You can handle this via JSON + prompt. In your JSON, include the end_call tool in the tools array. Then in your assistant’s prompt, tell the assistant when to use it, like: “If the user sounds finished or says goodbye, trigger the end_call tool.”