Extract form fields

Pull every AcroForm widget out of a filled-out PDF.

import * as mupdf from "mupdf";
import { getKeyValues } from "@nalinor/mupdf4llm";
import { readFileSync } from "node:fs";

const buf = readFileSync("application.pdf");
const doc = mupdf.PDFDocument.openDocument(
  new Uint8Array(buf),
  "application/pdf",
) as mupdf.PDFDocument;

try {
  const fields = getKeyValues(doc);
  const byName = Object.fromEntries(fields.map((f) => [f.name, f.value]));

  console.log("applicant_name:", byName.applicant_name);
  console.log("date_of_birth:", byName.date_of_birth);
  console.log("agreed_to_terms:", byName.agreed_to_terms === "Yes");
} finally {
  doc.destroy();
}

Filter by type / page

const textFields = fields.filter((f) => f.type === "text");
const onPageZero = fields.filter((f) => f.page === 0);

Combine with toMarkdown

If you want both the text and the form values in one shot:

import { toMarkdown, getKeyValues } from "@nalinor/mupdf4llm";

const md = await toMarkdown(buf);
const fields = getKeyValues(doc);

const fullMd =
  md +
  "\n\n## Form values\n\n" +
  fields.map((f) => `- **${f.label || f.name}**: ${f.value}`).join("\n");

A note on baking

toMarkdown internally calls doc.bake() so widget appearances are rendered into the page stream and AcroForm overlays become regular text in the markdown output. getKeyValues reads the widget objects directly, so it doesn't need baking.