Voice & chat have some small differences, for example on voice I want to repeat things and confirm spelling whereas on chat I don't need to do that. So far I've been using separate (but 90% similar) prompts to accomplish that. I'm wondering if there's a way to accomplish with a single prompt. Is the LLM aware of its context (chat vs voice) so that I can tell it e.g. "if you're in chat mode then you do not need to confirm the spelling..." What would be the best way to do that? Thanks!

That happens because most LLMs don’t automatically know whether they’re in chat or voice mode unless you pass that info through metadata or context variables. I can help you set that up are you using Vapi, Retell, or a custom LLM integration for your agent?

thanks @Joshua and sorry for the delayed reply i didn't see this. i'm using vapi

Got it since you’re using Vapi, it’s totally possible to make one unified prompt. You just need to pass a mode variable (like "context": "voice" or "context": "chat") through the session metadata or webhook. The LLM isn’t aware by default.I’ve set this up before want to message me? I can show you the exact way to structure the prompt so it automatically adapts to chat or voice mode.