One of the biggest challenges in Graph Retrieval Augmented Generation (RAG) systems is how to communicate knowledge from the ontology (e.g., subclasses, types, property values) to the LLM. The approach described in this post is very simple and easy to reuse but surprisingly effective.