LLM Wiki: Turn Your Dumb Documents Into a Smart, Self-Updating Knowledge Base
You know that feeling when you’ve got a pile of PDFs, markdown files, or meeting notes, and you just know the answer is in there somewhere—but you can’t find it? Or worse, you find it, but it’s outdated because nobody updated the doc from last quarter?
That’s the exact problem LLM Wiki solves. It takes your static documents and turns them into a live, searchable knowledge base that updates itself. No manual tagging, no database schema, no sysadmin nightmares. Just point it at your files and let the LLM do the heavy lifting.
What It Does
At its core, LLM Wiki is a CLI tool that watches a folder of documents (PDFs, plain text, markdown, you name it) and builds a searchable index using an LLM (like GPT or local models). When you ask a question, it retrieves the most relevant chunks from your docs, feeds them to the LLM, and spits out a contextual answer.
But here’s the twist: it doesn’t just build a one-time index. It watches for changes—new files, edits, deletions—and updates the knowledge base automatically. Think of it as a private, always-fresh Wikipedia for your team or project.
Why It’s Cool
Most “RAG” (Retrieval-Augmented Generation) tools require you to manually re-index every time you add a document. LLM Wiki says “nah, I got this.” The auto-update feature is genuinely useful because:
- No cron jobs or scripts. It uses filesystem watchers (like
inotifyon Linux) to detect changes in real time. - Zero configuration to start. Point it at a folder. Done. It even ships sensible defaults for chunk size and overlap.
- Local-first option. Don’t want to send your docs to OpenAI? It works with local models through Ollama or llama.cpp. Privacy fiends, rejoice.
- Plug-and-play with your existing workflow. Drop a markdown file into the watched folder, and within seconds it’s searchable.
Use cases? Perfect for:
- Product teams who keep their specs in a shared drive
- Open-source maintainers documenting their own repos
- Anyone who has a “knowledge base” that’s really just a dump of random notes
How to Try It
The repo is straightforward, Python-based, and requires minimal dependencies.
# Clone it
git clone https://github.com/nashsu/llm_wiki
cd llm_wiki # Install (preferably in a virtual env)
pip install -r requirements.txt # Point it at a folder of docs and run
python llm_wiki.py /path/to/your/docs
You can choose your LLM backend via environment varia