Register a file by URL, or get a presigned upload URL

Adds a file to your reusable corpus, either by URL or by uploading the bytes yourself. Two request shapes, both pure JSON — send one or the other, never both. **By URL.** `{ url, name?, processing? }` registers the file and fetches it over http(s). The media type is sniffed from the response, so any `mimeType` you send is ignored. You get the file back with `status: "processing"`. ```json { "url": "https://www.irs.gov/pub/irs-pdf/fw9.pdf", "name": "fw9.pdf" } ``` **By upload.** `{ name, mimeType, processing? }` reserves the record and returns `uploadUrl` plus `uploadExpiresAt` (15 minutes). `PUT` the bytes to that URL with a `Content-Type` header exactly equal to the `mimeType` you sent; parsing starts the moment they land. ```json { "name": "recording.mp3", "mimeType": "audio/mpeg" } ``` **Parsing engine** — `processing` accepts `auto` (the default, self-upgrades to OCR on scans), `ocr`, `simple`, `transcribe` and `transcribe_diarize`. Audio and video default to `transcribe_diarize`. <Note> Files registered here are persistent — no TTL — and reusable across any number of runs as `file: { "id": "…" }`. That is what makes parse-once-run-many work: pass the same id to [extract](https://docs.cloudraker.com/api/cloud-raker-api/capabilities/extract), [redact](https://docs.cloudraker.com/api/cloud-raker-api/capabilities/redact) and [pipeline](https://docs.cloudraker.com/api/cloud-raker-api/capabilities/pipeline) without re-parsing. </Note> **Learn more:** [Working with files](https://docs.cloudraker.com/developers/files)