What Powers Similarity Search in Milvus Vector Database?

This post introduces Knowhere, Milvus's core vector computation engine. It details Knowhere's concept, its role within the Milvus architecture, and its advantages over Faiss, including support for BitsetView, more similarity metrics for binary vectors, and AVX512 instruction sets with automatic SIMD instruction selection. The post also breaks down the Knowhere code structure, explaining base classes and various index types (IDMAP, IVF, Annoy, HNSW), and provides a guide on how to add new indexes to Knowhere.