Toward Explainable and Automated Software License Checking via RAG and Knowledge Graphs

Abstract

Software development increasingly depends on large ecosystems of libraries, frameworks, and platforms, enabling faster delivery yet at the same time intensifying the challenge of license incompatibility. Such conflicts often remain undetected until late in the development lifecycle, where they are more costly and disruptive to resolve. Addressing such issues early is therefore essential for ensuring software robustness and long-term sustainability. To this end, we present LARK, a system that integrates Knowledge Graphs (KGs) with Large Language Models (LLMs) through Retrieval-Augmented Generation (RAG) to automate the detection of license incompatibilities while offering context-rich, citation-backed explanations. In an evaluation spanning 4,000 open-source and 100 proprietary Python projects, LARK achieves 98.1% license detection accuracy, a 96.2% compatibility F1 score, and improves explainability compared to existing methods. We anticipate that LARK will serve as a practical and scalable solution to mitigate legal risks and strengthen compliance in software development.