Extreme Weather Early-Warning Dashboard — SDG 13.1
Helios is an AI-powered early-warning dashboard that helps operations and safety teams see which locations in a city or university campus are at risk from upcoming extreme weather (heat, storms, flooding) — and what actions they should take.
"Every minute of advance warning saves lives."
Helios provides a real-time operational view of weather threats across monitored locations:
- 🗺️ Interactive Map — Stamen Terrain tiles with risk-colored markers, analysis area boundaries, and click-to-inspect popups
- 🚨 Smart Alerts — Severity-ranked alert cards (High / Moderate / Low) that update based on time horizon and city
- 🧠 AI-Powered Reasoning — Click any alert to see a detailed modal explaining why the AI flagged it, how the prediction was made, and what to do about it
- 🔬 3-Model Consensus — Every prediction is cross-referenced by 3 independent AI models (Helios, NOAA HRRR, ECMWF-IFS) with confidence scores
- 📎 Source Verification — Clickable links to NWS, NOAA, USGS, and academic sources so humans can cross-verify every prediction
- ⏱️ Time Horizon Toggle — Switch between Today, Tomorrow, and 3-Day outlooks to see how risk evolves
- 🏙️ Multi-City Support — Switch between State College, Pittsburgh, and Philadelphia — each with unique locations, alerts, and analysis boundaries
The main dashboard showing the terrain map with risk markers and active weather alerts in the sidebar.
Clicking an alert opens a smooth animated modal with full AI reasoning, model consensus, predicted outcomes, recommended actions, and verification sources.
| Layer | Technology |
|---|---|
| Structure | HTML5 (semantic, single-page) |
| Styling | Vanilla CSS (custom design system with CSS variables) |
| Logic | Vanilla JavaScript (ES Modules) |
| Mapping | Leaflet.js with Stadia Stamen Terrain tiles |
| Fonts | Inter via Google Fonts |
| Build | None — zero build step, just open index.html |
# Clone the repo
git clone https://github.com/KingReaper6940/Helios.ai.git
cd Helios.ai
# Serve locally (any static server works)
npx -y serve .
# or
python -m http.server 8080Then open http://localhost:8080 in your browser.
- A modern web browser (Chrome, Firefox, Safari, Edge)
- A local HTTP server (ES Modules require serving over HTTP, not
file://)
Helios now expects the OpenRouter key to live on the server, not in browser code.
Create a local .env only for Vercel/local function runtime use:
cp .env.example .envRequired variables:
OPENROUTER_API_KEY=your_openrouter_key
OPENROUTER_MODEL=google/gemini-3-flash-previewDo not commit .env. The repo ignores it by default.
This project is a static dashboard plus one Vercel Function at api/openrouter.mjs, which securely proxies Gemini/OpenRouter requests.
- Import the repo into Vercel.
- In
Project Settings -> Environment Variables, add:OPENROUTER_API_KEYOPENROUTER_MODEL(optional, defaults togoogle/gemini-3-flash-preview)
- Redeploy the project.
If the dashboard still shows mock data after deploy:
- open
https://your-deployment.vercel.app/api/openrouterin the browser and confirm it responds with405 Method not allowed - if it returns a missing env error, the env var is not set in Vercel
- after adding or changing env vars, redeploy so the function picks them up
If you want the AI features locally, use Vercel's local runtime instead of a plain static server:
npx vercel devThat serves both the static files and the /api/openrouter function together.
Helios.ai/
├── index.html # Main SPA entry point
├── css/
│ └── styles.css # Complete design system (1400+ lines)
├── js/
│ ├── app.js # Orchestration — wires all modules together
│ ├── data.js # Mock data — 3 cities × 3 time horizons
│ ├── map.js # Leaflet map with markers & boundaries
│ ├── alerts.js # Alert card rendering & interactions
│ ├── modal.js # Alert detail modal with AI reasoning
│ ├── details.js # Location detail sidebar panel
│ └── timeline.js # Upcoming timeline component
├── assets/
│ └── logo.svg # Helios sun logo
├── Future_Implementations.md # Roadmap to production
└── README.md
This is a prototype with live weather data support and AI reasoning routed through a server-side proxy. Here's what works:
| Feature | Status |
|---|---|
| Light ops-dashboard theme | ✅ Working |
| Interactive map with terrain tiles | ✅ Working |
| City switching (3 cities) | ✅ Working |
| Time horizon toggle (Today/Tomorrow/3-Day) | ✅ Working |
| Alert cards with severity ranking | ✅ Working |
| Alert detail modal with AI reasoning | ✅ Working |
| 3-model consensus display | ✅ Working (mock) |
| Source verification links | ✅ Working (real URLs) |
| Loading states & animations | ✅ Working |
| Per-city analysis area boundaries | ✅ Working |
| Real-time weather API integration | ✅ Working |
| Gemini/OpenRouter server-side proxy | ✅ Working |
| Mobile app with push notifications | 🔜 Future |
Helios directly supports UN Sustainable Development Goal 13.1:
"Strengthen resilience and adaptive capacity to climate-related hazards and natural disasters in all countries."
By providing earlier, more actionable weather warnings to campus operations teams, Helios aims to:
- Reduce response times from hours to minutes
- Increase warning coverage across vulnerable populations
- Improve decision quality with AI-powered reasoning and multi-model consensus
- Build trust through transparent, source-verified predictions
See Future_Implementations.md for the full production roadmap, including:
- Multi-model AI prediction engine
- Real-time sensor integration (IoT weather stations, USGS stream gauges)
- Mobile app with location-aware push notifications
- Backend infrastructure (FastAPI, PostgreSQL + PostGIS)
- Campus pilot deployment plan
Built for hackathon demonstration purposes.
This project is open source under the MIT License.