Colossyan callback vs polling, and Sume terminal-only webhooks

Colossyan recommends a callback in production: it POSTs to your callback URL with a status of finished or failed, or you poll the job, and its example loop retries every 5 seconds. Sume's webhook is terminal-only (job.completed, job.failed, job.canceled), so use it and keep polling status_url as a backup.

Colossyan facts are from its Receiving a generated video page; Sume facts from the OpenAPI schema and Jobs and results, both read 2026-10-01.

What does the Colossyan callback carry?

On success the service issues a POST to the callback URL. The body includes the public url of the video file, a sharing link, a status that is finished or failed, and whatever you passed as callbackPayload and dynamicVariables. The polling alternative loops until the status is finished or failed.

How does Sume deliver the outcome?

Send webhook_url (alias callback_url) and, without a mode, you get mode: "webhook": a 202 with the job envelope and polling URLs while the callback is stored. The OpenAPI text says delivery is terminal-only, with no progress or partial callbacks, and tells you to keep status_url polling available as a backup.

What poll cadence is safe on Sume?

The submit response and the job envelope include next_poll_after_seconds; the OpenAPI schema says to honor it when a wait times out. Use that value rather than copying a fixed 5-second loop from another vendor.

If you hit the 30-second wait cap, the response is still 2xx with the current job state. Poll with the same job id and do not submit a second paid job for the same intent.

Should I use the webhook or polling?

Use both. The webhook tells you quickly, and polling covers a missed delivery. For retry behavior on the receiver side, see Sume webhook retry schedule.