Import document

Server-side pull: fetch a PDF from a caller-supplied URL (e.g. a presigned object-store GET) or an operator-registered storage connection into CloudPDF-owned storage, verify it, and commit it.

Default mode is synchronous and bounded: the response returns only after the transfer verified and committed (or failed). mode=async (connection sources only) answers 202 immediately and an in-process worker performs the transfer with leased, fenced retries; poll the document until ready/failed. The deployment import policy gates scheme, network range, and size; sources must declare a length. CloudPDF copies and owns the bytes — the source is never referenced in place. A 502 marks a retryable upstream failure: retry with the same idempotencyKey to resume the same document. URL sources are capabilities and never echoed back. Connection sources name operator-registered storage (bucket/prefix scope, allowed credential classes, and tenant bindings are deployment configuration); `revision` is provider-interpreted (S3 VersionId, GCS generation, Azure version id).

{
  "source": {
    "kind": "url",
    "url": "url"
  }
}

import { CloudPDFClient } from "@cloudpdf/sdk";

const client = new CloudPDFClient({
  baseUrl: "https://yourhost.com/path/to/api",
  token: "<token>",
});

await client.documents.importFrom({
  tenantId: "tenantId",
  source: {
    kind: "url",
    url: "url"
  }
});