This article delves into the transformative world of vector search technology, charting its evolution and widespread adoption across both new and established database platforms.

Manticore

Vector search has transformed database technology over the past few years, with both new vector-focused databases and established platforms adding vector capabilities. The technology uses mathematical vectors to represent data in multi-dimensional space, enabling similarity searches through methods like cosine similarity and euclidean distance. While sparse vectors (like TF-IDF) have been used for decades in traditional search, dense vectors from deep learning models capture richer semantic relationships but require specialized indexing like HNSW for performance. Modern implementations support K-nearest neighbor searches, hybrid approaches combining keyword and vector search, and increasingly offer automatic embedding generation to simplify operations.

Vector search in old and modern databases