Generation workflow

Generation progress and recovery

Distinguish queueing, active generation and download without submitting duplicate jobs.

When to use

Behavior

What the dialog shows

Generate project submits a background job using your signed-in account. Queued means that the request has been accepted but a worker has not started it. Running means a worker is processing it.

When the backend reports progress, the dialog shows preparation, shared files, API contracts, service generation, project files, documentation, archive finalization and upload. It can also show the current service and completed-service count.

The count is not a percentage or an estimate of remaining time: services differ in size, and shared files and upload take time too. Elapsed time includes waiting in the queue. Keep the job ID when reporting a problem.

Reloading and interrupted connections

After the server returns a job ID, the Designer stores job metadata in sessionStorage for the current account and generation API. Reloading the same tab can resume status checks after authentication is restored; it does not submit the model again.

Recovery requires browser storage and a received job ID. It is not guaranteed after closing the tab, clearing storage, changing accounts, or losing the initial submission response. If storage is unavailable, keep the page open.

A delayed heartbeat or a temporary network error is not proof that generation failed. The dialog reports the interruption and retries status requests. Reloading or signing out stops browser polling, not the server-side job.

Waiting limits and results

One waiting session makes up to 120 status checks with five-second pauses. Network request time can extend that period. A browser waiting timeout does not cancel the server-side job; with saved metadata, reloading the same tab or choosing Generate project resumes the existing job.

On success, the Designer requests a short-lived download URL and initiates the ZIP download. A server-reported failure or expired job ends the wait. If fetching the download URL fails, retained job metadata allows another attempt while the job remains available.