Maize MVP (single script)

The maize-mvp repository bundles all components and a setup.sh that deploys them on one host.

1. Configure

git clone https://github.com/datacrop/maize-mvp.git
cd maize-mvp
cp .env.example .env

Edit the root .env:

Variable	Set to
HOST_IP	The IP address or hostname browsers and workers use to reach this host. Required.
DOCKER_BIN_PATH	Path of the docker binary on the host (default /usr/bin/docker).
WME_SERVICE_TOKEN	Leave blank — generated on first run (shared backend ↔ worker secret).
CREDENTIALS_ENCRYPTION_KEY	Leave blank — generated on first run (AES-256-GCM key for stored secrets).

You only need component .env files for manual or distributed deployments; setup.sh writes them from the root .env.

2. Run

chmod +x setup.sh
./setup.sh

Choose:

1. Deploy everything — Keycloak, backend, editor, Airflow and a worker.
2. Everything except Keycloak — skip starting the bundled Keycloak.
3. Worker only — start the worker stack.

External Keycloak or a remote Airflow host

With a root .env present, setup.sh synchronizes component settings before the menu. It overwrites Keycloak URLs and the worker's AIRFLOW_IP with values derived from HOST_IP. Editing the component files beforehand does not preserve those external addresses. Use Manual setup for an existing external Keycloak and Adding workers for a worker on another host.

What the script does:

1. Checks docker and docker compose.
2. Validates HOST_IP (warns if it looks unset) and generates WME_SERVICE_TOKEN (openssl rand -hex 32) and CREDENTIALS_ENCRYPTION_KEY (openssl rand -base64 32) if missing, saving them in .env.
3. Creates any missing component .env from its example and synchronizes host addresses, Keycloak URLs, Airflow/Flower/Kafka/Kibana URLs and the shared secrets into them.
4. Creates the maize-wme-network Docker network and runtime folders.
5. Deploys, in order: Keycloak (waits for it to be healthy), backend (prepares config/extra-processors.json and a placeholder Logstash pipeline, pulls images), editor, Airflow, worker.
6. Prints the URLs.

3. Access

Service	URL	Login
Workflow Editor	http://HOST_IP:5173, or http://HOST_IP with UI_PORT=80	admin / admin (Keycloak)
Keycloak admin	http://HOST_IP:8180/admin	admin / admin
Backend API / Swagger	http://HOST_IP:9090/swagger-ui/index.html	Bearer token
Airflow	http://HOST_IP:8080	airflow / airflow
Kibana	http://HOST_IP:5601	elastic / elastic
AKHQ	http://HOST_IP:8081	—
Flower	http://HOST_IP:5555	—

Change the defaults

The MVP ships well-known passwords (Keycloak admin, realm user admin, Airflow, Elastic, MongoDB). Change them before exposing the host to any network you do not fully trust.

4. Initialize

Log in to the editor and run Settings → Initialize Resources once for each user (details). Then import the worker in Warehouse → Workers (Adding workers).

Optional: ship extra processors

Place an extra-processors.json in maize-model-repository/config/ before initialization to add your own processor definitions to every user's catalogue. Format: Model Repository → Predefined processor definitions.

Check and troubleshoot

# Run from the maize-mvp repository root.
docker compose -f Keycloak/docker-compose.yml ps
docker compose -f maize-model-repository/docker-compose.yml ps
docker compose -f maize-workflow-management-editor/docker-compose.yaml ps
docker compose -f maize-processing-engine-airflow/docker-compose.yaml ps
docker compose -f maize-processing-engine-worker/docker-compose.yaml ps

# Choose one component to inspect its logs.
cd maize-model-repository
docker compose logs -f

• Worker Runtime unavailable — backend and worker must share WME_SERVICE_TOKEN, port 8090 must be reachable and the agent needs /var/run/docker.sock.
• Port conflicts — the editor port (5173, or 80 with UI_PORT=80), 8080, 8180 and 9090 must be free.

Tear down

./teardown.sh            # stop and remove containers (same 3-option menu)
./teardown.sh --volumes  # also remove volumes

Components are stopped in reverse order and the network removed. The MongoDB volume (maize-model-repository_mongodb_data) is always kept so your catalogue survives.

Workflow Assistant

The MVP does not yet deploy the preview assistant-runtime service. It runs in the backend stack on the dev/ai-feature branch; see Model Repository → assistant-runtime.