How to Use Milvus Boost Ranker for Business-Aware Vector Search

This post introduces Milvus's Boost Ranker feature (2.6.2), a new capability for adjusting vector search result rankings based on scalar metadata. It details the internal mechanism of filtering and weighting candidate results after initial vector search, and provides use cases for prioritizing business-critical information (e.g., in-stock products, official documents) and demoting less relevant content. The post also explains how to configure Boost Ranker using the Function API, including parameters for weight, filtering, and optional random score for diversity. It demonstrates implementation with a Python code example for prioritizing official documents and compares results with and without the feature.