Yeah that's the LLM not vapi, fine tune your model on your function calls and it'll dramatically increase it's accuracy at callingthem