Exporting and syncing

Read every product with every asset we hold for it, one row per SKU. Ask for what changed since your last run, or freeze a selection into a file.

The export is the read a PIM or DAM sync wants. It is deliberately not a file list: relations are the payload, so your side never parses filenames to work out which product a render belongs to.

for row in pages:
    product = upsert_by_reference(row.product)
    for asset in row.assets:
        upsert_asset(product, asset.id, asset.url)

Read the catalogue

GET/v1/exports

Cursor-paginated products, newest first, each with every asset we hold for it.

{
  "product": {
    "id": "01HX5K2MZ7A3Q4FBNDC0EVDXY1",
    "reference": "TEE-SAGE-001",
    "name": "Crewneck Tee — Sage",
    "sex": "female",
    "deleted_at": null
  },
  "assets": [
    {
      "id": "01HX5K2MZ7A3Q4FBNDC0EVDXYS",
      "role": "source",
      "url": "https://…signed…",
      "url_expires_at": "2026-09-03T15:32:00Z",
      "mime_type": "image/jpeg",
      "width": 2000, "height": 2667,
      "ai_generated": false,
      "watermarked": false,
      "derived": false,
      "created_at": "2026-09-01T09:12:44Z",
      "task_id": null,
      "attributes": { "orientation": "front" },
      "also_features": [],
      "suggested_filename": "TEE-SAGE-001_front.jpg"
    },
    {
      "id": "01HX5K2MZ7A3Q4FBNDC0EVDXZ9",
      "role": "tryon",
      "url": "https://…signed…",
      "url_expires_at": "2026-09-03T15:32:00Z",
      "mime_type": "image/webp",
      "width": 1536, "height": 2048,
      "ai_generated": true,
      "watermarked": false,
      "derived": false,
      "created_at": "2026-09-02T16:40:03Z",
      "task_id": "01HX5K2MZ7A3Q4FBNDC0EVDXY2",
      "attributes": { "angle_shot": "full_body_relaxed_symmetric", "background": "park", "model_id": "01HX5K2MZ7A3Q4FBNDC0EVDXMD1" },
      "also_features": ["01HX5K2MZ7A3Q4FBNDC0EVDXPN"],
      "suggested_filename": "TEE-SAGE-001_tryon_park.webp"
    }
  ]
}

What an asset tells you

Field	Use it for
`role`	What the asset is: `source` (your photo), `beautifier`, `tryon`, `recoloring`. Edits keep the role and set `derived: true`. New roles may appear; treat unknown values as opaque.
`id`	An image id. One expired URL is re-fetched with `GET /v1/images/{id}`, never by re-running the export.
`url`, `url_expires_at`	A signed URL to the active version of the image, the one the dashboard shows. Valid for one hour.
`ai_generated`	`false` only for your own photos. For EU AI Act labelling.
`watermarked`	The bytes carry a watermark. This can change for a stable `id` when your plan changes.
`task_id`	The task that produced it, for `GET /v1/tasks/{id}`. Null for your photos and for generations made in the dashboard.
`attributes`	Flat, open facts per role: `orientation`, `angle_shot`, `background`, `model_id`, `edit_number`. Ignore keys you do not know.
`also_features`	Other products visible in this asset (tryon stylings).
`suggested_filename`	A filename, never a path, built from your organization’s export naming template with the real extension of the bytes.

Each asset appears exactly once, under the product it was made for. A tryon of a top styled with pants lives on the top; the pants are listed in also_features. Whether it also goes on the pants page is your merchandising call.

Filters

role=beautifier,tryon limits assets to those roles. Products are still listed when nothing matches; their assets is then empty.
limit (1 to 100, default 25) and cursor page through, exactly like GET /v1/products.

Only what changed

Add since=<timestamp> and the same endpoint returns only the products that changed at or after that instant: a field edited, a photo attached, a generation or edit completed, or the product deleted.

curl "https://api.veeton.com/v1/exports?since=2026-09-02T00:00:00Z" \
  -H "Authorization: Bearer $VEETON_KEY"

Each returned row is the product’s full current state, so upsert it whole; you never merge partial deltas. A deleted product appears with product.deleted_at set and an empty assets array. Remove it on your side.

Use the created_at of your previous run, minus a safety margin, as the next since. A nightly job is then a delta, not a re-export.

Freeze a selection into a file

POST/v1/exports

The same rows as a stable artifact: a JSONL manifest (one row per line) or a ZIP of the image files rendered by the same export pipeline the dashboard uses, with your organization’s naming template and per-workflow formats.

curl https://api.veeton.com/v1/exports \
  -H "Authorization: Bearer $VEETON_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{"format":"zip","product_ids":["01HX5K2MZ7A3Q4FBNDC0EVDXY1"],"role":["beautifier","tryon"]}'

Select up to 500 products with product_ids, or omit it for the whole catalogue visible to the API (same cap). Narrow with role.
format: manifest (default) is written immediately; the response is already completed with a signed url.
format: zip is rendered asynchronously. Poll GET /v1/exports/{id}, which carries Retry-After while processing, or subscribe to export.completed and export.failed.

{
  "id": "01HX5K2MZ7A3Q4FBNDC0EVDXEX",
  "format": "zip",
  "status": "processing",
  "url": null,
  "url_expires_at": null,
  "total_assets": 148,
  "error": null,
  "created_at": "2026-09-03T14:30:09Z",
  "finished_at": null
}

Asset URLs inside a manifest are valid one hour after it was written; the asset ids are permanent. The artifact URL on the job is re-signed on every read, so cache the file, not the URL.