This study investigates the integration of XML-TEI encoded digital editions of heritage collections into a Retrieval-Augmented Generation (RAG) system. Using the collections of the Valentin Haüy Association, we demonstrate that leveraging hierarchical divisions, titles, and paragraphs improves chunking and information retrieval. Results show that the RAG delivers precise, coherent answers faithfully reflecting the source documents while enabling traceability through TEI metadata. This approach highlights the potential of structured digital editions as a dynamic memory for RAG systems, offering enhanced exploration and accessibility of complex archival corpora within the digital humanities.
