mineru-rs 0.1.0

Native Python helpers for MinerU Rust

MinerU Rust

Rust client library, command-line tools, and local API server for parsing PDF, image, and Office documents with an external OpenAI-compatible MinerU VLM service. PDF rendering uses pure Rust; no model is downloaded or run locally.

Compatibility is pinned to the MinerU 3.4.4 vlm-http-client transport baseline. See the compatibility contract and usage guide. This is not a full MinerU compatibility claim.

Requires Rust 1.89 or newer.

Rust library

cargo add mineru

use mineru::{ClientConfig, MinerUClient, ParseOptions, PdfInput};

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let client = MinerUClient::new(ClientConfig::new(
        "https://example.test/v1",
        "model-id",
    )?)?;
    let document = client
        .parse_pdf(PdfInput::Path("input.pdf".into()), ParseOptions::default())
        .await?;
    println!("{} pages", document.pages.len());
    Ok(())
}

CLI and API server

cargo install mineru
mineru --help

The package installs the existing mineru, mineru-api, mineru-vlm, mineru-vlm-api, and mineru-office-convert binaries. See docs/usage.md for service configuration and complete options.

Python

Wheels support CPython 3.9 and newer:

pip install mineru-rs

import mineru_rs

print(mineru_rs.canonical_stem("a bad/pdf"))
mineru_rs.validate_pdf_options(0, None, True, True, True)

Python currently exposes only canonical stem handling and PDF-option validation. It does not expose asynchronous document parsing. Releases are wheels only; there is no sdist or source fallback for unsupported platforms or PyPy.

Node.js

npm install @alexsun-top/mineru

const mineru = require('@alexsun-top/mineru')

console.log(mineru.canonicalStem('a bad/pdf'))
mineru.validatePdfOptions(0, null, true, true, true)

Node.js currently exposes only canonical stem handling and PDF-option validation. It does not expose asynchronous document parsing. Node.js 18 or newer is required.

License

Licensed under either Apache-2.0 or MIT, at your option.