turbo-parsepdf-wasm

Fast native PDF text / table / image extraction in the browser, via WebAssembly (wasm-bindgen) over the pure-Rust turbo-parsepdf core. Output as a JS object, or HTML / Markdown / JSON strings.

npm install turbo-parsepdf-wasm

Benchmark

Same engine as the native Node binding — ~9.7× faster than pdf.js on a 100-page document (5.6 ms vs 54 ms, best-of-N), text byte-identical to PyMuPDF. Browser numbers track native within WebAssembly overhead. Full tables + harness: the workspace benches/.

import init, { parse, parseToMarkdown, parseToHtml, parseToJson } from "turbo-parsepdf-wasm";

await init(); // load the .wasm

const bytes = new Uint8Array(await file.arrayBuffer());
const doc = parse(bytes);            // { version, pages: [...] }
const md = parseToMarkdown(bytes);   // string

// Encrypted PDFs: pass the user or owner password.
parse(bytes, "secret");

A fatal parse fault rejects with a "<Code>: <message>" string. Same engine and feature set as the Node and Python bindings (cross-reference streams, all stream filters, font/Unicode decoding, tables, images, RC4/AES decryption); scanned pages are flagged needs_ocr.

Part of the turbo-parsepdf workspace. MIT.