Incident Reasoning Graph
The graph always stays visible. Lens buttons activate different circuits while the human approval gate remains connected to remediation, recovery, and postmortem.
Activation controls
These controls do not hide nodes. They light up the functional circuit while keeping the full incident brain visible.
Human-controlled neural canvas
Investigating: checkout API latency after a cache configuration change. Active path: deployment context → metric anomaly → hypothesis → approval gate.
Human-in-the-loop behaviour
Relation depth expands nearby context, while the human approval gate always keeps the downstream remediation → recovery → postmortem path visible.
Whole graph remains visible
Lens buttons now behave like neural activation circuits, not destructive filters.
Approval gate is explicit
The policy-to-approval edge is highlighted as the human-in-the-loop checkpoint before remediation.
Downstream record remains connected
Remediation, recovery, and postmortem stay visible so the graph shows the full operational consequence.