Adaptive Classifier

Guides you through adaptive-classifier, an open-source Python library for text classifiers built on pretrained language models that learn new examples and new classes without retraining from scratch. The workflow helps you decide whether the library fits your problem (small labeled sets, label sets that change, routing, intent and moderation use cases) or whether standard fine-tuning is the better choice for a large static dataset. It checks whether the library is installed and points you to the README install steps without running them, asks before anything downloads model weights or datasets, and walks through a quickstart, adding examples and new classes later, multi-label use, strategic (anti-gaming) prediction modes, and the standard versus ONNX inference paths. It explains how to evaluate honestly: a held-out split, per-class results, calibration checks, and a comparison against a trivial baseline. It covers saving and loading and states the limits plainly, including order dependence in online learning and the cost of wrong pooling or blend settings. This plugin contains instructions only; everything runs on your own machine.