Liteparse

LiteParse is a local parser for PDFs, Office files, and images that returns text with bounding boxes, which is the main reason to reach for it. If you're building layout-aware RAG or need to point back to where a citation came from on the page, the per-item coordinates save you from writing your own geometry pass. It also renders pages to PNG, so a multimodal agent can get screenshots alongside the parsed text for the same pages. Processing runs locally, with no cloud account needed, though LibreOffice is required for Word and Excel inputs and missing Tesseract language data downloads on first use. The source is careful about what parallelizes (OCR does, whole-document PDF parsing doesn't) and notes that the batch wrapper's JSON differs from the native CLI schema, so check that before mixing outputs. For dense tables or handwriting, the skill itself points you toward a cloud service.

skills.sh installs: 1.6K
GitHub stars: 56.6K
last updated: Jul 31, 2026

View on GitHub

Install

$ npx -y skills add K-Dense-AI/scientific-agent-skills --skill liteparse --agent claude-code

Installs into .claude/skills of the current project.

Files

LiteParse — Local Document Parsing

Overview

LiteParse is a fast, open-source document parser (Rust core, Python/Node bindings) focused on local, layout-aware text extraction with bounding boxes. It does not produce Markdown and does not call cloud LLMs. Outputs are plain text (layout-preserved) or structured JSON with per-page text_items (position, font metadata, optional confidence).

Version note: Examples target liteparse 2.0.0 (PyPI, May 2026). The upstream V1 branch is legacy; this skill documents V2 / main only.

For parser selection vs MarkItDown, the pdf skill, or LlamaParse, see references/choosing_a_parser.md.