I built a minimal full‑stack web UI for the incident investigation and live‑agent workflow, with a FastAPI backend in backend/app.py and a static‑file serving frontend.
I start the server with python -m uvicorn backend.app:app --reload --port 8000 and open the UI at http://localhost:8000/.
The live‑agent run endpoint is POST /api/live-agent/run {persona, enforcement, amount, ...}.
Each node lights up in the diagram and appears as a trace chip, one at a time, followed by a state‑snapshot of the key fields (persona, decision, amount, threshold, risk_score, outcome) once the run completes.
The same incident data structure can be rendered to PDF/JSON for regulators, or piped into a ticketing system.