Maintain Workspace State

Tascarrel retains state at different scopes:

• Workspace configuration and host-owned repository caches remain when its VM is stopped.
• VM disks, images, seeds, harness profiles, editor profiles, and declared caches remain until the workspace is deleted.
• A pod’s writable files and chats persist until the pod is destroyed.
• USB attachments last for the current VM runtime. Prompt queues last until the agent engine restarts.

Replace a Pod Generation

Image and seed updates affect only new pods. Existing pods remain pinned to their original files, hooks, and agent inputs. Review and publish valuable changes before destroying an old pod.

One explicit operation can extend a retained pod: a repository added to the workspace configuration can be imported from the running pod’s Repositories tab. The import adds an absent checkout but never replaces existing pod files.

See Build Pod Images for image and seed actions.

Remove a Pod or Workspace

Destroying a pod permanently removes its private state. Deleting a workspace removes all of its pods and retained workspace state:

tascarrelctl workspace delete demo

The CLI requests confirmation. --force skips it but does not make the data recoverable.

Retain Shared Caches

Removing a [[caches]] declaration unmounts the cache from new pods without deleting its backing contents. Reusing the same cache name makes those contents visible again.

Tascarrel has no general state-migration framework. Back up repositories and configuration before upgrading, and follow release-specific compatibility instructions.