This checks if a customer has finished speaking using a VAPI model. It's great for figuring out if someone is pausing to think, rather than actually done talking. Once we detect they're done, the request goes to

assistant.model.