wiki-graph-showcase
A static 3D knowledge graph built from my own Obsidian notes. Live: https://wiki.dynamic-dome.com/
Two datasets, rendered with three.js and 3d-force-graph: a competence graph on agentic engineering (the default) and a curated astrophysics nebula. Two visual themes (Crab Nebula, Dynamic Dome) and a "Gold Mode" slider that lets the edges pulse with the density of the graph. Click a node to open its summary.
What it shows
- Notes in, public site out. A Python build reads a private Markdown vault, keeps only the pages matched by the config, drops pages marked
private: true, resolves wikilinks and typed links, and writes graph.json plus one small JSON file per node. The browser loads node details lazily.
- A gate before every deploy.
tools/pre_deploy_sweep.py scans the built site and fails on internal paths, secret or private markers, status notes and pipeline jargon that should not leave the notes. It runs before every deploy.
- Static and strict. No server code, a Content Security Policy in
_headers, one shared three.js instance instead of two bundled copies.
- Tested. 138 pytest tests for the build and the sweep, Node unit tests for the frontend modules, Playwright tests for the rendered pages.
How it works
Markdown vault -> tools/build.py -> dist/ (graph.json, nodes/*.json, assets) -> Cloudflare Pages
| Path | What it is |
|---|
tools/build.py | Build orchestrator: select pages, parse, extract titles and summaries, emit JSON |
tools/parser.py, tools/filter_slice.py, tools/extract_page_meta.py | Link resolution, page selection, page metadata |
tools/pre_deploy_sweep.py | The deploy gate described above |
src/ | Frontend: HTML, CSS and ES modules (scene, themes, Gold Mode, detail panel, URL state) |
showcase.config.json, kompetenz.config.json | Which pages go into each dataset; vault_root is relative to the repository root |
docs/ | Architecture, capabilities, changelog, audits (mostly in German) |
Run it
The real graphs are built from private vaults that are not part of this repository, so a clone builds only with a vault of your own. Point vault_root in a config at any folder of Markdown notes, then:
npm install
npm run build # python -m tools.build --config showcase.config.json --out dist/
npm run serve:preview # http://127.0.0.1:8051
npm run sweep # the deploy gate
npm run test:py # pytest, uses small fixture vaults
Python 3.11+ with the standard library is enough for the build; Node is needed for the preview tooling and the Playwright tests.
License
MIT. The note contents shown on the live site are mine and not covered by this license.
