Browse / LLM wiki / llm-wiki-starter

llm-wiki-starter

aliepratama

A minimalist boilerplate for Karpathy’s LLM Wiki.

Open aliepratama/llm-wiki-starter → Checked against the four rules.

Works with any agent

ships AGENTS.md, the cross-tool instruction file.

In its own words

Aboutaliepratama/llm-wiki-starter

A plug-and-play folder structure and protocol starter for building persistent knowledge bases without the RAG overhead.

README.md

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:

The repository's own description, then 22% of its README. Read the rest on GitHub ↗

Read from the repository's git tree on 2026-08-17.