Read PDF text and tables

POST /pdf reads the text and tables that are already inside a digital PDF. No model is involved, so the result is exact and repeatable. For scanned pages, which are images of text, use OCR instead.

Python
cURL

import base64, os, requests
from pathlib import Path

pdf = Path("statement.pdf")
body = {
    "attachments": [{"name": pdf.name, "data": base64.b64encode(pdf.read_bytes()).decode()}],
    "mode": "all",            # all | text | tables
    "table_format": "both",   # rows | records | both
    "pages": "all",
}
r = requests.post("https://dummydomain/pdf", json=body,
                  headers={"Authorization": "Bearer " + os.environ["API_KEY"]})
r.raise_for_status()
for f in r.json()["files"]:
    print(f["pages_processed"], "pages,", f["table_count"], "tables, truncated:", f["truncated"])

# Build the JSON body with the file base64-encoded (see the Python tab), then:
curl --fail-with-body "https://dummydomain/pdf" \
  -H "Authorization: Bearer $API_KEY" -H 'Content-Type: application/json' \
  --data-binary @pdf-request.json --output pdf-result.json

Request

Field	Values
`attachments`	List of `{"name", "data"}`, where `data` is the file as base64
`mode`	`all`, `text` or `tables`
`table_format`	`rows`, `records` or `both`
`pages`	A maximum number of pages to read, or `"all"`. It is not a page range.

The body is JSON, not a multipart upload.

Response

Each item in files[] has text, page_texts, tables, table_count, pages_processed and truncated. Check truncated before treating the result as the whole document.

Limits

Up to 20 files per request, 50 MiB per PDF and 200 pages per file. Base64 makes the request about a third larger than the file, and your server’s request size limit still applies.