Exine

Universal Markdown extraction engine for Rust.

37+ formats. Zero external dependencies. 10–96× faster than Pandoc.

Performance

HTML (133 KB): 6.3ms. DOCX (37 KB): 25ms. Text: near-instant.

Supported Formats

PDF, DOCX, PPTX, XLSX, ODT/ODS/ODP, EPUB, RTF, SVG, HTML, EML, MSG, plain text — plus URL fetching and Vision AI escalation for images (Gemini, Claude, OpenAI, Mistral).

Installation

CLI

cargo install exine

Library

[dependencies]
exine = "0.1"

Usage

Library

use exine::extract::extract_by_extension;

let bytes = std::fs::read("report.pdf").unwrap();
let markdown = extract_by_extension("pdf", &bytes).unwrap();

CLI

exine report.pdf                    # Extract to stdout
exine report.pdf -o output.md       # Extract to file
exine https://example.com           # Fetch URL and extract
exine image.png --vision gemini     # Vision AI for images

Findability Shield

Protects content from AI scrapers while allowing search engines.

# Generate robots.txt
exine shield --robots > robots.txt

# Deploy to CDN (S3-compatible) with content
exine shield \
  --robots \
  --s3-bucket my-bucket \
  --s3-region us-east-1 \
  --content-dir ./output

Web Scraping

Stealth scraping with CAPTCHA solving and pagination.

# Crawl with stealth and pagination
exine crawl "https://example.com" --stealth --depth 2

# Crawl with CAPTCHA escalation
exine crawl "https://site.com" --captcha --render

Vision AI (Optional)

For scanned PDFs and images, Exine escalates to Vision AI:

export GEMINI_API_KEY=...
exine scanned.pdf --vision auto     # Auto-selects best available provider

Supported providers: gemini, claude, openai, mistral

Feature Flags

Contributing

See CONTRIBUTING.md for guidelines.

Built by NMA

Exine powers the FIELD ecosystem (GRID + SCALAR + STRIA) — AI-native tools for European and Israeli startup fundraising. https://nma.vc

License

MIT