v1.2.1
LatestFixes local assistants, which could fail to respond in 1.2, and adds new local models to choose from.
Highlights
Local assistants work reliably again
Local assistants could fail to respond in 1.2. This release fixes them: the first message no longer times out while the model is loading, new assistants warm up before you chat, and the first run greets you with a clearer, provider-aware message — with a "still working" hint when a model is slow to respond.
- Feature Add new local models to choose from when creating an assistant: Qwen3.5 9B and Ministral 3 8B and 14B Reasoning.
- Fix Fixed the first message to a local assistant timing out instead of responding.
- Fix Fixed the first-run experience for local assistants: a clearer, provider-aware greeting, a "still working" hint when a model is slow, and per-reply timing on hover.
- Fix Fixed a 404 when deleting an assistant and creating another in the same session.
- Other Updated the bundled inference engine (llama-server).