Generate batch PDFs in Node.js

Use renderPdfBatch() when many records share a design: account notices, letters, invoices, or statements. The engine compiles the template once per request and returns one PDF in record order. Longer records can wrap onto additional pages.

This guide uses Node package 0.9.0, powered by Fullbleed 2.5.46. With Node.js 22 or newer, install the pinned version:

npm install --save-exact fullbleed@0.9.0

For a Python application, use the Python compiled-batch guide.

Generate account notices

Save this as notices.mjs, then run node notices.mjs. The first account has a short notice; the second has a longer message. All data is fictional.

import { writeFile } from 'node:fs/promises';
import { renderPdfBatch } from 'fullbleed';

const records = [
  { account: '1042', message: 'Your account is up to date. Thank you.' },
  {
    account: '1043',
    message: 'Your scheduled account review includes the following information. '.repeat(90),
  },
];

const result = await renderPdfBatch({
  html: `<!doctype html><html lang="en"><body>
    <p class="label">ACCOUNT NOTICE</p>
    <h1>Account {{account}}</h1>
    <p class="message">{{message}}</p>
    </body></html>`,
  css: `@page { size: A5; margin: 18mm }
    body { font-family: Inter; font-size: 11pt; color: #23342f }
    .label { color: #47745d; font-size: 8pt; letter-spacing: 1pt }
    h1 { font-size: 25pt; margin: 8pt 0 18pt }
    .message { line-height: 1.5 }`,
  records,
  maxPages: 20,
  timeoutMs: 30_000,
});

await writeFile('notices.pdf', result.pdf);
await writeFile('record-pages.json', JSON.stringify(result.records, null, 2));
for (const row of result.records) {
  const first = row.pageOffset + 1;
  const last = row.pageOffset + row.pageCount;
  console.log(`Account ${records[row.recordIndex].account}: pages ${first}–${last}`);
}

Open notices.pdf. The console and record-pages.json identify each account's page range: account 1042 occupies page 1, and account 1043 occupies pages 2–5. recordIndex and pageOffset are zero-based; the printed ranges above are one-based.

Records contain plain strings. Format numbers and dates in your application; characters such as <, > and & remain literal text. Every record must have the same own enumerable fields, exactly matching the template's text bindings. Changes to image sources, attributes, CSS, or document structure use the ordinary rendering API or a different template.

Choose how each record flows

The default layout: 'reflow' lays out each record with its values, so a long notice can take more pages than a short one. Each returned pageCount describes the actual result.

Use layout: 'fixed' when values fit a reviewed template's geometry. Fixed mode retains the compiled layout and selected fonts. Review the PDF with realistic data before using that mode for a delivery workflow.

Compilation is scoped to one call. A later renderPdfBatch() call compiles its template again; JavaScript does not expose a persistent compiled-template handle.

Use the reports in your application

The result contains pdf, pages, and records. Each record reports its recordIndex, pageOffset, pageCount, and missingGlyphs.

Missing characters reject the batch by default with MISSING_GLYPHS. With allowMissingGlyphs: true, a PDF is returned and each affected record identifies the character's codepoint, char, occurrence count, and fontsTried. That last field identifies the first selected registered font where the character was absent. The top-level missingGlyphs count is the number of distinct missing code points across the entire batch.

These diagnostics cover the selected registered font chain. They exclude unembedded Standard 14 fonts and text inside images, and do not establish correct shaping, line fit, accessibility, or PDF conformance. Keep visual checks alongside the reports when adapting a design.

For order events or scheduled statements, let your application's durable job system collect and format the records. Render the batch, inspect the result, then store its PDF and page map together. A failed call does not deliver a partial PDF.

Bound work across requests

Use one render queue per application process and call its renderPdfBatch() method with the same options. A whole batch occupies one queue slot. Queue deadlines include waiting time, and cancellation waits for the worker or child process to stop. The queue controls concurrent jobs; your application owns scheduling, retries, authentication, and delivery.

Up to 10,000 records are accepted. HTML, CSS, and all record keys and values share a 4,000,000-byte UTF-8 budget. maxPages applies to the combined PDF. Records are copied when the request is accepted, before asynchronous rendering begins.

The PDF, reports, and optional previews stay in memory. Split larger workloads into batches that fit your host's memory and CPU budget. Use timeoutMs and an AbortSignal to bound work; use the existing isolation: 'process' option when your Node host permits child processes and that isolation suits the application.

Other JavaScript authoring paths

Component-authored templates can use document.renderPdf(renderPdfBatch, { records }); component authoring remains alpha. The same batch method is available on the browser renderer, and fullbleed/cloudflare exports renderPdfBatch for Workers. Workers have a smaller 1,000,000-byte combined source budget. The package's Workers verification covers local workerd, so qualify deployment against your actual hosting limits.

Node.js quickstart · Batch API reference