Resume after interruption

Before you begin

• The original local workspace and account/device context, if accessible.
• Any displayed Agent/run/native/lifecycle references and access to the affected host/document.

Follow the steps

1. Preserve the existing workspace

   Reopen the installed app and original project location. Reconnect an unavailable removable drive before relinking. Do not delete recovery data, reset history or reinstall to clear a hold.

   Checkpoint: The app can inspect the original local drafts, queues and journals.

2. Read unfinished-work status

   Inspect recovered work before sending anything new. Distinguish unsent drafts, queued items, active model requests, approvals, native operations and usage reservations.

   Checkpoint: Each unfinished item has its own identity and observed state; one finished model response does not clear a native unknown.

3. Restore identity and connection

   Sign in again if required. Reconnect the exact host and select the current document using fresh identity checks.

   Checkpoint: A fresh session/binding permits inspection; reconnect has not replayed the last command.

4. Continue from a proven checkpoint

   An unsent draft can be edited and sent deliberately. Undispatched queued work needs current admission. Refresh expired approvals. A lost model delivery needs its original request/status checked before a new inference.

   Checkpoint: The chosen resume action follows the actual recovered state instead of duplicating possible work.

5. Reconcile possible native effects

   If a native edit/save may have been dispatched, inspect its original operation and readback first. Import a known verified result rather than recreating it.

   Checkpoint: The original effect is proven or remains visibly unresolved with dependent writes held.

6. Choose a new continuation explicitly

   After the original state is known, review the current context, estimate and scope before continuing. Use the safe exit choices if you need to stop again.

   Checkpoint: No hidden paid continuation or duplicate native edit is caused by opening the app.

Check the result

• Drafts and known results remain intact, with each unfinished item resumed or held according to its own evidence.
• Offline history/status can remain readable while new account/provider admission is unavailable.

If something goes wrong

• Recovery cannot find the local drive or original journal.

  Restore the original location if possible and retain the missing reference. Ask Support before replacing or deleting the workspace.

  Resume from: Original workspace accessible or explicit unresolved recovery record.

• An update or migration was interrupted.

  Keep its backup and lifecycle journal. Follow the build-specific supported resume/compatible rollback path; do not overwrite an active runtime or force an older data format.

  Resume from: Known compatible update/migration checkpoint.

• Native state or provider usage is still unknown.

  Keep the affected operation/reservation held. Gather redacted references and request help; a timeout is not no-effect or zero usage.

  Resume from: Correlated terminal evidence for the original item.

Continue with a related guide

• Resolve an unknown native result
• Restore sign-in
• Recover an update
• Collect recovery evidence