yeah I was actually just looking into whether vapi accepts responses streamed from the custom LLM, so you could potentially use a wrapper function for long-running functions to send the please wait first, but seems like you're way ahead of me