LLM Wiki is a personal knowledge base — part blog, part digital garden, part encyclopedia. It's built around three ideas:
- Articles as nodes, not streams. Every page stands on its own and is reachable from many other pages. There is no "feed" to scroll; there is a network to explore.
- Connections are first-class. Internal links, backlinks, related articles, and a graph view make the structure of the knowledge visible. The map is the territory.
- Markdown, no fuss. All content is plain text. No database, no accounts, no comments, no editorial pipeline. You can read it in any editor, version it in git, and host it anywhere.
How to use it
Click around. Start at the homepage for featured articles and recent updates, or open the graph view to see the whole network. The search modal (⌘K) finds articles by title, description, body, or tag.
Each article has a sidebar for navigation, a table of contents on the right, and three sections at the bottom:
- Related articles — pages that this one links to.
- Articles that mention this — backlinks: pages that link here.
- References — the sources cited in the article.
The technology
Built with Astro for static site generation, Fuse.js for client-side fuzzy search, Shiki for syntax highlighting, and a hand-rolled D3-style force simulation for the graph view. No JavaScript framework — the whole site ships as static HTML with a small JS layer for search, theme, and the graph.
The collections
The wiki is organized into nine content types:
- Notes — quick observations, ideas, fragments
- Articles — long-form essays and tutorials
- Concepts — definitions, explanations, mental models
- People — notable figures
- Projects — active work and side projects
- Books — books read, in progress, or to-read
- Research — research questions and findings
- Timeline — events, milestones, releases
- Sources — external references
Source
Content is in a public Git repository as Markdown files. The schema is enforced at build time via Zod, so frontmatter is type-checked. Adding a new article is a PR.
License
Content is licensed CC BY-SA 4.0. Code is MIT.