owenliang/llm-wiki

This page was translated automatically. Technical details may differ from the original. View the original

85 stars Quality 50 Trend 14

The implementation of karpathy/llm-wiki.md

Overview

This project is an LLM-based intelligent knowledge base management system, integrating Qoder Agent, Obsidian, and QMD semantic retrieval tools to achieve automatic knowledge accumulation and efficient retrieval. LLM Wiki is an incremental knowledge base construction model. Unlike traditional RAG, it **s a structured, interconnected collection of markdown files, with knowledge accumulated rather than reretrieved each time. Skill file location: .qoder/skills/llm-wiki/SKILL.md - : Save original content to raw/articles/, automatically analyze and update wiki - : Answer user questions based on wiki content - : Scan all pages to check for contradictions, outdated statements, orphaned pages, etc Examples: - /llm-wiki https://qwen.ai/blog?id=qwen3.6 - /llm-wiki https://github.com/user/repo/blob/main/docs/guide.md Example: - /llm-wiki Search Agent platform - /llm-wiki What Agent platforms are available - /llm-wiki Differences between CoPaw and OpenClaw (No command prefixes): - "Add this article" + article content → Automatically analyze and consolidate it into wiki - "Query knowledge base" + Question → Answer based on wiki content - "Check wiki" or "lint" → perform health check 3. (Optional but recommended) - Set → Core Plugins → Enable the following plugins: - Graph view - Backlinks - Outgoing links

Overview (machine translated)

Project Origin

This project is built on Karpathy’s LLM Wiki concept, with the core concept being:

LLM Wiki is an incremental knowledge base construction model. Unlike traditional RAG, it continuously builds and maintains a structured, interconnected collection of markdown files, with knowledge accumulated rather than retrieved each time.

Project Architecture

Core Components

1. Qoder Agent + LLM Wiki Skill

Used to automatically analyze new articles and deposit them into structured wikis.

Installing Skill

Method 1 - Using the create-skill command (recommended):

Method 2 - Manual Creation:

Skill file location: .qoder/skills/llm-wiki/SKILL.md

Skill Features

Receive new articles: Save the original content to raw/articles/, automatically analyze and update the wiki
Query Knowledge Base: Answer user questions based on wiki content
Health Check: Scan all pages to check for contradictions, outdated statements, isolated pages, and more

How to use

This Skill supports the /llm-wiki command prefix and provides two main functions:

1. Add new articles

Example:

/llm-wiki https://qwen.ai/blog?id=qwen3.6
/llm-wiki https://github.com/user/repo/blob/main/docs/guide.md

2. Search Knowledge Base

Example:

/llm-wiki 搜索Agent平台
/llm-wiki Agent平台有哪些
/llm-wiki CoPaw和OpenClaw的区别

Natural Language Mode (No command prefixes):

"添加这篇文章" + Article Content → Automatically analyzed and stored on wiki
"查询知识库" + Question → Answered based on wiki content
"检查 wiki" or "lint" → perform health checks

2. Obsidian

Used for visual browsing and editing of wiki content.

Configuration steps

Install Obsidian

Download link: https://obsidian.md/
Install and start up

Open Vault

Select 打开本地仓库
Select the root directory of the llm-wiki project

Enable Core Plugins (Optional but Recommended)

Set → core plugins → enable the following plugins:
Graph view
Backlinks

This page was translated automatically. Technical details may differ from the original.

Recommended Tools

Try a different keyword or remove a filter.

Install

npx skillfish add owenliang/llm-wiki

New to MCP? Start with the mcpskill.org guide →