LLM Wiki cc
A personal Wikipedia an LLM maintains for you
12 followers
A personal Wikipedia an LLM maintains for you
12 followers
Drop in papers, notes, PDFs, or URLs and an agent compiles them into a cross-linked markdown wiki you own. Each new source makes every page richer, not just adds one. Open source, local-first, free to start with local Ollama models. An implementation of Karpathy's LLM Wiki pattern. Artificial Intelligence, Open Source, Developer Tools, Productivity, Notes
Hey Product Hunt, maker here. I built LLM Wiki to close the gap between two tools I use daily. RAG chatbots (NotebookLM, ChatGPT) forget your corpus the moment you close the tab. Notes apps (Obsidian, Notion) make you do all the writing, linking, and cleanup yourself.
LLM Wiki sits in between. You feed it sources, an agent writes and cross-links the pages, checks for contradictions, and keeps the index current, so knowledge accumulates instead of evaporating. It's all plain markdown you own, runs on your machine, and is free to start with local models via Ollama (or bring an OpenRouter key for frontier models).
Open source under MIT. I'd love feedback, contributors, and ideas, and I'll be here in the comments all day.
mailX by mailwarm
How do you handle citations and page updates when a new PDF contradicts something that’s already in the wiki?
@manal_essalek1 The LLM reads the new PDF against your existing wiki, so it edits the relevant page in place rather than appending a dupe (prior version always backed up to page-history). If they genuinely conflict, it won't silently overwrite, it flags the contradiction in the log and drops a > [!contradiction] callout on the page with both claims + sources, and Lint catches these on a separate pass too. Citations update automatically since every page tracks the exact sources it was compiled from, bidirectionally.
In simple terms, it updates what's already there instead of piling on duplicates, and when a new source disagrees, it tells you instead of quietly picking a side.