Is it possible to interrupt the model when it’s talking? I feel that’s an important part of conversation. Especially when you’re talking to an LLM, that might go off on a tangent.