Python library hysteresis-state prevents false state changes in noisy classifiers

A developer has released hysteresis-state, a pure Python library with no dependencies, designed to eliminate rapid false transitions in discrete state classifiers. The library works by requiring a configurable number of consecutive observations before confirming a state change, resetting the count if the signal reverts before reaching the threshold. It also supports asymmetric confirmation thresholds, allowing systems to fall into a failure state quickly while recovering more cautiously — mirroring the behavior of an electrical circuit breaker. The tool originated from a trading bot whose market-regime classifier flickered near decision boundaries, causing unwanted operational freezes. The library is available via pip and its source code and mutation-tested test suite are published on GitHub.