LLM Wiki: A Karpathy‑Inspired Personal Knowledge Base Now Available as a Desktop App

LLM Wiki is an open‑source cross‑platform desktop application that automatically converts your documents into an organized, interlinked knowledge base.

It differs from traditional Retrieval‑Augmented Generation (RAG) by letting the LLM incrementally build and maintain a persistent wiki instead of performing a fresh retrieval for each query.

Core Concept

Traditional RAG vs LLM Wiki

Knowledge storage : RAG uses temporary retrieval each time; LLM Wiki stores a persistent wiki.

Processing method : RAG follows retrieve‑then‑answer from scratch; LLM Wiki incrementally constructs and continuously maintains the knowledge.

Context linking : RAG relies on similarity‑based retrieval; LLM Wiki employs a knowledge graph with bidirectional links.

Knowledge accumulation : RAG has no accumulation, recomputes each query; LLM Wiki continuously accumulates and becomes smarter with use.

Human involvement : RAG is passive query; LLM Wiki enables human‑AI collaboration with asynchronous review.

Core Architecture

Raw Sources
    ↓ immutable, keep original text
Wiki
    ↓ LLM‑generated, structured knowledge
Schema
    ↓ rules and configuration

Key Design:

Human curation, LLM maintenance — humans define goals and direction, LLM executes and maintains.

Three‑layer architecture — Raw Sources → Wiki → Schema.

Obsidian compatibility — the generated wiki can be used directly as an Obsidian vault.

Installation Guide

Pre‑built Binaries

Download the appropriate release package from the GitHub releases page.

Build from Source

# Prerequisites: Node.js 20+, Rust 1.70+

git clone https://github.com/nashsu/llm_wiki.git
cd llm_wiki
npm install
npm run tauri dev      # development mode
npm run tauri build    # production build

Chrome Extension

Open chrome://extensions.

Enable “Developer mode”.

Click “Load unpacked” and select the extension/ directory from the releases download.

Quick Start

Launch the application and create a new project (choose a template).

Configure the LLM model.

Import source files or folders.

Observe the panel – the LLM automatically builds wiki pages.

Use the chat interface to query the knowledge base (e.g., ask for a single‑node K8s installation script).

Review generated projects; manually resolve any doubts.

Run periodic lint to keep the wiki healthy.

Configure Chinese language support if needed.

The workflow can be expressed as:

Ingest (edit/import) → Govern (analyze/clean) → Link (graph/community) → Insight (detect issues) → Research (deep supplement) → Output (high‑quality dialogue)

The system continuously iterates, discovers knowledge structures, and conducts deep research, making it suitable for users who need to handle large, complex information, conduct research, or engage in long‑term learning.

Reference

Project repository: https://github.com/nashsu/llm_wiki