Link Search Menu Expand Document

Bodhi Map πŸƒ β€” Documentation

An open, interactive knowledge graph of modern AI.

πŸ“– This is the project documentation. Browsing on GitHub? Every page below is a markdown file in this docs/ folder.

Bodhi Map turns the scattered landscape of modern AI β€” from tokens and attention all the way to LoRA, DPO, reasoning models, agents, and quantization β€” into a single navigable map. Every concept is a node; typed edges show how techniques relate, improve on, combine with, or build on each other. It runs fully offline in a browser with zero dependencies β€” just open index.html.

New here? Jump to Getting Started to run it, or the User Guide to learn what it can do.


At a glance

Β  Β 
Concepts 123 across 11 clusters
Relations 229 typed edges (is-a, improves-on, alternative, requires, combines, builds-on)
Runtime Single folder, no build step, no internet β€” works on a plane
Source format Open Knowledge Format (OKF) bundle β€” markdown, git-shippable, agent-ready
License MIT
Current release v0.4.0

Why it exists (philosophy)

Modern AI moves faster than anyone can read, and the knowledge that matters is real but scattered across hundreds of papers, blogs, and docs. The hard part isn’t finding definitions β€” it’s seeing the connections: how LoRA relates to full fine-tuning, when DPO beats PPO, where quantization fits. Bodhi Map draws that map in the open. Guiding principles:

  • Understanding over hype β€” every concept cites a real source and explains why and when to use.
  • Connections are the content β€” the typed edges are where the insight lives.
  • Open & offline by default β€” no accounts, no CDNs, no telemetry.
  • A path, not a pile β€” clustered from foundations to the frontier.

Bodhi (ΰ€¬ΰ₯‹ΰ€§ΰ€Ώ) is Sanskrit/Pali for awakening β€” the moment scattered facts resolve into understanding.

How the project is organized

knowledge/ is the source of truth β€” an OKF bundle of one markdown file per concept. A tiny zero-dependency build compiles it into js/data.js, which the browser app renders. See Architecture for the full pipeline.

knowledge/      ← OKF bundle (SOURCE OF TRUTH β€” edit here)
okf.config.js   ← clusters, levels, relations, release, interactive flows
scripts/        ← build + validate tooling (no deps)
js/data.js      ← GENERATED (do not hand-edit)
js/graph.js     ← the app (Cytoscape.js, vendored offline)
index.html      ← landing page (front door) Β· app.html ← the interactive map

Documentation map

  • Getting Started β€” run it in 30 seconds.
  • User Guide β€” graph navigation, search, level & cluster filters, the New highlight, the decision wizard, guided path, compare tables, export-for-agents, themes, deep links, keyboard shortcuts.
  • Concept Authoring β€” add or edit a concept (the most common contribution): the frontmatter schema, relation verbs, levels, sources, and marking something new.
  • Configuration Reference β€” extend the controlled vocabulary and interactive flows in okf.config.js.
  • Architecture β€” how the bundle β†’ build β†’ runtime pipeline works, and why it’s offline.
  • Contributing β€” the full fork β†’ PR workflow and quality bar.
  • Release Process β€” how to cut a versioned release and highlight what’s new.
  • Roadmap β€” direction and candidate additions.
  • Changelog β€” notable changes per release.
  • FAQ β€” quick answers.

Contributing in one line

Adding a concept is just writing a markdown file in knowledge/ and running npm run build (which validates). Start with Concept Authoring.