Anchor-based placement

Find text anchors like assinatura, CPF, or CNPJ and turn them into clamped PDF signature rectangles.

Use findPdfTextAnchors when the business document already names the signing area. It replaces app-side pdfjs-dist scans and transform math with a package-owned matcher over SignatureKit text boxes.

anchor-placement.ts

import { findPdfTextAnchors } from "@signature-kit/pdf/anchors"
import { type PdfSignaturePage, type PdfTextBox } from "@signature-kit/pdf/config"
import {
  liteParseWorkerBrowserLayer,
  parsePdfTextBoxesBrowser,
} from "@signature-kit/pdf/liteparse-browser"
import { prepareAndSignPdf } from "@signature-kit/pdf/workflow"
import { Effect, Layer } from "effect"

const pages: Array<PdfSignaturePage> = template.documents[0]?.pages ?? []
const extractedTextBoxes = await Effect.runPromise(
  parsePdfTextBoxesBrowser(pdf, pages.length).pipe(Effect.provide(liteParseWorkerBrowserLayer)),
)
const pageTextBoxes: Array<Array<PdfTextBox>> = extractedTextBoxes.map((page) =>
  Array.from(page),
)

const stampRects = await Effect.runPromise(
  findPdfTextAnchors({
    pages,
    textBoxes: pageTextBoxes,
    matchers: [
      { text: "assinatura:" },
      { digits: "12345678901" },      // CPF/CNPJ-style numeric-only match
    ],
    stampSize: { width: 180, height: 54 },
    placement: "below",
    offset: 6,
  }),
)

const signedPdf = await Effect.runPromise(
  prepareAndSignPdf({
    pdf,
    pages,
    pageTextBoxes,
    stampRects,
    badge: {
      header: { text: "ASSINADO DIGITALMENTE" },
      rows: [[{ label: "Signatário", value: "Maria Souza" }]],
      footer: [{ text: "ICP-Brasil" }],
    },
    rubric: { initials: "MS" },
    signing: { policy: "pades-icp-brasil", name: "Maria Souza" },
  }).pipe(Effect.provide(Layer.merge(layer, liteParseWorkerBrowserLayer))),
)

digits normalizes both the matcher and the PDF text to numeric-only strings, so 123.456.789-01 and 12345678901 match. Returned rectangles are clamped to the page bounds.

Browser text extraction comes from parsePdfTextBoxesBrowser and LiteParse WASM. If your runtime cannot load that WASM module, pass pages and textBoxes from your own extraction seam and keep placement in findPdfTextAnchors.