Skip to main content
You need a LiveKit project (URL, API key, API secret) and three provider keys: Deepgram, Cerebras, Inworld. A free LiveKit Cloud project works.
1

Copy the env file

Fill in LIVEKIT_URL, LIVEKIT_API_KEY, LIVEKIT_API_SECRET, DEEPGRAM_API_KEY, CEREBRAS_API_KEY and INWORLD_API_KEY. Everything else has a working default in compose or in code.
2

Bring up the stack

That is Postgres, the backend, the voice worker and the console. LiveKit itself is external: the worker connects out to LIVEKIT_URL.The backend runs alembic upgrade head on startup, so there is no migration step.
3

Start a test call

Open http://localhost:5173, go to Agents, open the agent, choose Test call, then Start test call. Allow the microphone, then talk.
docker compose reads the root .env and nothing else. The agent/.env, backend/.env and frontend/.env files exist for running a service by hand. A value written into the wrong one produces a config that works one way and not the other.

​
What is published where

Both are bound to 127.0.0.1, so nothing is reachable from the network.

​
If nothing speaks

Do not guess. Every failure in this stack is quiet: a mismatched agent name mints a valid token, opens a real room, and logs no error anywhere.
It names the cause and the fix. See Troubleshooting for what it is checking and why.

​
Running the services by hand

Three terminals, one per service. Each has its own .env.example.