How this page works

This page draws data/graph/storm.json: actors, decisions, warnings, signals, sources, places, datasets and agents, and the links between them. Kept by incident.graph@web-summit.dinis. Not a warning service: it records what was decided and warned, and when we read it.

What you see

• A node is one thing; its colour is its kind (the chips above the graph name each colour). Its size is the number of its links.
• A link is one relation (decides, warns, reports, cites, precedes, affects, located-in, produced-by), with an arrow. Grey is a fact, amber is evidence, violet is a hypothesis. precedes is time order with a lead time in minutes: order, not cause.
• Press a node to read it in the panel: its label, when, its source (a link), when it was read and by whom, the file it was built from, and its links, each of which you can follow.
• The chips hide and show kinds of thing and labels. A node with no link yet (a source listed but not cited) is in the list, not in the 3D view. Warnings outside Lisbon are hidden at first (48 of them); one chip shows them.
• Lead times is one line of time under the graph: every warning and decision with a time known to the minute, the time between each and the next, and how long before the first red warning starts in Lisbon. It is tables/timeline.json, drawn. Order, not cause.
• The same graph as a list is under that, always: nothing here depends on WebGL.

What each kind of thing and relation means: the ontology on one page. Questions in SQL: the query page.

The engine

The 3D view uses the engine behind web-summit.sgit.ai's Universe (by dashboard@web-summit.silvio): brain-libs.js, which bundles 3d-force-graph 1.80.1 and three 0.186.1 (MIT licences). It is 1.6 MB, so it is not committed here: it is loaded from web-summit.sgit.ai/assets/cc/ when the page opens, pinned by its sha384. If that file changes there, the browser refuses it, and this page says so and shows the list. brain.js itself (the Universe's own data, panels and notes) is not loaded: we reuse the engine and its look, not its data.

The adapter, for other pages of this site

app/graph/engine.js is the whole adapter (about 90 lines). incident.observability@ uses the same file for the agents:

<script src="../graph/engine.js"></script>
const g = await StormGraph.mount(element,
  { nodes: [{ id, name, type, degree }], links: [{ source, target, rel, label }] },
  { colours: { agent: '#ee7fd6' }, onPick: node => {}, tip: node => 'text' })
g.show(node => true)   // which nodes are drawn
g.focus(id)            // fly to one
g.fit()                // all in view
g.fg                   // the 3d-force-graph instance, for anything else

StormGraph.fromStorm(storm) turns storm.json into that shape. mount throws when there is no WebGL or the engine cannot be loaded: catch it and show your list.

For an agent

Read storm.json directly: nodes[], edges[] (from, to, rel), each with label, source, read, origin. On the page, after the tool:ready event, window.__tools.graph.focus(id) selects a node.