Move your existing AI inference workload to GKE.
Recent revisions are kept. Restoring one creates a new revision.
