speech-to-speech

Build local voice agents with open-source models