Unveiling Milvus 2.4: Multi-vector Search, Sparse Vector, CAGRA Index, and More!

Milvus 2.4 introduces multi-vector search, enabling storage and querying of multiple vectors per entity and optimization of reranking algorithms. It adds Grouping Search for efficient large-scale grouping queries on scalar fields. Beta support for sparse vector embeddings is included, allowing hybrid search with models like SPLADEv2 and BM25, and supporting the Inner Product metric. The CAGRA index, a GPU-based graph indexing technology, is now supported for accelerated graph indexing and low-latency similarity search. Other enhancements include regex support for metadata filtering, a scalar inverted index, and a CDC tool.