LlamaIndex published LiteParse, an open-source document parser released under the Apache-2.0 licence. It extracts text and layout from PDFs locally, with no cloud service or language model involved. Word, Excel, PowerPoint files and images are converted first, then parsed. Output can be Markdown, JSON or plain text, with positions for each text region. Built-in text recognition uses Tesseract, and other recognition servers can be plugged in. A complexity check estimates whether a file needs heavier processing before a full parse. It installs from npm, pip or cargo, and also runs in the browser. Python and Node.js users get worker processes with per-file timeouts. LlamaIndex says hard documents such as dense tables and scans still suit its paid cloud parser better. The earlier version of the project remains on a separate branch.