Naweather

• 

Inspiration

In many parts of Nigeria and other developing regions, weather forecasts are either inaccurate, too broad, or not localized enough to help people make real-time decisions. Farmers, commuters, market traders, and everyday citizens are often caught off-guard by rain or extreme heat due to lack of timely, location-specific data. We wanted to solve this by building a smart assistant that combines weather APIs, AI, and crowdsourced data to give users reliable, hyperlocal forecasts.

What it does

Naweather is a mobile-first weather assistant that allows users to:

• Check real-time weather by town, street, or LGA

• Ask weather-related questions in natural language (“Will it rain in Ariaria at 3 PM?”)

• Receive smart, AI-generated advice for activities like commuting, drying clothes, farming, etc.

• View a live weather map showing rainfall, clouds, and temperature zones

• Submit local weather reports, improving crowd accuracy

• Get location-based weather alerts via push or SMS

  How we built it

  Frontend: React + Tailwind CSS + Mapbox for map visualizations

Backend: Node.js with Express (API layer)

Database: Supabase (PostgreSQL + Realtime)

AI Layer:

GPT-4 (via OpenAI) to process queries and summarize weather trends

LangChain to orchestrate prompt templates and user memory

Weather APIs: OpenMeteo

Deployment: Netlify

Challenges we ran into

Combining crowdsourced, unstructured reports with structured API weather data

Handling multiple Nigerian dialects and phrases in the natural language layer

Rate-limiting and latency with GPT-4 responses

Designing a clean map UI that shows weather icons clearly on mobile screens

Generating accurate micro-forecasts with limited historical data

Accomplishments that we're proud of

Successfully integrated live user weather reports with location filtering

Created a hyperlocal alerting system that notifies users before the weather changes

Designed a map UI that reflects real-time weather in Nigerian towns

Developed a reliable fallback system when API data is missing — using GPT summarization

What we learned

How to combine traditional API data with user-generated reports

The importance of geospatial awareness in weather applications

Prompt engineering and memory context for weather-specific use cases

How to deploy end-to-end AI features on a full-stack app under time pressure

What's next for Naweather

Expand coverage across all 36 Nigerian states and rural regions

Add voice support for hands-free weather queries (with Whisper + GPT-4o)

Train a custom AI model for 1–3 hour rainfall prediction using local weather history

Partner with schools, farms, and market cooperatives to deploy in offline-friendly modes

Build a Weather for Business API for logistics, agrotech, and event planners

Built With

• javascript
• next.js
• openmeteo
• react
• supabase