Scikit-LLM wraps language models in the scikit-learn estimator API, so it drops into a Pipeline or a cross-validation loop natively.