Weather Oracle ~ LLM-Enhanced Forecasting App

Weather Oracle is a full-stack TypeScript application that combines real-time weather data with LLM-generated narrative output.

Users enter a city and receive both structured weather data (temperature, feels-like, humidity) and a generated “oracle” interpretation built from those conditions.

This project explores how deterministic data can be transformed into a more engaging, story-driven UX without losing its underlying structure.

Key features:

• Server-side API orchestration (no client-side secrets)
• City → coordinates → weather retrieval via Meteoblue
• Normalised weather data used to construct LLM prompts
• Gemini 2.5 generates contextual narrative output
• Single endpoint: GET /v1/oracle?city=…

Example output:
Instead of:
20°C, partly cloudy

The app returns:
The clouds drift like slow omens, and the day carries a warm promise.

GitHub: [weather-oracle](https://www.twine.net/signin