Repolode
Submit
run-llama

run-llama/liteparse

A fast, local PDF and document parser in Rust with bounding-box text extraction, Tesseract or HTTP OCR, and Markdown, JSON, or text output.

Open-source alternative to:Adobe Acrobat

View on GitHubWebsite
Stars
12,884
Forks
886
Open issues
44
Last commit
8 days ago

What is it?

LiteParse is a standalone open-source PDF parser focused on fast, lightweight spatial text extraction. It uses PDFium to pull text with bounding boxes and runs entirely on the local machine, without LLM features or cloud calls. Office files and images are converted to PDF first through LibreOffice or the built-in Rust image pipeline, then parsed.

Output can be Markdown with reconstructed headings, tables, lists and links, structured JSON with text and layout blocks, or plain layout-preserved text. Pages that need OCR can use the built-in Tesseract engine or any HTTP server that follows the published OCR API spec, such as EasyOCR or PaddleOCR wrappers. A complexity check command reports which pages need OCR before a full parse.

The core is a Rust workspace with bindings for Node.js, Python, WASM in the browser, and a `lit` CLI, and it supports Linux, macOS and Windows.

Key features

  • 01Spatial text extraction with bounding boxes using PDFium
  • 02Built-in Tesseract OCR, or any HTTP OCR server that follows the published API
  • 03Markdown, JSON and plain text output, including tables and layout blocks
  • 04Complexity check to route or reject documents before a full parse
  • 05Page screenshots for LLM agents, plus batch directory parsing
  • 06Bindings for Rust, Node.js, Python and browser WASM

Use cases

  • Feeding PDFs and Office documents into RAG pipelines as clean Markdown
  • Batch converting a folder of reports to JSON with positional data
  • Triaging a document queue to find which files need OCR before parsing
  • Running document parsing inside a browser or a locally isolated environment

Quick start

  1. 1

    Install the CLI via npm

    npm i -g @llamaindex/liteparse
  2. 2

    Parse a document to Markdown

    lit parse document.pdf --format markdown -o output.md

Things to know

  • Built-in OCR and advanced parsing for scanned or complex documents need extra setup or the paid cloud LlamaParse service.
  • Benchmark scores are modest on charts, math and handwriting, where it does not perform well.

Topics

#pdf#pdf-parser#ocr#text-extraction#document-processing#rust#markdown#document-ocr#ocr-recognition

FAQ

What is liteparse?
A fast, local PDF and document parser in Rust with bounding-box text extraction, Tesseract or HTTP OCR, and Markdown, JSON, or text output. LiteParse is a standalone open-source PDF parser focused on fast, lightweight spatial text extraction. It uses PDFium to pull text with bounding boxes and runs entirely on the local machine, without LLM features or cloud calls. Office files and images are converted to PDF first through LibreOffice or the built-in Rust image pipeline, then parsed.
Is liteparse free to use? What license does it have?
liteparse is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is liteparse still maintained?
Actively maintained. The last commit to liteparse was on Oct 12, 2026.
What can I use liteparse for?
Feeding PDFs and Office documents into RAG pipelines as clean Markdown Batch converting a folder of reports to JSON with positional data Triaging a document queue to find which files need OCR before parsing Running document parsing inside a browser or a locally isolated environment
How do I install liteparse?
Install the CLI via npm: npm i -g @llamaindex/liteparse Parse a document to Markdown: lit parse document.pdf --format markdown -o output.md
Where can I get liteparse?
The source code and installation instructions are on GitHub: https://github.com/run-llama/liteparse

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects