open-llm-wiki

A lightweight, local-first starter architecture for collaborative Human-AI knowledge management, inspired by Andrej Karpathy's LLM-Wiki pattern and hardened with enterprise data governance principles.

🌟 Philosophy

Most AI-assisted documentation setups fail due to context drift and file sprawl: agents generate loose files everywhere, forget conventions, and break internal links.

Open LLM-Wiki solves this with five core rules:

• Simple UX ( 10-Second Setup): Zero external libraries, zero databases, zero build steps. Works out of the box with Python 3 and any AI coding assistant (Google Antigravity, Cursor, Claude Code, Windsurf, Aider).
• Own the Migration Moment: Turn unstructured meeting notes, transcripts, and brain dumps into structured knowledge in minutes with the built-in doc_ingest skill.
• AI is the Operating Interface: AI agents don't just read documents; they maintain the wiki index, cross-reference concepts, and keep links healthy following AGENTS.md.
• Indestructible Core: A standalone, zero-dependency validation script (scripts/check_docs.py) acts as a deterministic gate against broken links, missing metadata, and orphan documents.
• 100% Local-First Data Ownership: All content is pure Markdown on your local file system. No proprietary cloud database, no vendor lock-in.