Anyone able to successfully use a fine tuned openAI model outside of the vapi UI ? When I receive a call the ai does it’s first message and then as soon as I respond it ends the call and says custom LLM pipeline error with no further info… but in the UI it works and it’s really annoying me because I need it with twilio outside of the UI