Genie Voice Runtime

External voice runtime for the Genie ecosystem.

genie-voice-runtime owns the audio pipeline:

• wake word and push-to-talk activation
• voice activity detection
• capture and playback device handling
• speech-to-text
• text-to-speech
• acoustic echo control and denoise stages
• streaming voice session events

It does not own the agent.

genie-claw remains the agent layer: prompt policy, memory, tools, smart-home intent, safety gates, audit, and conversation state. The voice runtime turns audio into text and text back into audio; GenieClaw decides what the user meant and what actions are allowed.

Boundary

microphone / speaker
        |
        v
genie-voice-runtime
  wake, VAD, STT, TTS, audio streaming
        |
        | transcript / spoken reply events
        v
genie-claw
  agent policy, memory, tools, smart-home actions
        |
        v
genie-home-runtime / Home Assistant

The protocol types in this crate are the stable contract between GenieClaw and the voice runtime. The runtime implementation can evolve without forcing GenieClaw to own audio details.

Design Rules

• No agent prompts, memory policy, tool routing, or smart-home authorization in this repo.
• No Home Assistant or genie-home-runtime device logic in this repo.
• No LLM provider logic in this repo.
• Voice hardware should be portable across NVIDIA Jetson Orin 8GB, Raspberry Pi, other SBCs, Linux laptops, and development machines where possible.
• NVIDIA Jetson Orin 8GB remains the flagship tested deployment.

Evaluation Data

Public datasets can be used, but only for the layer they actually measure:

• genie-voice-runtime owns audio, wake/VAD, STT/TTS, transcript quality, and noisy utterance robustness.
• genie-claw owns BFCL tool-call scoring, family memory retrieval, smart-home intent routing, and deterministic device-state questions.

See docs/evaluation-data.md for the allowed data sources, license notes, and where each dataset belongs.

Status

Initial boundary scaffold. The current production voice code still lives in genie-claw and will move here in slices:

1. protocol and config contract
2. STT/TTS process wrappers
3. streaming session events
4. wake/VAD/audio-device handling
5. GenieClaw external runtime client
6. removal of the internal GenieClaw voice pipeline