Headline
- Supports Ryzen AI SW 1.6 for hybrid NPU/GPU execution of Qwen3 models with 4k prompt size
- Use models downloaded from any source with Lemonade
- Lots of quality-of-life improvements
What's Changed
- Fix: Prevent port conflicts on macOS in run command (#394) by @HyunhoAhn in #449
- Update models page with recent changes and FLM by @jeremyfowers in #453
- Gracefully handle no FLM on linux by @jeremyfowers in #452
- Continue.dev documentation updates by @itomek in #454
- Prevent suggested=false models from being recommended by the model manager by @jeremyfowers in #459
- unblock lint and ci by @jeremyfowers in #467
- keep fixing lint by @jeremyfowers in #468
- Implement smart polling in the web ui by @jeremyfowers in #469
- Enhance version handling and model list by @ZaneNi in #461
- Don't show RAI/FLM models on unsupported systems by @jeremyfowers in #470
- Fix CI by avoiding operations that may require elevated permissions by @danielholanda in #474
- Update Docs to Include Dify, Copilot by @vgodsoe in #471
- Feat: Log Filter Extension by @siavashhub in #473
- Fix major testing issues in server.py by @jeremyfowers in #476
- Fix server spec bugs by @jeremyfowers in #479
- Add Ryzen AI 1.6 support by @ramkrishna2910 in #463
- Feature to upload local models to the UI by @meghsat in #456
- Fix thought process content display by @ramkrishna2910 in #486
- Improve CI reliability: Fix tests and prevent silent failures by @danielholanda in #489
- Rev version for 8.2.0 release by @jeremyfowers in #488
New Contributors
- @HyunhoAhn made their first contribution in #449
- @meghsat made their first contribution in #456
Full Changelog: v8.1.12...v8.2.0