The Rust crate is simply anydoc — the native form of the whole library. Every other binding (Node/Python/WASM) is compiled from this code, so in Rust you get the ceiling of performance.

Install

cargo add anydoc

The three core APIs

use anydoc::ConvertError;

fn main() -> Result<(), ConvertError> {
    // 1. From a path
    let markdown = anydoc::to_markdown("report.docx")?;
    println!("{markdown}");

    // 2. From bytes
    let bytes = std::fs::read("slides.pptx")?;
    let md = anydoc::to_markdown_bytes(&bytes, None)?;

    // 3. Stop at the document model (keep embedded assets)
    let document = anydoc::to_document(&bytes, None)?;
    // document keeps raw bytes and media types of images/objects

    Ok(())
}

Error handling

match anydoc::to_markdown("scanned.pdf") {
    Ok(md) => println!("{md}"),
    Err(ConvertError::Unsupported) => println!("Unsupported format or scanned PDF"),
    Err(ConvertError::Malformed) => println!("Corrupt file"),
    Err(e) => println!("Other error: {e:?}"),
}

The ConvertError enum has six variants: Unsupported, Malformed, Encrypted, ResourceLimit, MissingPart, Io — details in Error Handling & Limits.

Writing embedded assets to disk

let document = anydoc::to_document(&bytes, None)?;

// Walk the document's asset nodes (sketch)
for asset in document.assets() {
    // asset.bytes: raw bytes
    // asset.media_type: e.g. "image/png"
    std::fs::write(format!("assets/{}", asset.name), &asset.bytes)?;
}

In the Markdown output these assets render as alt text; the raw bytes live only on the document model.

Common scenarios

A tiny CLI tool

let args: Vec<String> = std::env::args().collect();
let md = anydoc::to_markdown(&args[1])?;
print!("{md}"); // pipeline-friendly, redirect with your shell

The Rust code behind WebAssembly

Compiling this crate to wasm produces the browser build @firecrawl/anydoc-wasm. Want to customize the wasm behavior (e.g. asset policy)? Compile your own from source.

Tips

• The second argument of to_markdown_bytes / to_document is the options slot — pass None for defaults
• The same code compiles to both native and wasm: Rust is the "write once, ship everywhere" source
• For server-side concurrency, pair with rayon or tokio::spawn_blocking as you like

Contributors: souvc