02 · Deploy & Run
Every supported way to run Eddytor, in production and locally.
Eddytor runs from prebuilt public images - no source, no build. Pick the path that matches your scale.
Choose a path
- Docker Compose (single host) - the whole stack on one machine. Best for evaluation and small deployments. The Quickstart installer uses this.
- Kubernetes with Helm - for scale and HA. The stateless engine scales out across replicas. Cloud guides for AKS, GKE / EKS, and local clusters.
Before production
- Requirements & sizing - what you need and how much.
- Bundled vs external Postgres & object store - the bundled datastores are evaluation-only; graduate to managed ones for anything you keep.
- Deploy the optional Web UI - Eddytor is API-first; the browser UI is opt-in.
- Upgrades & forward-only migrations - snapshot Postgres first.
- Production checklist - lock it down before exposing it.
The shape of every deployment
However you deploy, it's the same four parts -
server, engine, Postgres, object storage.
Compose runs one server + one engine over direct TCP; Kubernetes runs the
:k8s images where the server discovers engine replicas via DNS.