Anomaly and Precursor Timeline UX

A shared timeline that shows precursors emerging, growing, and resolving — turning the anomaly ensembles' outputs into a story an operator can read.

Anomalies have a history, not just a state

A single red/green anomaly indicator throws away the most useful information: the trajectory. The L3 anomaly ensembles surface sub-threshold quench and disruption precursors well before any threshold trips, and their value is in the trend. L6 renders anomalies on a timeline so operators see a precursor emerge, strengthen, and either resolve or escalate — reading intent, not just an instantaneous flag.

The precursor lane

Each detector contributes a lane on a shared timeline. A lane shows the precursor's strength over recent history and, where the ensemble provides it, a projected lead time to the event. Lanes are ordered by current triage score, so the most consequential emerging precursor is at the top. Sub-threshold activity is visible — not hidden until it crosses a line — which is what gives operators time to act early.

• Breeder: disruption precursors, quench precursors, magnet delta-T trends
• Burner: plug-density excursion precursors, ambipolar-potential drift
• Cross-machine: diagnostic-health degradation feeding provenance
• Each lane: strength history, confidence, projected lead time, contributing signals

From precursor to action

A lane is not just a chart; it is a starting point for action. Selecting a precursor reveals the features driving it (the anomaly ensembles are feature-attributable) and links to the recommended response and, if warranted, the mitigation path. This closes the loop from detection to decision: the operator sees what the model saw, how sure it is, and what to do about it. Confidence and provenance ride along so a precursor from a degraded, imputed input is weighted accordingly.

The timeline is the tissue connecting detection to the countdown and to replay: an escalating breeder lane hands off to the disruption countdown, and every lane is preserved in incident replay so lead times can be validated after the fact. Emergence cues are tuned per the notification budget.