This is completed and released! See docs here: https://dev.hume.ai/reference/empathic-voice-interface-evi/chats/get-audio

Having access to the raw audio data is 100% necessary for auditing the chat sessions since transformer STT models are susceptible to hallucination.

I know I can capture the audio on my own, but I'd rather not double the bandwidth usage for the user.

Glad to see that this is in progress.