From docs/releases/v2.2.16.md in the repository (generated, always the current text).
The local AI in one place, and AI plans that work.
#New
- Settings > Local AI: everything about the optional local AI model in one section: switch it on or off, choose the downloaded model or your own llama.cpp and GGUF file, download or remove it, start or stop it now, and how many idle minutes it stays loaded. Every AI button shows the same status line ("AI: on (Qwen3 8B) - not loaded right now, starts with your next request") with a link there, and offers to switch the AI on when it is off.
- Check your relations: Manage with AI can check every relation against all its aircraft and untick the aircraft where the command does something else.
#Fixed
- Manage with AI sent its first request with the AI switched off after a start, and showed "no offline rule" even when the model had answered. It now uses the model, shows its reply, and says plainly why the model could not start when that happens.
- Requests like "expand my relations to the new modules: ..." or "all modules in DCS" are understood, also without the model; an empty answer from the model falls back to JoyPro's own reading of the request.
- Settings: a section opens and closes from its arrow and the whole header row; the section chips at the top open or close their section (click) or show only that section (double-click).
- Settings > Components shows the local AI as ready (green) when it is set up and on.
- "F-14B" in a request includes the F-14 and its RIO seat.