- Your own model server works: Ollama, LM Studio and llama.cpp are in the service list (Settings → API keys), plain
http://is accepted on this phone and your own network (including Tailscale), and no API key is needed there. See Local models. - Saving your own server again reads its model list again, so models you've added show up
- Clearer message when an address isn't accepted
Full Changelog: v2.2.24...v2.2.25