my-agent-voice

Voice layer for my-agent — speak to your terminal agent

my-agent-voice is the JARVIS layer for my-agent, adding real-time spoken interaction to a terminal-first agent workflow. It streams audio over LiveKit so you can talk naturally while the agent is planning, coding, or running commands.

Key Features:

  • Live Streaming Pipeline: Low-latency audio transport with interrupt-friendly turn-taking for fluid back-and-forth conversations
  • Model-Agnostic Voice Runtime: Designed to work with configurable STT/TTS voice providers and swap models without changing core workflow logic
  • Deep my-agent Integration: Reuses my-agent memory, conversation state, and tool execution loop so voice and text stay in sync
  • Multilingual-Ready: Supports multilingual interaction paths through the configured speech pipeline, including mixed-language sessions
  • Latency Focus: Optimized for “speak -> think -> respond” cycles that feel interactive instead of batch-oriented