Author here! I built voxlocal as an educational experiment to understand how the magic behind voice agents work. I wanted to have a raw CLI setup that does the bare minimum tasks needed for me to debug and understand the stuff behind the hoods. I have left many things out of the scope like VAD and async streaming as they might have complicated the journey.

Looking forward to learn more from the HackerNews community and their experiences about voice agents.