Your LLM Wiki Deserves a Canvas, Not a Folder

The LLM Wiki pattern is powerful: compile raw knowledge into interlinked pages that compound over time. But when your wiki lives in markdown files inside Obsidian, you lose the most important thing — spatial context.

Levelry gives your compiled knowledge a visual canvas. Concepts become objects. References become connections. Your wiki has a shape you can actually see.

Start for free See full features

What is the LLM Wiki pattern?

The term comes from Andrej Karpathy's widely-shared idea: instead of retrieving knowledge on the fly with RAG, compile it once into a persistent, interlinked wiki. Each concept gets its own page. Pages link to each other. Your explorations compound.

The pipeline is simple: dump raw sources into a folder, run them through an LLM, and get structured wiki pages back with [[wikilinks]] between related concepts.

The problem

Every implementation outputs markdown files. You open them in Obsidian, get a text-based graph view, and browse them like any other note collection. But knowledge has shape. Systems have dependencies. Lore has layers. Mechanics connect to each other in specific ways. A folder of files can't show you that.

The fix

Levelry takes the same compiled knowledge and puts it on an infinite canvas. Instead of browsing files, you see the structure. Instead of clicking through links, you zoom and pan across relationships. Instead of maintaining two tools (a compiler and a viewer), you have one: a visual knowledge base with AI copilot and MCP support.

Obsidian vs Levelry for LLM Wiki

	Obsidian	Levelry
Compiled knowledge input	Markdown files	Canvas objects
Visual structure	Graph view approximation	Infinite spatial canvas
Documents per concept	Separate .md files	Embedded in every object
Layer separation	Folders/tags	Canvas layers
AI copilot	Plugin-dependent	Built-in, canvas-aware
MCP support	Community plugin	Native
Connections between concepts	Wikilinks in text	Visual links on canvas

How it works

Compile: Paste a URL, drop a file, or type raw notes. Levelry extracts concepts and connections.
See: Every concept becomes an object on the canvas. Every reference becomes a visual connection.
Explore: Zoom out for the big picture. Click into any object for full documentation.
Grow: Ask the AI copilot questions. Save answers as new objects that become part of your wiki.
Automate: Connect via MCP to let your agents read and write directly to the canvas.

For game designers

Game Design Documents are wikis by nature. Mechanics connect to systems, systems connect to lore, lore connects to characters. LLM Wiki gives you the compilation step. Levelry gives you the visual layer where all those connections actually make sense.

Your GDD stops being a folder of documents and becomes a map you can navigate, rearrange, and expand without losing the context of how everything fits together.

For knowledge workers

Researchers, writers, and anyone building a personal knowledge base: the LLM Wiki pattern solves retrieval. Levelry solves comprehension. When you can see how 50 concepts relate, you make decisions faster than clicking through 50 files.