The API supports MP3 streaming outputs, but I'd like to stream data in as well. In my application I have a text stream that I'd like the AI to speak in real time. Currently I have to collect all the text into a single string and send it one-shot, then I can stream the output, but I'd like to open the stream as soon as I have even a little bit of text and have the AI read it in real time as the text becomes available.