Business-Aware Vector Search Ranking

Milvus 2.6.2 introduces Boost Ranker, a lightweight, rule-based reranking feature that adjusts vector search results using scalar metadata fields. This allows for business context to be incorporated into search rankings without rebuilding indexes or changing embedding models. It operates entirely within Milvus, applying filters and weights to candidate results post-vector search. The feature supports boosting important results, demoting less relevant ones, and adding controlled randomness for diversity. It can be combined with other rankers like RRF or model-based rankers, and is configured via the Function API.