Extraction guide

Extract a dividend statement PDF to JSON

Everything below is the response the live API returned for a dividend statement, generated when this page was built. Not a description of what it would return — the output itself, including what it scored low on.

Who this is for

Portfolio trackers and tax-reporting tools reading broker statements that have no API behind them.

The job, in the words people search for: “extract dividend payments from a broker statement”.

What makes this shape awkward

Withholding tax is blank for domestic holdings and populated for foreign ones, in the middle of the table. A gap in a middle column is harder than a gap at the end: the cells to its right must not shift left to fill it.

The measured result

Pages read	1
Time	16 ms (16 ms/page)
Labelled fields found unprompted	4
Tables	1 (12 rows)
Input	dividend-statement.pdf (28 KB — synthetic, generated from a spec in the repo; real documents of this type cannot be published)

Fields, with a schema

Ask for the fields you want by the label printed on the page. Each answer carries the text exactly as printed (raw), the value coerced to the type you asked for, a confidence, and a bounding box you can draw on the page to check it.

curl -X POST https://api.pdfcraft.dev/v1/extract \
  -H "Authorization: Bearer $PDFCRAFT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "file": "<base64 of your dividend statement>",
    "schema": {
      "account": "string",
      "total_gross": "currency",
      "total_net": "currency"
    }
  }'

What came back

Field	raw	value	confidence
`account`	ACC-77-41290	`"ACC-77-41290"`	60%
`total_gross`	USD 4,118.62	`4118.62` USD	70%
`total_net`	USD 3,702.18	`3702.18` USD	70%

Every label it found without being asked

Send no schema and you get all of these, keyed by the label as printed. Useful for discovering what a new supplier’s layout actually contains before you write a schema against it.

Account, Period, Total Gross, Total Net

Tables

Pay date · Security · Shares · Per share · Gross · Withheld · Net · 12 rows × 7 columns · page 1

Table confidence 90%. Header confidence 100% — the first row was promoted out of the data.

Pay date	Security	Shares	Per share	Gross	Withheld	Net
2026-01-05	ACME	377	0.5421	204.38	—	204.38
2026-02-06	BRIT	878	1.4982	1,315.40	—	1,315.40
2026-03-07	CALD	65	1.3613	88.49	13.27	75.21

First 3 of 12 rows.

With options.rows_as_objects, the same row keyed by its header:

{
  "Pay date": "2026-01-05",
  "Security": "ACME",
  "Shares": "377",
  "Per share": "0.5421",
  "Gross": "204.38",
  "Withheld": "",
  "Net": "204.38"
}

What this does not do

No OCR. This reads the PDF’s text layer. A scanned or photographed dividend statement has no text layer and returns 422 extraction_failed with the page numbers that were blank — deliberately, and quickly, so you can route it somewhere that does OCR rather than waiting on a guess.
No model. Run the same file twice and you get the same bytes back. That is the trade: it cannot infer a field that is not printed, and it cannot hallucinate one either.
Ambiguity is reported, not resolved. A date like 03/04/2026 stays a string, and a bare $ returns a null currency, because guessing wrong on either is not something you can recover from downstream.

Try it on your own dividend statement

The playground takes a file and shows the same JSON, with every bounding box drawn over the page. No key and no signup for the first few; a free key gives you 100 pages a month.