Artifacts and Downloads

Last updated: May 11, 2026

Jobs produce artifacts. Artifacts are described through the API and downloaded through short-lived signed URLs.

Read artifact metadata

Use the job response or call:

GET /v1/artifacts/{id}

That returns artifact metadata plus a signed download_url.

Artifact metadata reads are rate limited to 240 requests per minute per account.

Download artifact bytes

Use the returned signed URL directly:

GET /v1/artifacts/{id}/download?expires=<unix>&sig=<hex>

Signed artifact downloads are rate limited to 600 requests per minute per IP address.

Why the flow is split

• Storage keys stay private.
• Download URLs can expire quickly.
• Delivery rules can change without changing the primary artifact metadata API.

Integration pattern

1. Poll the job.
2. Read the artifact IDs.
3. Request artifact metadata.
4. Download using the signed URL.
5. Cache the artifact on your side if your application needs longer-term access.

If the signed URL fails with invalid or expired download signature, fetch the artifact metadata again and use the new download_url. See Errors for more detail.

Retention and policy

Artifact retention may be shaped by:

• Plan policy
• Account-level quota overrides
• Retention sweeps

If your workflow depends on long-term storage, do not assume generated artifacts remain available indefinitely at the origin.