LLM Wiki Starter

Welcome to the LLM Wiki Starter! A framework for building a personal knowledge base using AI (LLM) agents. Instead of starting from zero every time you ask an AI a question, this system helps the AI build a persistent, interlinked wiki that gets smarter the more you use it.

✅ Simple RAG alternative: No vector databases or hard config required. ✅ Token-efficient: Uses text-based Knowledge Graphs to read fewer files. ✅ Plug-and-play: Persistently remembers your settings across all chat sessions.

Folder Structure

This project is organized into three main folders:

• 📂 raw/: Drop your source materials here (text files, articles, data, images). The AI reads these but never edits them. This is the source of truth.
• 📂 wiki/: The "brain" of the project. The AI will create and link Markdown files here to summarize concepts, entities, and facts based on your raw materials.
• 📂 outputs/: Any final work generated by the AI (like code, reports, presentations) goes here.

There are also a few core configuration files:

• 📄 AGENTS.md: The rulebook for the AI. It tells the agent how to manage folders, synthesize knowledge, and keep things organized.
• 📄 CONFIG.md: The settings file where you can turn additional AI modules (like Knowledge Graph) ON or OFF.
• 📄 wiki/graph/network.md: A lightweight text-based Knowledge Graph. The AI maps relationships block here (e.g., [[A]] --(is related to | W:0.9)--> [[B]]) so it doesn't have to read every file to understand the big picture.

(Note: The AI will generate index.md and log.md automatically as you use the system).

How to Use This (For Beginners)

You don't need to be a programmer to use this! You just need an AI Assistant that has access to your files (like GitHub Copilot in VS Code, Cursor, or Claude Code).

🚀 Quick Start (via NPX)

To instantly generate this starter in your computer, open your terminal and run:

npx create-llm-wiki-starter my-project
cd my-project

Step 1: Open the Project

1. Open your newly created my-project folder in your code editor.
2. Open the AI Chat panel.

Step 2: Add Knowledge (Ingest)

1. Copy an article, note, or document into the raw/ folder.
2. Tell your AI Assistant:

   "I just added a new file to the raw folder. Please read it, extract the key takeaways, and update the wiki folder based on the Ingest rules in AGENTS.md."

Step 3: Ask Questions & Brainstorm (Query)

1. When you want to explore the knowledge you've built up, ask the AI:

   "Based on the wiki, what are the main concepts we've gathered about [topic]?"

2. If the AI gives you a really good insight or summary, tell it to save it:

   "That's a great synthesis. Please file this back into the wiki as a new page and update the index."

Step 4: Keep it Tidy (Lint)

1. Every couple of weeks, ask the AI to clean things up:

   "Please review the wiki folder. Check for contradictions, dead links, or missing cross-references as described in the Lint workflow in AGENTS.md."

🔌 Optional: Knowledge Graph Mode (Plug & Play)

This starter includes a built-in, text-based Knowledge Graph (wiki/graph/network.md) to make your AI smarter and save processing power (tokens). It maps connections automatically with priority weights, like: [[Concept A]] --(relates to | W:0.9)--> [[Concept B]].

To activate Knowledge Graph Mode persistently:

1. Open the CONFIG.md file in this folder.
2. Change the checkbox from [ ] to [x] next to the knowledge graph module.

The AI Agent will automatically check this configuration file every time you start a new chat session. You no longer need to instruct it manually!

Why turn this on?

• Super Token-Saving: The AI doesn't need to read 100 notes to find an answer. It looks at the text map (wiki/graph/network.md) first, then only opens the 2 or 3 files it actually needs.
• Smart Priority: By parsing connection weights (W:0.9 vs W:0.2), the AI knows which relationships are critical and which are just trivia.
• Zero Setup: It runs entirely on plain text reasoning. No database scaling or extra coding required!

Enjoy building your AI-powered second brain!

🤝 Contributing

This is a personal open-source project meant to help everyone. Feel free to use it, modify it, or contribute back!

If you want to suggest a new AI module, fix a typo, or test edge cases:

1. Check the Contributing Guidelines on how to submit a Pull Request.
2. Found a bug where the AI gets confused? Open an Issue.

📄 License

This project is licensed under the MIT License - free to use for personal or commercial work!