anydoc compiles to WebAssembly and runs in the browser — which makes it special: conversion happens entirely on the user's device, and files never leave it. The converter on this site's home page is exactly this approach.

Install

npm install @firecrawl/anydoc-wasm

Works in any frontend project — no backend required.

Basic usage

import init, { toMarkdownBytes, toDocument } from '@firecrawl/anydoc-wasm'

// 1. Initialize the wasm (once, at app startup)
await init()

// 2. Convert from a File object
const file = inputElement.files[0]
const bytes = new Uint8Array(await file.arrayBuffer())
const markdown = toMarkdownBytes(bytes)

// 3. Stop at the document model (keep embedded assets)
const document = toDocument(bytes)

A complete upload-and-convert example

<input type="file" id="doc" accept=".docx,.pptx,.xlsx,.pdf,.doc,.ppt,.xls,.odt,.ods,.odp,.rtf,.epub,.csv" />
<pre id="out"></pre>

import init, { toMarkdownBytes } from '@firecrawl/anydoc-wasm'

await init()

document.getElementById('doc').addEventListener('change', async (e) => {
  const file = e.target.files[0]
  if (!file) return
  const bytes = new Uint8Array(await file.arrayBuffer())
  try {
    const md = toMarkdownBytes(bytes)
    document.getElementById('out').textContent = md
  }
  catch (err) {
    document.getElementById('out').textContent = `Conversion failed: ${err.code || err}`
  }
})

The privacy pitch: why files never leave the device

• Conversion logic is Rust compiled to wasm, running inside the browser sandbox
• No upload request, no backend relay — after file.arrayBuffer(), the bytes live in memory
• Perfect for sensitive documents (contracts, medical records, financial data) that need in-browser conversion

Performance & size

• The wasm binary is a few MB (less gzipped); the first load costs a moment, then it's resident
• Conversion speed is close to native — same Rust core, running in the wasm sandbox
• This site's converter is built on it: drag a file in, Markdown out in milliseconds

Tips

• Call await init() once and reuse the instance — don't re-initialize per conversion
• Very large files (hundreds of MB) will consume memory proportional to the file; keep that in mind
• Scanned PDFs still return unsupported — there's no OCR capability in the browser build, and that's a deliberate boundary

Contributors: souvc