Background work

Queued and running

A job can wait while another transcription, generation or export uses the worker or a required service. Queued does not by itself mean failure. Check the active work and let it finish before starting a duplicate.

Closing the browser leaves managed background work running. Use Activity & logs to inspect it, or reopen the project through Manager.

If work fails

Read the error and check the service, model, source and provider connection. Retry after resolving the cause. Keep a usable earlier artifact or take until the replacement has been reviewed. Updating or repairing a service while it is generating can interrupt that job.

MCP monitoring

pandrator_get_work reads a returned work handle and its status. Use the durable ID to follow the same job after reconnecting. A successful job means the operation completed; listen to or inspect its output before treating it as a finished creative result.

Before starting again

Check the current operation and its returned work ID. If the client timed out, the app may still be processing the request. Reconnect and inspect that work before submitting another generation or export.

A model can appear in the catalogue before its package and required runtime files are ready. Check the selected model, service readiness and compatible voice as separate facts. For an API route, check the configured deployment and account access. Keep logs for the specific failed operation when reporting a problem.