Deployments

Every time the AI makes a change you accept, OctopusLab produces a deployment. They're the immutable, addressable artefacts of your project — every state your site has ever been in is one click away.

Preview vs Production

Two environment kinds, identical otherwise:

  • Preview — every AI change. Each gets a unique URL (<slug>-<short>.octopus-lab.app). Use it to share work in progress; nothing routes traffic from the production subdomain.
  • Production — exactly one Preview is "live" at any moment for your custom domain / subdomain. Promote any Preview to Production from the Deployments tab — it's an alias swap, not a rebuild, so it's essentially instant.

Statuses

  1. queued — accepted, waiting for a builder slot. Almost never visible because we run plenty of concurrency.
  2. building — code is being compiled inside the sandbox; framework builds (Next.js, Astro, plain static) run here.
  3. ready — passed the build and is reachable at its URL.
  4. failed — build error. Click the deployment to see the captured build log; the AI in the chat panel can read it too and propose a patch.
  5. canceled — superseded by a newer build of the same source. Older builds are canceled automatically to save credits.

Instant rollback

The kebab menu on every row exposes Instant Rollback. It points the production alias at the older deployment in < 1 second — no rebuild. If a regression ships, rollback first and debug second.

Promote, Redeploy, Inspect

  • Promote — turn a Preview into the current Production. Disabled if the row is already Production.
  • Redeploy — rebuild the same source again. Useful if a transient dependency error broke a build.
  • Inspect — opens the deployment page with build log, file list, and the git-style commit short id.

API

See Endpoint reference for the public REST API to list deployments and trigger redeploys from your own scripts.