{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/018c20b00b84470da28c89616f870a76\" frameborder=\"0\" width=\"1662\" height=\"1246\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":1246,"width":1662,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":1246,"thumbnail_width":1662,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/018c20b00b84470da28c89616f870a76-08ca5df02c2e6377.gif","duration":340.148,"title":"Adjacent Demo: Cold-Start Recommendations via Lazy Graph Construction","description":"In this demo, I walk through how Adjacent operates using Neo4j for visualization, the Adjacent API for triggering recommendations, and Grafana for monitoring latency and inference over time. I demonstrate the evolution of product relationships starting from a cold state, where querying a product leads to the formation of a local subgraph and the inference of relationships, ultimately improving system performance. Initially, we see a latency of 733 ms drop to around 66 ms as the graph becomes denser and we rely less on vector search. I encourage you to explore the repository for deeper insights into the design and assumptions behind the system. Your feedback and questions are welcome as we continue to refine this approach."}