Driftoracle

Reach for this when you need to monitor production ML models for data and prediction drift. You get 15 tools covering the essentials: PSI (Population Stability Index) calculations, Kolmogorov-Smirnov tests for distribution shifts, configurable alerting when drift crosses thresholds, and evidence pack generation for incident reports. It's designed for the operational side of ML, giving you programmatic access to the statistical tests you'd otherwise run manually in notebooks. The streamable HTTP transport means you can integrate drift checks into existing monitoring pipelines or trigger them from Claude conversations when investigating model degradation.