I am an archivist in Stanford Libraries. Currently, I’m investigating using Neo4j LLM Knowledge Graph Builder to extract entities and relationships in archival collections. My hope is to have a system which allows users to define nodes and relationships to be extracted in collections. Users can then browse collections by extracted entities and relationships. And also improve vector search results with the knowledge graph generated. The target materials are primary source materials like unpublished papers and book drafts, correspondence (including letters and emails), oral history, photos, audio and video recordings, and more. Due to copyrights, privacy concerns, and other factors, these materials are only available in our reading room, so the system needs to be deployed locally.