sdyckjq-lab/llm-wiki-skill

基于 Karpathy llm-wiki 方法论的个人知识库构建 Skill，支持多平台！

What it solves

llm-wiki transforms fragmented information into a structured, interconnected knowledge base. It prevents the need for an AI to re-derive information from raw documents every time a query is made by "compiling" knowledge once and maintaining it continuously.

How it works

The project operates as a monorepo with two primary interfaces:

  1. Skill Mode: A set of instructions and tools that can be installed into AI CLI agents (such as Claude Code, Codex, OpenClaw, or Hermes). The agent then manages the knowledge base directly via shell commands and file operations.
  2. Agent Workbench: A local development environment (currently for developers) that provides a chat-centric interface and an interactive knowledge graph.

It processes various inputs (PDFs, Markdown, HTML, and web content from X, YouTube, and WeChat) into a structured Wiki format using [[bi-directional links]]. It uses a two-step "analyze then generate" process to ensure data quality and employs SHA256 deduplication to prevent redundant entries.

Who it’s for

Users of AI CLI agents who want to build a persistent, structured personal knowledge base from diverse digital sources, and developers interested in interactive knowledge graph visualization.

Highlights

  • Interactive Knowledge Graph: Generates a self-contained HTML graph with a "digital landscape" aesthetic, allowing offline exploration of nodes, communities, and relationships.
  • Multi-Platform Support: Easy installation for Claude Code, Codex, OpenClaw, and Hermes.
  • Structured Wiki Generation: Automatically creates entity pages, topic pages, and material summaries with confidence labels (e.g., EXTRACTED, INFERRED).
  • Conversation Crystallization: Allows users to save valuable AI chat outputs directly into the knowledge base as permanent pages.
  • Obsidian Compatibility: All content is stored as local Markdown files, making it fully compatible with Obsidian.

Related

  • Project
    ussumant/llm-wiki-compiler

    A Claude Code and Codex plugin that compiles scattered markdown files and codebases into a synthesized, topic-based wiki to reduce AI agent context costs.

  • Project
    xoai/sage-wiki

    A graph-based knowledge base that uses LLMs to compile raw documents into an interlinked wiki and knowledge graph for shared human-AI memory.

  • Project
    Astro-Han/karpathy-llm-wiki

    A reusable agent skill that enables LLMs to build and maintain a structured wiki of synthesized knowledge, moving beyond simple RAG by compiling sources into durable, cross-linked markdown pages.

  • Project
    zosmaai/pi-llm-wiki

    A self-maintaining, Obsidian-compatible knowledge base for AI agents that transforms raw sources into a durable, interlinked wiki to prevent ephemeral RAG synthesis.

  • Project
    SamurAIGPT/llm-wiki-agent

    A coding agent skill that converts source documents into a persistent, interlinked markdown wiki with automated entity extraction and contradiction flagging.