[CodeBuildder]

Closes #96

Root cause

The console polls /cluster-overview immediately after startup. Although that FastAPI route is async, it performs synchronous Kubernetes configuration and API calls. A stale or unreachable kubeconfig can therefore block the event loop, causing /health and /simulate-threats to time out as well. /infra-observability had the same behavior. The command-center simulation request also had no client timeout, leaving its loading state active indefinitely whenever the backend stopped responding.

What changed

• launch the backend with explicit ARGUS_LOCAL_DEMO=true and IN_CLUSTER=false settings from make demo-local
• bypass all Kubernetes discovery in both cluster overview and infrastructure observability when local-demo mode is active
• continue returning deterministic incident-derived fallback telemetry to the console
• expose mode: local_demo through /health for troubleshooting
• avoid noisy Kubernetes fallback warnings during the intentionally cluster-free path
• bound command-center simulation and refresh requests to five seconds
• always clear the Simulating state and show an actionable success, timeout, or failure message
• add regression tests for health mode and both fallback endpoints

User impact

make demo-local no longer becomes unresponsive because an unrelated kubeconfig points at a missing cluster. Threat generation remains entirely local and should return immediately. If the backend becomes unavailable for another reason, the UI now stops spinning and tells the operator to restart the demo. Cluster-backed launches retain the existing Kubernetes discovery behavior.

Verification

• 53 action, enrichment, and reasoning tests passed
• 17 local-demo, simulation, and World Model tests passed
• focused local-demo and simulation suite passed after the UI follow-up
• UI TypeScript and Vite production build passed
• git diff --check passed