Experience report: working with realtime audio in Streamlit

Hi Streamlit Community!

I wanted to share an experience report I put together while using Streamlit for the first time.

My project was a demo of putting a voice interface (Hume’s EVI 2) in front of Anthropic’s Computer Use API. Anyway, Anthropic built their demo with Streamlit, so because I wanted to extend it, I found myself learning Streamlit too, and quickly became a fan!

I ran into some struggles with real-time audio, custom components, and overall trying to get the inherently “event-driven” of the voice AI to work well with Streamlit’s execution model, but ultimately I overcame them. I have some code examples in the post that might be useful for anybody else trying to do something similar. Anyway, give it a read if you’re interested!

1 Like