Deploy
A Maize deployment consists of five Docker Compose stacks: Keycloak, the Model Repository backend (with MongoDB, the Elastic stack, Kafka and AKHQ), the Workflow Editor, Airflow and one or more workers. See Architecture for how they fit together.
Choose a path
| Path | Use it when | Start here |
|---|---|---|
| Maize MVP (single script) | Evaluation, demos, single-host installations. One .env, one script, everything on one machine. | Maize MVP |
| Manual per-component | Production or distributed installs: Keycloak you already run, Airflow on its own host, workers on many machines. | Manual setup |
Even with the MVP, the manual pages explain what each component does and what every variable controls.
Before you start
- Read Requirements & topology for host sizing and the ports each component needs.
- Decide which AI provider (if any) the AI features should use: AI providers.
After deploying
- Log in and run Settings → Initialize Resources for each user (details).
- Register your workers (Adding workers).
- Follow the Quickstart from step 4 to verify everything end to end.