Deploying flowcat-server

flowcat-server runs one agent (defined in a YAML/JSON config) over HTTP — health probes, a Plivo media WebSocket + answer XML, and, with the webrtc feature, a browser playground at /. No control plane, no database.

Docker Compose

cp deploy/.env.example deploy/.env      # fill the key(s) your topology uses
docker compose -f deploy/docker-compose.yml up --build

• Health: curl localhost:6210/healthz → {"status":"ok"}
• Readiness: curl localhost:6210/readyz (200 once the topology's provider key is set)

For the browser playground (str0m WebRTC), build the webrtc feature:

FLOWCAT_FEATURES=webrtc docker compose -f deploy/docker-compose.yml up --build
# then open http://localhost:6210/  (set FLOWCAT_WEBRTC_BIND_IP to a reachable host IP)

Without Docker

cargo build --release -p flowcat-server --features webrtc   # or just "server"
GOOGLE_API_KEY=… ./target/release/flowcat-server --config deploy/agent.example.yaml

Configuration

• Agent config — see agent.example.yaml and the schema in flowcat-server/src/config.rs (agent graph, realtime/cascaded topology, transport, bind).
• Provider keys come from the environment as <PROVIDER>_API_KEY (e.g. GOOGLE_API_KEY, DEEPGRAM_API_KEY), the Gemini family also accepts GOOGLE_API_KEY. Keys are never read from the config file.

The build compiles the full provider connector set, so the build stage needs protobuf-compiler, cmake, and clang (already installed in the Dockerfile) for the gRPC + local-Whisper connectors. Trim --features to speed it up.