Upgrades

Dark Factory is pre-MVP: any release may change the configuration, the database schema or the web UI incompatibly, and there are no migration guarantees between releases yet. Read the release notes on the releases page before every upgrade.

Upgrade a release

1. Back up the database (Backup and restore).

2. Let active runs finish, or accept that they are cut off: a run whose sandbox disappears ends with sandbox_lost and its task is queued again automatically.

3. Upgrade with the new chart version and your values:

   helm upgrade dark-factory oci://ghcr.io/prskr/dark-factory/charts/dark-factory \
     --namespace dark-factory \
     --version 2026.11.1 \
     --values values.yaml

4. The migrate job runs first (pre-upgrade hook) and applies the new schema migrations; then the control plane and the gateway roll. Check:

   kubectl -n dark-factory get pods
   kubectl -n dark-factory rollout status deploy/dark-factory-control-plane

5. The profile menu shows the new version and links to its docs.

The control plane, runtime, gateway and plugin images follow the chart’s appVersion unless you pinned them; a pinned plugin image keeps its version until you change it.

Rolling back

Schema migrations only go forward: there are no down migrations. helm rollback restores the old images and values but not the old schema, and an older control plane may not start on a newer schema. To go back, restore the database backup taken before the upgrade, then roll back the release.

Release-specific steps

Some releases need a step beyond helm upgrade, for example removing the forge host from the egress rules after the switch to the Git proxy. They are listed in the release notes and in the chart README.