Novus Incident Copilot

Novus Incident Copilot – How It Works: 5 autonomous steps from detection to resolution
Live Incident Dashboard – Incidents tracked in real time with severity classification
Gemini AI Analysis – Root cause, AI summary, and suggested fix generated automatically
Real-time Signal Feed and Timeline– DROP, FRICTION, and SPIKE signals streamed live from Novus.ai
Live Incident Dashboard – Incidents tracked in real time with severity classification
Novus Memory – Product context mapped: 3 Personas, 5 Product Areas, 5 Key Flows

Inspiration

We watched product teams waste hours every week in monitoring dashboards — manually correlating metrics, guessing root causes, and by the time they understood an issue, users had already churned. Traditional incident management tools are reactive dashboards, not proactive intelligence systems.

What it does

Novus Incident Copilot is the answer. It's an intelligent incident management system that: Detects real-time signals (friction, errors, drops, dead flows) Clusters correlated signals into prioritized incidents Analyzes root causes using Google Gemini 1.5 Flash AI Advances incidents through resolution workflows autonomously Notifies teams instantly via Slack/email Suggests AI-powered fixes for fast resolution

How we built it

Frontend: React 19 + Vite + Tailwind CSS 4 (glasmorphic dark UI) Real-time: TanStack Query with 3-second polling AI: Google Gemini 1.5 Flash API (root cause analysis) State Management: React hooks + TanStack Query cache Design System: Glasmorphic cards, glow effects, smooth animations (0.2–0.4s) Responsive: 3-column desktop, 2-column tablet, 1-column mobile

Challenges we ran into

Live Incident Dashboard

Real-time incident grid (state badges: new/investigating/fixed/resolved) Signal timeline showing related anomalies Impact metrics (users affected, % of traffic) State selector (change incident status with one click) Metrics bar (total, by severity, average impact)

AI-Powered Analysis

One-click "Analyze" button triggers Gemini API Auto-generated summaries: root cause + suggested fix Typewriter animation (2–3 sec) for engaging UX Loading states with helpful feedback Graceful error handling with fallbacks

Autonomous Demo Engine (The Magic)

Auto Mode toggle: turn on, demo runs forever Speed slider: 1–10 second intervals (default 3s) Every 2–5 seconds (random delay): 35% chance: Gemini AI analysis 30% chance: State advancement (new → investigating → fixed → resolved) 35% chance: New incident creation

Toast notifications announce all actions Reset button wipes + re-seeds 10 incidents + 7 signals

What we learned

Speed builds trust: Investors instantly "get it" when they see autonomous updates happening
Design matters: Glasmorphic UI + glow effects make AI feel approachable, not scary
Demo mode is a superpower: Let the product speak for itself (no slides needed)
TanStack Query is incredible: Polling + optimistic updates = responsive UX without WebSockets
Animations tell stories: Typewriter effect makes 2–3 sec wait feel intentional + engaging
Component reusability: 70% of Bolt code transfers directly to Claude Code full-stack rebuild