A research framework that trains and evaluates LLMs for accurate API and function calling, with models, benchmarks, and datasets.