About Volt-AI
Turns Colombia's open electricity data into forecasts, explanations and answers, with every forecast checked in public.
What it is
Volt-AI reads the public data of the national power market every day, trains forecasting models, checks them honestly against the past and publishes the results here. An AI analyst answers questions using only validated lookups.
How it works
Daily pipeline
Python on GitHub Actions: fetch, validate, train, forecast, brief.
Database
PostgreSQL stores series, forecasts, anomalies and briefings.
This site
Next.js reads the database with a read-only role and renders cached pages.
Data sources
- XM: demand, spot price, dispatch cost, contracts, generation, reservoirs and inflows (public API)
- Open-Meteo: temperature and rainfall in four cities
- NOAA Climate Prediction Center: Oceanic Niño Index
What was taken care of
- Days with incomplete data are held back, never shown as if they were final.
- Forecasts are tested only on days they could not have seen.
- Every chart has a text summary and a table.
- The site works in Spanish and English and without JavaScript for reading.
Built with
- Python 3.12
- pandas
- pandera
- LightGBM
- statsmodels
- SHAP
- PostgreSQL
- Next.js 16
- TypeScript
- Tailwind CSS 4
- Apache ECharts
- Vercel AI SDK
- Gemini
- GitHub Actions
Author
Built by Miguel Angel, data and AI developer.
Source code