Skip to content

Release v0.1.0

Latest
Latest

Choose a tag to compare

View all tags
@mahimairaja mahimairaja released this 16 Jun 21:16
v0.1.0
d09e778

First public release of the LiveKit Voice AI Starter.

Most voice AI demos are a single script. This is the whole loop, structured the way you would actually ship it, and split into three pieces you can run, deploy, and swap independently.

What's inside

  • agent/: LiveKit voice worker. Deepgram nova-3 STT, OpenAI gpt-4.1-mini, Cartesia TTS, Silero VAD, and the multilingual turn detector. Web and SIP on the same worker, explicit dispatch.
  • backend/: FastAPI token server. Mints LiveKit room tokens at POST /api/v1/token using the standard LiveKit schema, with a clean API to service to repository layout and a JWT User slice to copy.
  • frontend/: React, Vite, and Tailwind on LiveKit's Agents UI. Audio visualizer, live transcript, and text chat.

Highlights

  • One command per service to run locally, one .env.example each.
  • Docker Compose brings up Postgres, backend, agent, and frontend together.
  • Web and telephony (SIP) on the same agent.
  • Swappable STT, LLM, and TTS providers; self-hosted or LiveKit Cloud with a one-line change.
  • Standard token endpoint, so LiveKit client SDKs connect with no glue.
  • Typed end to end, tested, linted, with CI and a pre-commit hook across all three packages.

Stack

STT, LLM, TTS: Deepgram nova-3, OpenAI gpt-4.1-mini, Cartesia (all swappable). Realtime: LiveKit Agents, WebRTC, Silero VAD, turn detector. Backend: FastAPI, async SQLModel and Postgres, PyJWT. Frontend: React 19, Vite, TypeScript, Tailwind v4.

Get started

git clone https://github.com/mahimairaja/livekit-starter
cd livekit-starter
cp .env.example .env   # add LIVEKIT_* + OPENAI/DEEPGRAM/CARTESIA
docker compose up --build

Then open http://localhost:5173 and click Start conversation. See the README for manual setup and per-package docs.

License

MIT.

Assets 2
Loading