Retrieval Augmented Generation (RAG) is starting to be used in professional settings like the legal domain. In classic information retrieval (IR) systems, the effectiveness of the system is often evaluated using measures that rely on relevance judgements. Van Opijnen and Santos [1] have created a framework of relevance for information retrieval systems for legal professionals, to ground these evaluation methods. This paper takes their framework of relevance, and extends it for use with RAG applications. The aim of this framework is to make visible how legal RAG differs from other RAG applications, and how that might impact evaluation.