ReliefLink AI is an AI-powered emergency management and response platform designed to orchestrate coordination during natural disasters (floods, earthquakes, cyclones). Built using a multi-agent framework powered by Google Gemini API, it bridges the communication gap between citizens in distress, local volunteer networks, relief shelter managers, and rescue authorities.
During crises, communication channels become overloaded and information is scattered. Citizens don't know where to go, authorities struggle to prioritize calls, and volunteers lack navigation routes.
ReliefLink AI deploys 7 Specialized AI Agents operating inside a "Coordination Chamber" to ingest SOS details, prioritize incidents, provide instant medical first-aid guidance, map safe routes, allocate shelter resources, and dispatch volunteers—all in real-time.
When a citizen broadcasts an SOS trigger, the backend launches a coordinated thought cycle among these agents:
graph TD
SOS[SOS Distress Signal Received] --> CA[Citizen Agent: Extract details & urgency]
CA --> RA[Rescue Agent: Triages danger & sets priority Critical/High]
RA --> MA[Medical Agent: Compiles immediate First-Aid advisory]
MA --> NA[Navigation Agent: Maps safe paths avoiding waterlogged roads]
NA --> ResA[Resource Agent: Allocates nearest shelter & reserves capacity]
ResA --> VA[Volunteer Agent: Dispatches closest available volunteer]
VA --> CommA[Communication Agent: Translates alert to Hindi/English for the citizen]
- Citizen Agent: Ingests the raw SOS (text/speech), extracts metadata, and acknowledges receipt.
- Rescue Agent: Assesses danger level and sets urgency status.
- Medical Agent: Analyzes the hazard and injects emergency first-aid protocols.
- Navigation Agent: Calculates safe routes to shelters, avoiding local hazards (e.g. waterlogged subways in Mumbai).
- Resource Agent: Checks shelters' occupancy and locks resource inventory.
- Volunteer Agent: Matches and dispatches the closest active volunteer using GPS distance calculations.
- Communication Agent: Translates the coordinated instructions into local languages (English & Hindi) for the citizen, and sends a technical report to the authorities.
- Multi-Agent Coordination Chamber: Stream logs showing real-time agent reasoning processes.
- Interactive Live Operations Map: Displays active SOS signals (🔴), safe shelters (🔵), and active volunteer responders (🟢) via Leaflet.js (OpenStreetMap).
- Citizen SOS Console: Broadcast SOS, select category, map pin-pointing, and voice simulation.
- Volunteer Duty Hub: Self-registration console, coordinate tracker, active dispatch route view, and status updates.
- Authority Dashboard: Statistics overview cards, shelter occupancy metrics, and incident resolution control.
- AI Reasoning Engine: Gemini API (
gemini-2.5-flash) - Backend Framework: FastAPI (Python 3.13)
- ORM & Database: SQLAlchemy & SQLite (local file)
- Frontend Web App: React 19, Vite, Lucide React (Icons)
- Mapping Library: Leaflet.js (React Leaflet)
- Styling: Custom Vanilla CSS (Glassmorphism & Dark Mode)
- Python 3.10+
- Node.js v18+ & npm
Navigate to the backend directory, set up your Python environment, and start the FastAPI application:
cd backend
# Create virtual environment
python -m venv .venv
# Activate virtual environment (Windows)
.venv\Scripts\activate
# Install dependencies
pip install -r requirements.txt
# Start backend (Runs on http://localhost:8000)
python main.pyNote: Create a .env file in the backend/ directory and configure your GEMINI_API_KEY to enable real Gemini agent responses. The system will fall back to smart local mock agent simulation if no key is supplied.
Navigate to the frontend directory and start the Vite React dev server:
cd frontend
# Install packages
npm install
# Start frontend (Runs on http://localhost:5173/)
npm run devOpen your browser and navigate to http://localhost:5173/ to access the ReliefLink AI Platform.