A2
Track A — Conversational & RAG-Powered Assistants

WeatherGPT — Conversational Weather & Local Alert Assistant

Agents that answer questions by retrieving the right information and responding conversationally, in the right language.

Background

Weather information is scattered across multiple portals, bulletins, satellite products, and forecast systems, making it hard for common users, researchers, and disaster managers to get quick, actionable insights.

Objective

Develop an AI-powered chatbot platform, WeatherGPT, that integrates meteorological data, forecasting models, and alert systems to deliver accurate, contextual, multilingual weather intelligence through natural conversation.

Key Features

1. Real-time weather information retrieval 2. Natural language querying for forecasts 3. Tool-selection logic (current vs forecast vs historical data source) 4. Extreme weather alerts and proactive warning dissemination 5. Location-based forecasting and advisory generation 6. Multilingual support for Indian languages 7. Climate trend and historical weather analysis 8. Voice-enabled interaction for rural accessibility

Expected Solution

A conversational AI platform (web or mobile) with tool-calling Backend integration with a weather API/data source An LLM-based query-understanding and tool-selection engine A scheduled alert mechanism for saved locations

Suggested Technology Stack

Backend: Python / FastAPI / Node.js Agent framework: LangChain / LangGraph (tool-calling) Weather data: OpenWeatherMap API / WeatherAPI.com (free tier) Notifications: Twilio / WhatsApp Cloud API Database: PostgreSQL / MongoDB

Expected Outcomes

Faster dissemination of weather information Improved public accessibility to forecasts Better disaster preparedness and response at the individual level

Possible Use Cases

Farmers seeking crop-weather advisories Commuters checking hyperlocal conditions Event planners checking multi-day forecasts Basic flood/heavy-rain alert dissemination

Evaluation Parameters

Accuracy and relevance of weather data Quality of tool selection (does it pick the right data source?) Response latency Multilingual capability Working alert mechanism