date: 2026-07-14 title: "Local Ollama models; quieter background observation"
Local Ollama model support
- If Ollama is installed on your computer, the assistant can now use its local models directly — no configuration needed.
- Fixed picking Ollama during initial setup but silently getting a cloud model instead — choosing a local model now really means local.
Quieter background observation
- When the assistant checks the screen in the background to prepare proactive suggestions, it no longer flashes a window highlight — the highlight only appears when it is actually operating your computer for you.
Other fixes
- Fixed deleted conversations reappearing after a while.
- Fixed the "raw data" toggle in action details not responding.
- Fixed some system notifications showing incomplete content.