LlamaIndex Releases LiteParse: A CLI and TypeScript-Native Library for Spatial PDF Parsing in AI Agent Workflows

LlamaIndex released LiteParse, llamaIndex released a new library named LiteParse.

LiteParse is TypeScript-native and runs on Node.js, liteParse is implemented without Python dependencies.

Uses PDF.js (pdf.js-extract) for text extraction and Tesseract.js for OCR, text extraction uses PDF.js and local OCR uses Tesseract.js.

Projects text onto a spatial grid to preserve page layout, the library outputs preserved-layout text rather than converting to Markdown.

Can generate spatial text files, page-level screenshots, and JSON metadata, liteParse can produce multiple output types during parsing.