Appendix B: the knowledge graph explorer

This appendix is an alternative interface to the same model, not new content. Every chapter closes by saying there is more in the model; the explorer is where a reader goes to see it. It is rendered from the same graphs the site is rendered from (the vocabulary, the sources, the rulings, the essentials, the shapes, the canonical model graph and the measles evaluation), and it is regenerated by the gate and diffed byte for byte. Every node of the record is flagged synthetic in the data, so the page can tell synthetic content from the rest.

Each button is a view, with the perspective it encodes stated in two halves, what it brings into focus and what it leaves out: the vocabulary, the rulings, the process, the wiring, the record, the essentials, and the crosswalk from the front page. Click a node to read everything the graph says about it, follow a reference, and jump to the page where it is rendered. Double-click a node to focus on it: only its neighbourhood at a chosen distance (1, 2 or 3) stays in view, the rest fades or hides, the legend counts what is shown, Escape restores the view, and the address bar carries the focus so a neighbourhood can be shared as a link. The ogc tool is the machine interface to the model; this page is the human one, and every node names the ogc command that prints it.

The graph view works from a file on disk; the SPARQL box needs the page served over http, because it fetches the Turtle files and the WebAssembly engine (from a clone, serve the repository root, not the explorer/ folder, since the page loads its data and engine from ../explorer/: uv run python -m http.server 8765 at the root, then open http://localhost:8765/explorer/index.html). The explorer is served as static files next to the site, under explorer/; the data it draws is explorer/graph.json, the same nodes, links and views, for anyone who wants the graph without the page. The SPARQL box loads one merged file, explorer/data/all.ttl, every graph in one default graph, which is what the shapes and the record’s queries assume.

The worked example is a synthetic case, and every node of its record and its agents says so; the button beside the focus bar includes or excludes synthetic data from every view (sheet 10-43), and the deep link keeps the choice.