You're right, llama-cpp-python OpenAI compatible endpoint works with `stream:true` and you can interrupt generation anytime by simply closing the connection. I

You're right, llama-cpp-python OpenAI compatible endpoint works with `stream:true` and you can interrupt generation anytime by simply closing the connection. I use this in a private fork of Chatbot-UI, and it just works.