Downloads

You can retrieve the original uploaded document or a generated result file.

Choose one of three response styles:

• Raw bytes for direct streaming
• Base64 JSON for JSON-only clients
• A temporary URL for browser downloads and larger files

Retrieve the original document

Request a temporary URL:

curl \
  "https://sandbox-api.shipwell.com/document-store/documents/<document-id>/presigned-url" \
  -H "Authorization: Token <user-token>"

import requests

response = requests.get(
    "https://sandbox-api.shipwell.com/document-store/documents/<document-id>/presigned-url",
    headers={"Authorization": "Token <user-token>"},
    timeout=30,
)
response.raise_for_status()
temporary_url = response.json()["download_url"]

Use GET /document-store/documents/{id}/file to stream the original bytes. Add encoding=base64 for a JSON response.

Base64 responses have an inline size limit. If the endpoint returns 413, use a temporary URL.

The temporary URL response includes download_url and expires_in. Treat the URL as sensitive and do not store it as a permanent document reference.

Retrieve a split result file

When split processing creates a generated PDF for a result, use the result identifier rather than the parent document identifier.

Use GET /document-store/results/{id}/file for raw bytes, add encoding=base64 for JSON, or use GET /document-store/results/{id}/presigned-url for a temporary URL.

Result file endpoints return 404 when the result does not have a generated file.

Find result identifiers

Retrieve every result for a completed run:

Use GET /document-store/documents/{id}/results?run_id={run_id}.

Use a row's id with the result file endpoints. output_file_ref indicates that the result has a generated file, but applications should use the authenticated download endpoints rather than that reference directly.

Download recommendations

• Store document_id or result_id , not temporary URLs.
• Follow the returned content_type instead of assuming every source is a PDF.
• Stream large responses rather than buffering them entirely in memory.
• Validate filenames before writing to a local filesystem.
• Request a fresh temporary URL after the previous URL expires.
• Do not expose temporary URLs in analytics, logs, or error messages.