PageIndexFrom the builder
OpenKB: An Open-Source LLM Knowledge Base
Inspired by Andrej Karpathy's LLM Knowledge Bases workflow, PageIndex open-sourced OpenKB, a command-line tool that turns raw files in a folder into an interlinked Markdown wiki compiled by an LLM, with querying, health checks and Obsidian support. Karpathy had flagged long books and PDFs as the hard part, so OpenKB indexes PDFs above a threshold, 20 pages by default, with a PageIndex tree that the model reads instead of the full text. It handles PDF, Word, PowerPoint, Excel and more through Microsoft's markitdown and can watch the folder for new files.
Drop raw files into a directory; an LLM compiles them into a structured, interlinked wiki of Markdown files.
Saved . A shortened excerpt from the official blog; the original publisher retains authorship.