"Turning chaos into coordination with AI-powered precision."
ReliefRoute is an AI-driven, offline-first disaster relief platform designed for coordinators operating in unstable network environments. It automates supply chain routing to prioritize the most critical, high-density, and low-accessibility zones following a disaster.
This project is built as part of the Solution Challenge 2026 – Build with AI, organized on the Hack2Skill platform.
The challenge focuses on leveraging Artificial Intelligence and Google technologies to solve real-world problems and create impactful, scalable solutions.
"Where problems aren’t missing — just waiting to be solved."
- Renu Kumari Prajapati
- Muskan Yadav
- Arushi Thalur
During disasters, relief operations face critical challenges:
- ❌ Lack of internet connectivity in affected areas
- ❌ Inefficient and manual resource allocation
- ❌ Poor communication between coordinators
- ❌ Unequal distribution of resources (remote areas ignored)
ReliefRoute is an AI-powered, offline-first disaster relief platform designed to:-
- 📍 Identify high-priority zones
- 🚚 Optimize resource distribution
- 📡 Enable communication without internet
- ⚖️ Ensure fair and unbiased allocation
It empowers coordinators to make fast, data-driven decisions even in extreme conditions.
- Offline-First Architecture: Continuous operation without internet. IndexedDB queues actions, caches the last-known state, and automatically synchronizes when connection is restored.
- Mesh Communication: A 3-mode relay system (Internet -> Local Mesh WebSocket -> SMS Fallback) ensures field coordinators can always send and receive critical updates.
- AI Allocation Engine: Automates the distribution of food, medicine, and shelter using Haversine distance, severity floors, and population density metrics.
- Conversational Reallocation: Ask Claude to intelligently shift resources via natural language when ground conditions change.
- Bilingual Interface: Full support for English and Hindi (Devanagari) with instant, zero-reload switching.
- Equity Auditing: Proactively detects and flags systematic bias against remote zones, ensuring minimum resource thresholds are always met.
- Frontend: React 18, Vite, TypeScript, Tailwind CSS, Zustand, React-Leaflet, Workbox (Service Workers), IndexedDB
- Backend: Express, TypeScript, Better-SQLite3 (WAL Mode)
- AI Integration: Google Gemini API(Gemini 2.5 Flash)
- Node.js (v18 or higher)
- API Key from Google Gemini
# Install root dependencies (concurrently, etc.)
npm install
# Install server dependencies
cd server
npm install
# Install client dependencies
cd ../client
npm installCreate a .env file in the root directory and add your Anthropic API key:
ANTHROPIC_API_KEY=sk-ant-api03-xxxxxxxxx
PORT=3001
MESH_WS_PORT=8765
EQUITY_THRESHOLD=0.40Return to the root directory and start the entire stack:
npm run devThis command concurrently runs:
- The React Vite server (
http://localhost:5173) - The Express backend (
http://localhost:3001) - The WebSocket Mesh Relay (
ws://localhost:8765)
To reset the system state and load the initial disaster scenario (Uttarakhand Flood):
npm run seedserver/: Contains the Express API, the SQLite database logic (db.ts), the matching and scoring algorithms, the mesh WebSocket server, and the Anthropic integration service.client/: The React UI. Notable structures includeuseStore.tsfor Zustand global state,offlineQueue.tsfor IndexedDB caching, andmeshComm.tsfor the 3-mode communication logic.data.db: The local SQLite database, operating in WAL (Write-Ahead Logging) mode for durability.
MIT License © 2026 Error_404
Built for Solution Challenge 2026 – Build with AI Powered by innovation, AI, and teamwork 🚀