Vector search provides powerful semantic matching capabilities. However, relying solely on vector similarity may not always yield the most relevant results, particularly when exact term matches or specific keyword constraints are important.

This chapter focuses on hybrid search approaches, which mitigate these limitations by combining vector search with established methods like keyword-based retrieval using algorithms such as BM25. We will examine strategies for implementing these combined systems, including techniques for merging results from different sources (result fusion) and effective re-ranking methods. You will gain practical knowledge on building and tuning hybrid systems to enhance search relevance for demanding applications.