
llm-wiki-compiler
The knowledge compiler. Raw sources in, interlinked wiki out. Inspired by Karpathy's LLM Wiki pattern.
The Lens
llm-wiki-compiler turns a pile of raw sources into a clean, interlinked wiki you can browse and search. Feed it markdown, URLs, PDFs, images, even AI chat transcripts, and it runs an LLM pipeline to compile typed pages with wikilinks and citations that point back to the exact source lines. It's MIT licensed, and you bring your own LLM key.
It's a Node CLI, so setup is light: install it, point it at your sources, run compile. The engineering touches are what stand out. It does hybrid semantic and keyword retrieval, tracks changes by hash so re-compiles are incremental, scores the result's health and citation coverage, and ships a read-only local viewer plus an MCP server for Claude or Cursor. You can run it free against a local Ollama model or pay per token with Claude or GPT.
There's no clean commercial equivalent. Hosted tools like Mintlify, GitBook AI, and Notion AI live in the same neighborhood, but they're managed docs platforms, not a local compiler that turns arbitrary sources into a cited wiki. Solo builders and small teams drowning in scattered notes and docs: this is worth a real try.
The catch: the author calls it early software, and it shows in the edges. Quality depends on the model and its citation judgments, every recompile costs LLM tokens unless you run local, and the viewer is read-only with no freshness watchdog yet.
Free vs Self-Hosted vs Paid
fully freeFree: The tool itself, under MIT. Ingesting, viewing, and linting need no credentials at all.
Self-hosted: It's a local Node CLI: npm i -g, then llmwiki compile. The only real dependency is an LLM provider.
The cost is the LLM: Compiling and querying call a model, so you either pay per token (Claude, GPT) or run a local model free with Ollama. Every recompile spends tokens unless you go local, so a large or frequently-changing corpus is where the bill shows up.
Free and open source (MIT). You bring your own LLM key, or run a local model free via Ollama.
Get tools like this every Wednesday
One featured tool, three on the radar. No fluff.
License: MIT License
Use freely, including commercial. Just keep the license.
Commercial use: ✓ Yes
About
- Owner
- Atomic Strata (Organization)
- Stars
- 1,492
- Forks
- 155
Explore Further
More tools in the directory
everything-claude-code
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
212.9k ★hermes-agent
The agent that grows with you
190.3k ★ollama
Get up and running with Kimi-K2.5, GLM-5, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.
173.7k ★