EcoPulse Stadium is a GenAI reasoning engine that ingests live stadium telemetry — crowd density, energy consumption, smart bin fill levels, and HVAC/lighting status — and produces three synchronized outputs every reasoning cycle:
| Output Block | Purpose |
|---|---|
| 📊 Organizer Sustainability Metrics | Real-time efficiency scorecard with per-sector breakdowns and DATA GAP flags |
| 🔔 Ground Staff Actions & Alerts | Short, push-notification-style instructions ordered by priority (safety → overflow → routine) |
| ⚡ Automated Utility Commands | Direct HVAC, lighting, and ventilation commands — or explicit BLOCKED notices when safety overrides them |
Core differentiator: one reasoning layer handles both sustainability and operations together, resolving trade-offs (e.g., "can we dim lights here without affecting egress safety?") in a single pass. The safety-first priority rule is a hard constraint — not a preference — and its enforcement is always visible in the UI.
| Scenario | Match State | Key Demonstration |
|---|---|---|
| ⚽ Pre-Match | 30 min before kickoff | Fans arriving; pre-cooling baseline; routine optimization approved |
| 🔥 75th Minute | Peak density | North packed, South near-empty but HVAC still at 100% — waste detected and reduced |
| 🚪 Post-Match Egress | Full-time | Safety rule fires visibly — HVAC and lighting reductions BLOCKED in active egress corridors |
| 📡 Data Gap | 62nd minute | West Concourse IoT node offline — DATA GAP protocol activates, sector commands default to MAINTAIN |
Live Telemetry Inputs
(crowd density | energy/HVAC | smart bin levels | lighting)
│
▼
┌─────────────────────────────┐
│ EcoPulse Reasoning Engine │
│ (engine.js) │
│ Safety > Sustainability │ ← hard-coded non-negotiable constraint
│ DATA GAP → MAINTAIN │ ← never act on stale/missing sensors
└─────────────────────────────┘
│
┌────────────────┼────────────────┐
▼ ▼ ▼
Organizer Ground Staff Automated Utility
Dashboard Alert Feed Commands
(efficiency (priority- (HVAC/lighting/
scorecard) ordered push ventilation adj.
notifications) or BLOCKED)
The JS reasoning engine is a rule-logic mirror of the LLM system prompt (prompts/system_prompt.txt), implemented for demo reliability so the dashboard works fully offline without API keys.
Designed to swap for a live Antigravity/Gemini API call in production — the three output block structure, the safety constraint, and the DATA GAP handling are identical. To integrate a live LLM:
- POST the telemetry snapshot (JSON) to the Gemini/Antigravity API with the system prompt as context
- Parse the three output blocks from the model response
- Render them in the same three panel containers in
index.html
PRIORITY RULE (non-negotiable):
Fan and staff safety always overrides sustainability optimization.
Never issue a command that reduces visibility, ventilation, or emergency
egress capacity in an active or exiting crowd zone, regardless of
projected energy savings.
When this rule fires (Post-Match Egress scenario), the Utility Commands panel shows:
⚠️ BLOCKED: HVAC reduction rejected — active egress zone (EAST_GATE)
Est. impact: ~38% sector power reduction (not applied — safety override)
The block is always visible and explicit — never silent.
DATA HANDLING RULE:
If telemetry for a sector is stale (not updated within the expected
interval) or missing, flag it as "DATA GAP - [sector]" and default
that sector's automated commands to MAINTAIN CURRENT SETTINGS.
The Data Gap scenario demonstrates this with the West Concourse sensor going offline at 62'. The sector is flagged in all three panels and excluded from optimization.
ecopulse-stadium/
├── index.html ← Dashboard (HTML + inline JS controller)
├── style.css ← Design system (dark glassmorphism, animated gauges)
├── engine.js ← EcoPulse reasoning engine (rule-logic)
├── telemetry.js ← Mock sensor data simulator (4 scenarios)
├── mock-data/
│ └── sample_telemetry.json ← Reference 75th-minute snapshot
├── prompts/
│ └── system_prompt.txt ← LLM system prompt for production integration
└── README.md
| Layer | Tool |
|---|---|
| Structure | HTML5 semantic |
| Styling | Vanilla CSS (glassmorphism, dark mode, micro-animations) |
| Logic | Vanilla JavaScript (no framework, no build step needed) |
| Fonts | Google Fonts — Outfit + JetBrains Mono |
| Data | Mock JSON telemetry (telemetry.js) |
| LLM (production) | Antigravity / Gemini Flash |
No build step required. Just open index.html in any modern browser:
# Option 1: direct open
open index.html # macOS
start index.html # Windows
# Option 2: local dev server (recommended for production swap)
npx serve .| Metric | Target |
|---|---|
| HVAC energy reduction (low-density zones) | Up to 40% per sector |
| Bin overflow events | Eliminated via predictive alerts |
| Staff response time | Reduced via plain-language push notifications vs raw dashboards |
| Safety blocks honored | 100% (hard constraint, not a preference) |
- All crowd data is aggregated zone-level only — no individual tracking
- Safety rule is hard-coded into the reasoning engine, not left to model discretion
- Designed as an integration layer over existing IoT/HVAC systems — no new hardware required for pilot
- DATA GAP handling ensures no automated action is taken on incomplete sensor data
Built for the FIFA World Cup 2026 hackathon. The 2026 tournament runs across 16 stadiums in 3 countries, with 104 matches and 5.5M+ attendees. Sustainability operations (HVAC, lighting, waste) and staff operations run as separate silos today — EcoPulse closes the loop with a single reasoning layer.
See the full technical context in docs/ and prompts/system_prompt.txt.