loading locally running models with ollama is broken

[TobiasHT5]

In the settings, I tried to connect to my locally running llama.cpp server. I typed in the open-AI compatible url that llama.cpp was serving on and openworker confirmed it was running but whenever I went back to start a new session, there was no model loaded and I had to go through the process again and again.

[niutech]

Have you tried the OpenAI provider with the custom endpoint URL pointing to http://localhost:8080/v1?

[TobiasHT5]

@niutech @rohitprasad15 @andrewyng

Screen.Recording.2026-09-01.at.9.43.13.PM-1.mov