[beaniez]

yeah. I think logic to make llm inference calls could be moved to frontend to resolve that concern.