Install pdf-inspector - Five Ways to Parse PDFs

Pick the integration that fits your toolchain. Every channel below is an official release (crates.io / npm / PyPI) — safe to use.

Prefer not to install? Parse in your browser

No setup needed — drop a PDF in, see its type verdict and Markdown instantly. Everything runs locally.

Try It Online

Rust

Native + CLIInstall from crates.io

Node.js

npm · TS types includedInstall from npm

Python

PyPI · type stubs includedInstall from PyPI

Browser

WebAssemblyInstall from npm

CLI

pdf2md / detect-pdfSee CLI usage

Install commands per channel

Node.js

npm install @firecrawl/pdf-inspector
# or
bun add @firecrawl/pdf-inspector

Python

pip install pdf-inspector

Rust

cargo add pdf-inspector

Browser WASM

npm install @firecrawl/pdf-inspector-wasm

Optional: OCR runtime

The default build contains no OCR models, PDFium, or ONNX Runtime. You only need them when you explicitly enable selective OCR and pages actually get routed:

Set PDFIUM_LIB_PATH and ORT_DYLIB_PATH when the shared libraries are not on the platform search path.
The pinned PP-OCRv6 Small model set is downloaded and checksum-verified on the first routed page; offline environments can warm the cache or set modelDirectory with offline mode to prohibit network access.

Details in the official OCR runtime setup guide.

Before you install

Only use official releases (crates.io / npm / PyPI). Avoid unofficial mirrors or third-party downloads.
On Node.js servers, prefer the async variants (processPdfAsync, etc.) for heavy calls so the event loop stays free.
The browser WASM build is synchronous — call it from a Web Worker for large documents.

Hit a snag? Check these common questions.

How do I parse my first PDF?

The CLI is the most direct: pdf2md annual-report.pdf prints Markdown, --json returns structured output, --pages 1-3 limits pages, --compact enables token-saving output; use detect-pdf document.pdf --analyze --json to classify only. Full walkthrough in the getting started guide.

How do I use it in Node.js?

import { classifyPdf } from '@firecrawl/pdf-inspector'
const result = classifyPdf(pdfBuffer)
console.log(result.pdfType)         // "TextBased" | "Scanned" | "Mixed" | "ImageBased"
console.log(result.pagesNeedingOcr)

The complete API is in the API quick reference.

How do I use it in Python?

import pdf_inspector
result = pdf_inspector.process_pdf("document.pdf")
print(result.pdf_type)
print(result.markdown)

The complete API is in the API quick reference.

How do I use the WASM build in the browser?

import init, { processPdf } from '@firecrawl/pdf-inspector-wasm'
await init()
const result = processPdf(pdfBytes)
console.log(result.pdfType)
console.log(result.markdown)

The demo on this very site is built with this package. See try online.

What about scanned PDFs?

Classify first: classifyPdf tells you the document type and which pages need OCR. The web demo has no OCR; native packages offer selective OCR — only quality-rejected pages go to OCR while the rest stay on the fast local path.

How fast is parsing?

Official opendataloader-bench benchmark: the full 200-PDF corpus completes in 0.470 seconds with an overall score of 0.875 (#1). Text-based PDFs finish end-to-end within 200ms; classification alone takes ~10–50ms. See the benchmark.