https://eventyay.com/e/4c0e0c27/session/9590

Write talk:

MariaDB Vector Search – The fastest vector search in relational databases
Talk 25 min
0

speaker
Daniel Black
Chief Innovation Officer, MariaDB Foundation
AI & Data Science
Room 3
14 Mar, 2025 2:00 PM (+07)

Vectors are essential for AI models to represent data semantics, making vector search a crucial feature for databases supporting AI applications. This talk will cover MariaDB Vector, its inner workings, use cases, and future plans.

MariaDB Vector adds a high-level interface for indexing in MariaDB Server, enabling custom indexing strategies. Vector search requires a special index.

MariaDB and many vector databases use the Hierarchical Navigable Small Worlds (HNSW) algorithm. We’ll explore HNSW, why it provides approximate results, and how tuning parameters affect performance.

We’ll also explain the difference between Generative and Embedding AI models and their use in building Retrieval Augmented Generation apps with MariaDB as a datastore.

Lastly, we’ll review the current landscape of vector databases, their pros and cons, helping you decide between a dedicated vector database or a traditional relational database with vector search support.