near mirage
I use openclaw with local llm qwen3.6 on lm studio
But there's a problem - every time openclaw sends it's request to lm studio first of all it asks for available models (I've got like 10+ downloaded), then it tries to load model i've set as a default (even if it's already loaded on the lm studio server), and then it sends it's prompt to it. Every time, on every tool call, on every request it tries to load another instance of a default model , 20-40 times per my message, it:
- asks for available models
- tries to load another instance
- sends it's request to the loaded instance (when loading a new one fails)
How do i set my config so it doesn't try to load another instance and just use the current loaded model?
I only need one llm instance loaded for all of my tasks
OpenClaw has to ask lm studio to load the model only if there's no model loaded on the server