MariaDB VDB Provider Adds Native Vector Search to Drupal AI Without External Databases

A new Drupal module enables vector search directly within MariaDB, removing the need for external vector database services in AI-powered Drupal applications.

The MariaDB VDB Provider, developed by Juan Martinez, uses MariaDB’s VECTOR data type and indexing capabilities introduced in version 11.7. By storing embeddings in the same database as application data, it maintains transactional consistency while simplifying architecture compared to external vector database setups.

The module supports multiple distance metrics, including Euclidean distance and cosine similarity, and relies on MariaDB’s modified Hierarchical Navigable Small Worlds (HNSW) algorithm for efficient vector search. It integrates with the Drupal AI module and Search API, enabling use cases such as semantic search and retrieval-augmented generation workflows.

Configuration follows standard Drupal search workflows, including creating a Search API server with the AI Search backend, selecting MariaDB as the vector provider, and indexing content. The module requires MariaDB 11.7 or later (with 11.8 recommended), Drupal 10.2 or higher, PHP 8.1, and the PHP mysqli extension.

The module joins other Drupal AI vector database providers such as Pinecone, Milvus, and PostgreSQL-based options, but distinguishes itself by consolidating vector storage and application data within a single system. The project was created on 22 February 2026, with a stable 1.0.0 release published on 13 March 2026 under Drupal’s security advisory coverage.