Practical Vector Search: Configuration, Tuning, RAG, and the Road Ahead

Vector search, sometimes known as semantic search, offers a huge opportunity to enhance content discoverability for your users while also giving you a key tool to feed info to chatbots and agents. Vector databases (DBs) and LLM-like embedding models power today’s implementations of semantic search as well as AI tooling for retrieval-augmented generation (RAG).

Thanks to the efforts of Drupal’s AI initiative and a growing stable of options for open-source vector DBs and open-weight embedding models—implementing this sophisticated technology is more accessible than ever. While you can call out to hosted APIs for the models or vector storage, it’s easier than ever to roll your own, or even run everything locally for testing and tinkering.

We’ll move beyond basic RAG demos to explore practical setup details like chunking strategies, field configuration, and hybrid filtering within the AI Search and Search API ecosystem. This includes identifying where the current stable of Drupal modules needs further community contribution.