AI Search retrieval quality

A built-in evaluation (Beta) that measures how relevant different retrieval strategies are on a managed Delta Sync AI Search index; it launched as Vector Search retrieval quality. It samples documents from the source table, has an LLM write natural-language and keyword queries, runs each one with ANN, hybrid and full-text retrieval with and without the reranker, and has an LLM judge score each result on a 4-point scale. A dashboard reports DCG@10 as the main metric next to NDCG, recall, precision, MRR and MAP, and keeps past runs for comparison. Any user with query access to the index can start a run from the index page.

Status and dates are extracted from public sources only. We don't see the private AE-shared roadmap. If something looks wrong, tell us.