Skip to contentATP Token
POST/v1/files

/v1/files

Upload a file to the Gateway and reference it in subsequent chat or messages requests by the returned id (format an_<ULID>).

Upload a file#

curl https://api.atptoken.ai/v1/files \
  -H "Authorization: Bearer atp-..." \
  -F "file=@./input.pdf"
# → 201 { "id": "an_01H...", "object": "file", "bytes": 152340, "filename": "input.pdf", ... }
The file ID field is `id`

The response follows OpenAI Files: the file ID is in id. Uploading identical content returns the same ID, with HTTP 200 instead of 201 (SHA-256 deduplication).

Endpoints#

POST /v1/files

multipart/form-data upload with a file part. Max 20 MB. Identical bytes are deduped by SHA-256.

GET /v1/files/:id

Returns 302 to a short-lived presigned URL on the object store. Follow the redirect to download.

Use an upload as a media reference URL#

A /v1/files id is not an asset:// reference, and the media endpoints cannot fetch it by id. Seedance video accepts asset:// URIs only from the separate asset API; see video generation. To feed an uploaded image to an image-edit or image-to-video model, read the Location header of GET /v1/files/{id} without following the redirect and pass that URL:

Upload → media reference URL
  1. Upload the file
    POST/v1/files201 with id (an_…)
  2. Resolve without following the redirect
    GET/v1/files/{id}read the 302 Location header
  3. Create the task right away
    presigned URLlives about 15 minutes
curl -sD - -o /dev/null https://api.atptoken.ai/v1/files/an_01H... \
  -H "Authorization: Bearer atp-..." | grep -i '^location:'
# → location: https://<object-store>/gateway-files/...?<presigned>

That presigned URL is downloadable without authentication — which is what the upstream provider needs — and lives about 15 minutes. Resolve it immediately before creating the task; do not cache it. See image generation and video generation.

Next steps

Previous/v1/models/{model}:generateContentNextOpenAI SSE
Was this page helpful?