xanderteng/TangerangCast — reverse-engineered prompt

Build me TangerangCast, a Streamlit web app that helps people in Tangerang, Banten see where rain is likely now and in the near future. It should pull weather data from the Open Meteo API for a 20 by 20 grid covering Tangerang, save clean local CSV data for historic, current, future, and temporary records, and refresh the live data every 6 hours. Use a machine learning rain prediction pipeline with saved models, plus a simple way to retrain from the latest data later.

Make the app feel like a dashboard, with a home page, a live interactive map showing rain probability and risk zones, and a data insights page that explains model and dataset performance in plain language. Please include sensible error handling when the API or models are unavailable, add basic tests, and make it easy to run locally with Docker or by running the Streamlit app directly.