nixtla

Why we included this project

TimeGPT is a pre-trained time series transformer, and this repository is its official Python client. You give it a pandas DataFrame with timestamps and values, make one forecast call, and get predictions with confidence intervals back. There is no model training or feature engineering to set up, and the same API also covers anomaly detection, fine-tuning on your own data, multiple series in one call, exogenous variables like holidays or prices, and cross-validation. That makes it a fast path for teams that need solid retail, electricity, finance, or IoT forecasts without building a bespoke pipeline. If you already work with Spark, Dask, or Ray, the client scales across large datasets, and a Snowflake deployment option runs forecasts inside your warehouse. Worth a look for anyone who wants production forecasts in a few lines of code and the option to fine-tune later.