Skip to content

Install

Install: overview

Four paths to the same four containers (web, orbit-api, nocturne, orbit-stack). All four obey CONTRACT.md: the same ports, the same environment variable names, the same /healthz//readyz, the same hardening. Pick one; don't mix two on the same host.

PathUse it forNeedsGuide
Local k3dKicking the tyres. A real, disposable Kubernetes cluster on your laptop, ~1.2 to 1.5 GiB of RAM.Docker, k3d, kubectl, helmlocal-k3d.md
Docker ComposeSingle host, no Kubernetes. The fastest way to a running stack, or a CI smoke test.Docker Engine + Compose v2docker-compose.md
HelmAny real cluster, staging or production.Kubernetes 1.30+, Helm 3.xhelm.md
Plain kubectlTeams that cannot or do not run Helm.Kubernetes 1.30+, kubectlkubectl.md

If you're not sure which one

  • You have Docker and nothing else → docker-compose.md.
  • You want to see the Kubernetes hardening (NetworkPolicy, PDB, topology spread) do something, on a laptop → local-k3d.md. (k3s enforces NetworkPolicy through its bundled kube-router controller, so the policies do something there too; see security.md.)
  • You're installing into a cluster you already run, and Helm is allowed → helm.md. This is the production path; it's the only one with a rollback story that survives an image tag being repointed.
  • You're installing into a cluster you already run, and Helm is not allowed → kubectl.md.

Before any of them

Every path needs credentials you generate yourself; nothing in this repository ships one. Read configuration.md → "Generating a static token" once, before you open any install page; each page tells you exactly where the token goes for that path, but none of them generate it for you.

After you're up

  • Confirm it is healthy, not only running: run the "Verify" section of the page you followed.
  • What to watch once it's running: operations.md.
  • First failure: read troubleshooting.md before you improvise. Its first three entries are failure modes already hit on these install paths.