellation.
🙏 Contributors
Thanks to @Vect0rM for their contributions to this release!