Artifacts & storage

This page covers the files your image and video requests produce — where they are stored and how long they are kept. Text and embedding calls produce no files.

Where these files appear now

The separate Artifacts gallery has been consolidated into Workspace › Assets. Task outputs are now ordinary assets, kept until you delete them. Individual image and video results are opened from their entry under Serverless › Requests, and still follow the retention rules below.

The unit is the request, not the file

A request that generated 4 images produces 4 files sharing one request id. Retention and deletion always operate on the whole request’s output — there is no per-file delete.

Where results are stored

Each request resolves its destination by priority:

1. storage_credential_id on the API request — per-call override.
2. The deployment’s default bucket — set in the Serving Edit drawer.
3. Managed — Meshive-operated storage. The default.

• Your buckets connect under Settings → Integrations: AWS S3, Cloudflare R2, MinIO, GCS (S3-compatible).
• Verified credentials only — an unverified/broken credential fails explicitly (422 at submit, or storage_upload_failed), never a silent fallback elsewhere.
• Sync calls follow the same priority. With your bucket configured, sync results are always uploaded there (even b64_json responses, for retention).

Retention

What	Where	Kept for
Async results	Managed	The retention window — 10 days by default — then auto-deleted. Each result carries its own delete at time
Async results	Your bucket	Forever — Meshive never deletes from your buckets
Sync image results (response_format=url)	Managed	Link lasts ~1 hour, object removed shortly after — download immediately, use b64_json, or set an output bucket
Sync image results	Your bucket	Forever (links signed ~7 days; re-fetch to re-sign)

• The retention window is platform configuration and may change — each result’s delete at always shows the live value.
• Need something past the window? Download it, or point the deployment (or request) at your own bucket.

Finding a result

• Open the request under Serverless › Requests — its detail page renders the output and links to the original file.
• Links don’t go stale — URLs are re-signed each time you open the page. A tab left open for hours may 403; refresh it.
• Managed results also age out on their own at delete at; your buckets never auto-expire.

Result links elsewhere

Surface	Link lifetime
GET /v1/requests/{id} → output_urls	~1 hour (Requests)
GET /v1/videos/{id}/content	302-redirect to a ~1-hour link (Videos API)
Request detail page	~6 hours, re-signed every visit

Links expiring is normal. Permanent access = downloading the file, or owning the bucket it lives in.