Back to Gap Map Model components / Document conversion & OCR

LiteParse

LlamaIndex
open source / Overall score: 3.4

LiteParse is a local document parser with a Rust core that extracts text with its position on the page and reconstructs layout into Markdown, JSON or plain text. It reads PDFs through PDFium, converts Office files and images first, runs Tesseract or a plug-in OCR server on scanned pages, and renders page screenshots for agents. LlamaIndex publishes it for Python, Node.js, Rust and the browser.

Openness

5 high confidence
5.0
license
Apache-2.0 (OSI. the LICENSE body is the Apache License 2.0)
source
public (run-llama/liteparse)
core features withheld
no — no ee, enterprise, commercial or proprietary path. the README names LlamaParse as LlamaIndex's separate cloud parser

LiteParse is released under the Apache License 2.0 with its full source public, and everything it does runs locally. LlamaIndex sells LlamaParse, its hosted parser, as a separate product rather than as a paid edition of this one.

Adoption

3 high confidence
3.0

Adoption is measured as PyPI downloads of liteparse, the Python binding published from the repository. The npm package is declared as well, and its downloads are not part of this figure.

Capability

4 medium confidence
4.0

LiteParse rebuilds a page's layout into Markdown with headings, tables and lists, and returns each piece of text with its position. It sits a step below Docling because it uses no vision-language model and returns no fields a user defines.

  • https://raw.githubusercontent.com/run-llama/liteparse/main/README.md recorded 2026-11-12

    README Overview: "Fast Text Parsing: Spatial text parsing using PDFium"; "Markdown Output: Structured Markdown with headings, tables, lists, images, and links"; "Bounding Boxes"; OCR via bundled Tesseract or HTTP servers; input formats PDF, DOCX, XLSX, PPTX and images.

Verified 2026-11-12