dOCR
Guides

Extract a document

Extract structured data from PDFs, images, and DOCX files.

The POST /api/v1/extract endpoint accepts a single file and returns structured data. dOCR automatically picks the right pipeline for the file you send.

PDFs

Digital PDFs (with a text layer) are parsed directly — fast and exact.

curl https://app.docr.dev/api/v1/extract \
  -H "Authorization: Bearer $DOCR_API_KEY" \
  -F "file=@invoice.pdf" \
  -F "documentType=Invoice"

Scanned PDFs (no text layer) are detected automatically and read with a vision model — no extra parameters needed.

Images

JPG, PNG, BMP, and WebP images are read with a vision model. This is ideal for photos of receipts, IDs, and other documents.

curl https://app.docr.dev/api/v1/extract \
  -H "Authorization: Bearer $DOCR_API_KEY" \
  -F "file=@drivers-license.jpg" \
  -F "documentType=Driver's License"

DOCX

Word documents are converted to text and extracted.

curl https://app.docr.dev/api/v1/extract \
  -H "Authorization: Bearer $DOCR_API_KEY" \
  -F "file=@statement.docx" \
  -F "documentType=Bank Statement"

The response

Every extraction returns the same shape regardless of input format:

{
  "extraction": {
    "status": "completed",
    "documentTypeName": "Invoice",
    "confidence": 0.98,
    "pagesProcessed": 1,
    "outputJson": {
      "documentType": "Invoice",
      "fields": { "vendorName": "…", "total": 729.61 }
    }
  }
}

Limits: 10 MB and 15 pages per document. Larger files return 422.

On this page