AtomicBot-ai/Atomic-Chat v2.0.44 on GitHub

🚀 New Features

Generate images with stable-diffusion.cpp using Qwen 2.1, Krea 2 Turbo, FLUX and other models
Connect to your local models from your phone, laptop or any other device over LAN or a free Cloudflare tunnel using a link or QR code
Launch Z.ai's official ZCode harness in one click with Atomic Chat integrations
Choose reasoning effort for cloud models

Gemini now uses the correct reasoning settings, and models that require thinking default to Low
Update notices no longer cover the setup screen or active downloads
Setup shows which models fit your memory and puts them first
You get a warning before downloading a model that exceeds your memory, and downloads check disk space before starting
Setup stays open until your model finishes downloading and importing, then opens the chat
Active downloads show up in the model picker with a Cancel button
The model picker is more compact and keeps long provider names on one line
Long code blocks wrap, and code and reasoning stay stable while responses stream
Changing reasoning effort no longer changes earlier messages
Image attachments prompt you to download or select a vision model when needed
Agent approvals sit next to the message box, and enabling Full access asks for confirmation
Model loading shows progress and lets you cancel, while backend setup no longer hangs indefinitely
Atomic Chat checks for an installed model engine before downloading one
Image generation uses less memory, shows clearer progress and handles reference images better
Connectors have clearer names and status indicators, and plugins keep their icons
Project files show upload and indexing progress, with a Retry button when a file list fails to load
Changes to skill files take effect without a version bump
This release also includes more than 50 stability and UI fixes across sidebar layouts, dialogs and model loading feedback

Thanks to @s-chudmunge, @Vect0rM and @danyurkin for their contributions to this release!