pdfjs-serverless

A redistribution of Mozilla's PDF.js for serverless environments, like Deno Deploy and Cloudflare Workers with zero dependencies. The whole export is about 1.4 MB (minified).

PDF.js Compatibility

[!NOTE] This package is currently using PDF.js v4.6.82.

Installation

Run the following command to add pdfjs-serverless to your project.

# pnpm
pnpm add pdfjs-serverless

# npm
npm install pdfjs-serverless

# yarn
yarn add pdfjs-serverless

Usage

Since PDF.js v4, the library migrated to ESM. Which is great. However, it also uses a top-level await, which is not supported by Cloudflare workers yet. Therefore, we have to wrap all named exports in a function that resolves the PDF.js library:

declare function resolvePDFJS(): Promise<typeof PDFJS>

So, instead of importing the named exports directly:

// This will NOT work at the moment
import { getDocument } from 'pdfjs-serverless'

We have to use the resolvePDFJS function to get the named exports:

import { resolvePDFJS } from 'pdfjs-serverless'

const { getDocument } = await resolvePDFJS()

[!NOTE] Once Cloudflare workers support top-level await, we can remove this wrapper and re-export all PDF.js named exports directly again.

🦕 Deno

import { resolvePDFJS } from 'https://esm.sh/pdfjs-serverless'

// Initialize PDF.js
const { getDocument } = await resolvePDFJS()
const data = Deno.readFileSync('sample.pdf')
const doc = await getDocument({
  data,
  useSystemFonts: true,
}).promise

console.log(await doc.getMetadata())

// Iterate through each page and fetch the text content
for (let i = 1; i <= doc.numPages; i++) {
  const page = await doc.getPage(i)
  const textContent = await page.getTextContent()
  const contents = textContent.items.map(item => item.str).join(' ')
  console.log(contents)
}

🌩 Cloudflare Workers

import { resolvePDFJS } from 'pdfjs-serverless'

addEventListener('fetch', (event) => {
  event.respondWith(handleRequest(event.request))
})

async function handleRequest(request) {
  if (request.method !== 'POST')
    return new Response('Method Not Allowed', { status: 405 })

  // Get the PDF file from the POST request body as a buffer
  const data = await request.arrayBuffer()

  // Initialize PDF.js
  const { getDocument } = await resolvePDFJS()
  const doc = await getDocument({
    data,
    useSystemFonts: true,
  }).promise

  // Get metadata and initialize output object
  const metadata = await doc.getMetadata()
  const output = {
    metadata,
    pages: []
  }

  // Iterate through each page and fetch the text content
  for (let i = 1; i <= doc.numPages; i++) {
    const page = await doc.getPage(i)
    const textContent = await page.getTextContent()
    const contents = textContent.items.map(item => item.str).join(' ')

    // Add page content to output
    output.pages.push({
      pageNumber: i,
      content: contents
    })
  }

  // Return the results as JSON
  return new Response(JSON.stringify(output), {
    headers: { 'Content-Type': 'application/json' }
  })
}

How It Works

First, some string replacements of the PDF.js library is necessary, i.e. removing browser context references and checks like typeof window. Additionally, we enforce Node.js compatibility (might sound paradox at first, bear with me), i.e. mocking the canvas module and setting the isNodeJS flag to true.

PDF.js uses a worker to parse and work with PDF documents. This worker is a separate file that is loaded by the main library. For the serverless build, we need to inline the worker code into the main library.

To achieve the final nodeless build, unenv does the heavy lifting by converting Node.js specific code to be platform-agnostic. This ensures that Node.js built-in modules like fs are mocked.

See the rollup.config.ts file for more information.

Inspiration

• pdf.mjs, a nodeless build of PDF.js v2.

License

MIT License © 2023-PRESENT Johann Schopplich