AirGuardian AI

Inspiration

AirGuardian AI Delhi Air Quality & Public Health Companion AirGuardian AI is an intelligent public health assistant that turns real-time and historical air quality data from Delhi into actionable, personalized health support. It uses Elasticsearch for hybrid search, geospatial queries, and semantic RAG, combined with an AI agent mesh (LangGraph) and a machine learning AQI predictor (XGBoost).

Key Features AQI & Weather Dashboard: Live views of pollutant metrics (PM2.5, PM10, O 3 , N O 2 , etc.) and current weather metrics. Personalized Health Risk Assessment: Calculates user-specific health risk scores (Low, Medium, High, Critical) based on age, pregnancy, asthma/COPD, and activity levels. Actionable Recommendations: Clear, contextual health advisories regarding masks, air purifiers, outdoor activities, commutes, and hydration. Agentic RAG Chatbot: Multi-agent chatbot powered by LangGraph, pulling clinical and governmental insights dynamically from WHO guidelines and Delhi's Graded Response Action Plan (GRAP) indexed in Elasticsearch. Interactive Health Map: Displays nearby hospitals, clinics, and pharmacies, complete with geo-spatial filtering (e.g., within 5 km of user's coordinates). XGBoost AQI Prediction: Predicts air quality forecasts for the next 6, 12, and 24 hours. Tech Stack Frontend: Next.js 15, React, TypeScript, TailwindCSS, ShadCN UI, Recharts, React Leaflet, React Query Backend: Python 3.10+, FastAPI, Pydantic, Uvicorn, LangGraph, LangChain Database & Search: Elasticsearch Cloud (Vector DB, Geospatial, Hybrid search) ML Framework: XGBoost / Scikit-learn AI Models: Gemini 2.5 Flash / text-embedding-004 Project Structure airguardian-ai/ ├── README.md ├── .gitignore ├── docs/ # Detailed technical documentation ├── frontend/ # Next.js 15 App Router frontend └── backend/ # Python FastAPI backend & agents Getting Started Prerequisites Python 3.10 or higher Node.js 18 or higher Elasticsearch Cloud instance or local installation Setup Instructions For details on setting up both the backend and frontend, please refer to:

setup.md architecture.md

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for AirGuardian AI

Built With

• 2.5
• ai
• fastapi
• flash
• framework:
• frontend:-next.js-15
• gemini
• geospatial
• hybrid
• langchain-database-&-search:-elasticsearch-cloud-(vector-db
• langgraph
• ml
• models:
• pydantic
• react
• react-leaflet
• react-query-backend:-python-3.10+
• recharts
• scikit-learn
• search)
• shadcn-ui
• tailwindcss
• typescript
• uvicorn
• xgboost