Debug a slow Sume job with GET /v1/jobs/:id/events
A slow Sume job is queued, running, or waiting on your webhook. The events timeline separates them: job.queued, job.started, terminal, webhook.delivery.
To find out why a Sume job is slow, read GET /v1/jobs/:id/events. The public timeline shows whether the time went to waiting in the queue, running, or delivering your webhook. Read it for the one job in question; it is a pull snapshot, not a stream.
Events and what they mean
The docs list the public events. Provider task ids and raw provider URLs are never exposed in them.
| Event | What it tells you |
|---|---|
| job.created | The request was accepted and a job exists |
| job.queued | Waiting for a concurrency slot; normal |
| job.started | A worker moved it to processing |
| generation.submitted | Work was handed to generation |
| job.completed, job.failed, job.canceled | The terminal outcome |
| webhook.delivery | A delivery attempt to your URL |
Reading the gaps
A long gap from created to started means queue pressure: concurrency is plan-based, and more jobs than slots wait as queued by design. A long gap from started to terminal is generation time. A terminal event followed by repeated webhook.delivery entries means your endpoint is slow or failing, since each attempt times out at 10 seconds.
What to do next
For queue pressure, check generation_limits on submit responses and pace submits. For delivery trouble, check your handler and use redeliver. There is no SSE stream, so do not wait on this endpoint; poll status and read events when something looks off.
Sources
Related posts
More in Developers
- reference_video_urls or video_url? Reference footage vs edit source
On Sume, reference_video_urls guide a new clip; video_url is a source you edit or swap. They cannot be combined on Gemini Omni Flash. Which model takes which.
- Sora video ids in your database after the shutdown: what to keep
OpenAI's Videos API shut down 2026-09-24. Old video_ids no longer resolve, so store your own file URL and the model used. Schema fields included.
- Speech-to-text audio too large? Sume STT takes up to 10 MB, hosted
Sume's stt_create needs a public HTTPS audio URL on the Sume media host, 10 MB at most. What fits: 16 kHz mono wav versus mp3, and how to cut a long file.
- SSML in text to speech: Sume takes a plain transcript, no ssml field
Does Sume's text to speech accept SSML? The tts_create body has a plain transcript and rejects unknown keys. What to use for speed, volume, emotion and pauses.
Written by Sume