Introduction

Among all index algorithms supported by Milvus, the graph-based index algorithm is Hierarchical Navigable Small World (HNSW), which can perform quick query and achieve a high recall rate, but consumes a large amount of memory resources. To extend the graph-based index algorithm and enhance the query efficiency while ensuring a high recall rate, Kunpeng BoostKit proposes the Kunpeng Blazing-fast embedding similarity search thruster (KBest) algorithm.

KBest is an efficient, Huawei-developed graph search algorithm. It optimizes the performance and precision of the nearest neighbor search by using methods such as quantization and vector instructions, delivering search capabilities equivalent to the open source Faiss HNSW algorithm.

This feature connects KBest to the open source Milvus database as a patch file to provide graph search functionality. For details, see Installation and Usage Description.