streamcal keeps a deployed model's binary probabilities calibrated as outcomes arrive, without retaining its prediction history. It fits a monotone (isotonic) correction from fixed-size per-bin sufficient statistics, can down-weight stale history, and predicts before observing each outcome so a forecast never uses its own label.
Cite: Gaurav Sood. "streamcal: Streaming Probability Calibration." https://gsood.com/research/streamcal-streaming-probability-calibration/
BibTeX:
@software{soodstreamcal,
author = {Gaurav Sood},
title = {{streamcal: Streaming Probability Calibration}},
url = {https://github.com/finite-sample/streamcal}
}