This guide is the shortest path for a new contributor to run DeepEye locally and validate that the development environment works.
- Docker and Docker Compose
uv- Node.js 20 and npm, only when running frontend checks outside Docker
- An LLM API key and model name
cp env.example .envEdit .env and set at least:
COMPOSE_PROJECT_NAMEHOST_GATEWAY_PORTLLM_API_KEYLLM_BASE_URLLLM_MODELJWT_SECRET_KEYPOSTGRES_PASSWORDMINIO_ACCESS_KEYMINIO_SECRET_KEYDOCKER_CONTROL_API_KEY
Use unique values on shared machines to avoid container, volume, and port conflicts.
make compose-configThis checks that Docker Compose can parse the stack and environment.
docker compose up --buildOpen:
http://localhost:8080
If you changed HOST_GATEWAY_PORT, use that port instead.
In another terminal:
docker compose psThe important services are:
gatewayfrontendbackend-apibackend-workerruntime-controlpostgresredisminio
For a deeper smoke test after the stack is running:
make compose-smokeUse the UI to create a session, attach a small CSV, and ask for a simple report or summary. For reference data and scenario notes, see:
Fast backend/core check:
make check-pythonFull local CI equivalent:
make checkInstall dependencies before checking:
make check-installSet a different HOST_GATEWAY_PORT in .env, then restart the stack.
Confirm your user can run:
docker psIf not, fix local Docker permissions before running DeepEye.
Check LLM_API_KEY, LLM_BASE_URL, LLM_MODEL, and provider network access.
For local debugging only, you can temporarily set:
STARTUP_WARMUP_STRICT=false
This removes local containers and volumes for the current Compose project:
docker compose down -vOnly use this when you intend to delete local development data.