run-llama/liteparse

A fast, helpful, and open-source document parser

What it solves

LiteParse is a fast, lightweight, and standalone open-source PDF parsing tool. It solves the problem of extracting high-quality spatial text and structure from documents locally, without requiring proprietary LLM features or cloud-based dependencies.

How it works

Built with a Rust core, LiteParse uses PDFium for fast text extraction and a grid projection system for spatial layout reconstruction. It supports multiple input formats (PDF, DOCX, XLSX, PPTX, and images) by converting them via LibreOffice or Rust-based image crates. For documents requiring OCR, it offers a built-in Tesseract engine or the ability to plug in external OCR servers (like EasyOCR or PaddleOCR) via a standard HTTP API. The tool can then classify pages into blocks (headings, tables, lists, etc.) to render structured Markdown, JSON, or plain text.

Who it’s for

It is designed for developers building RAG pipelines and LLM agents who need a local, high-performance document parser that provides precise bounding boxes and structured output without the latency or cost of cloud services.

Highlights

  • High Performance: Spatial text parsing typically takes 2-5ms per page.
  • Local Execution: Runs entirely on the user's machine across Linux, macOS, and Windows.
  • Multi-language Support: Available as libraries and CLI tools for Rust, Node.js/TypeScript, Python, and the browser (WASM).
  • Flexible OCR: Bundles Tesseract for zero-setup OCR and supports custom OCR servers via a standard API.
  • Structured Output: Generates Markdown (with tables and headings), JSON, and plain text, including precise bounding boxes for text and layout blocks.
  • Agent-Ready: Can be integrated as an agent skill and generates high-quality page screenshots for LLM agents.

Related

  • Project
    docling-project/docling-parse

    Docling Parse is a fast, configurable PDF‑parsing library (Python API with a C++ core) that extracts text, words, lines, vector shapes and bitmap images together with their coordinates. It supports both sequential and multi‑threaded parsing, lets you choose what to compute or materialize per page, and includes rendering utilities. Ideal for AI workflows that need layout‑aware document data.

  • Project
    arcships/light-ocr

    light‑ocr is a cross‑platform, offline OCR library for Node.js and C++. It bundles the model, PDF renderer, and a Chinese fallback font in a single npm package, provides pre‑built native binaries for macOS, Linux and Windows, and can accelerate inference with Core ML or WebGPU. The API returns text lines with confidence scores and quadrilateral coordinates, and a CLI offers quick command‑line usage. Model sizes range from a 6 MB Tiny tier to a 139 MB Medium tier, all under Apache 2.0.

  • Project
    firecrawl/pdf-inspector

    pdf‑inspector is a Rust‑based library (with Python, Node, and WASM bindings) that rapidly classifies PDFs, extracts position‑aware text, and converts it to clean Markdown. It can route only the pages that need OCR to a local model, saving time and cost. Benchmarks show it faster and more accurate than similar tools on a 200‑PDF test set.

  • Project
    drmingler/smart-llm-loader

    A Python package that converts documents and scanned images into clean, structured markdown chunks using OCR and LLM-powered semantic chunking for RAG systems.

  • Project
    PaddlePaddle/PaddleOCR

    PaddleOCR is an open‑source OCR and document‑AI toolkit that converts images or PDFs into structured, LLM‑ready text, tables and markdown/JSON. It offers multilingual scene‑text models (PP‑OCRv6) and lightweight vision‑language models (PaddleOCR‑VL, PP‑StructureV3) with state‑of‑the‑art accuracy, fast CPU/GPU inference, and deployment options ranging from Docker to browser SDKs. Integrated with RAG platforms like Dify and RAGFlow, it serves as a data‑engine for building AI agents and retrieval‑augmented generation pipelines.