Introducing the Magnolia Vector DB for semantic search

Your corporate content hub is full of answers. The problem is finding them. An editor searches for "sneakers" and gets nothing back — even though the landing page for the new running shoe collection has been sitting in the system all along. When content discovery fails, it directly impacts your user engagement and conversion rates, revealing the three ways your website is losing customers.

Classical full-text search tools index only the specific words your team used during content creation. Anyone searching with different words — even with exactly the same meaning — frequently comes up empty.

Today, we're introducing the Magnolia Vector database (DB) integration, a semantic search capability built for enterprise scale within the Magnolia AI Accelerator. The Vector DB understands what your content actually means, not just what it says — so the people who manage your content, and the AI agents that assist them, can finally find the right content, no matter how the question is phrased. Its primary surface is Context Search, a semantic search capability available everywhere in Magnolia DXP: as an AI task, through the Find Bar, and via a documented REST API.

Traditional search engines store words. Vector databases map meanings.

At the heart of the Vector DB are context vectors (also called embeddings): numerical representations of text, calculated by a specialized class of AI model — an embedding model forming an advanced AI-backed search ranking system.

Each piece of content becomes a point in a high-dimensional space, where distance reflects semantic similarity. "Sneakers", "running shoes", and "trainers" all land in the same neighborhood while "lawn mower" sits far away.

When someone searches, the query goes through the same embedding model and the Vector DB runs a nearest-neighbor search in vector space: it finds the content whose meaning sits closest to the intent of the query. An editor assembling a seasonal campaign searches "outfit for a rainy commute" and finds the "waterproof city jackets" collection page, no shared keyword required.

And because real-world queries often mix concepts with exact tokens — product names, SKUs, brand terms — the Vector DB doesn't choose between the two worlds. Every search runs hybrid retrieval: semantic vector similarity and classical full-text ranking, fused into a single result list. You get the recall of meaning-based search without losing the precision of exact-match search.

The Magnolia Vector DB brings profound capabilities to your central content hub, spanning content discovery, strict governance, and automated workflows:

Smarter search inside Magnolia DXP. Context Search plugs into Magnolia's Find Bar (Periscope), so editors searching the admin interface find content by meaning. "summer campaign visuals" finds the right assets even when no one ever tagged them that way.

Find the best asset, automatically. Capabilities like Doc-to-Page use Context Search to pick the most fitting images and documents for newly generated pages; semantic matching replaces manual asset hunting. Read more in Doc-to-Page: from brief to on-brand page in minutes.

Ground your AI in your content. Context Search is the retrieval layer for Retrieval-Augmented Generation (RAG): The Magnolia AI agent uses it as a tool to answer questions from your project content instead of generic model knowledge, with sources they can cite, and only content the user is allowed to see, effectively making AI real for content teams.

Build your own experiences. The documented REST API lets you bring semantic retrieval into your own applications, chat-based authoring assistance, intelligent content recommendations, or internal knowledge tools, all backed by the same permission-aware index driving advanced ecommerce personalization.

The Vector DB is a foundational building block of Magnolia's AI strategy, anticipating that the future is agentic. This same retrieval layer powers the Magnolia Agent and will power upcoming agentic workflows, chat-based authoring assistance, and context-powered automation. On the roadmap: additional vector database providers, higher-quality semantic chunking, and cross-encoder reranking for even sharper relevance.

Semantic search shouldn't require a data-science team. With the Magnolia Vector DB, you connect an embedding model, choose what to index, and run docker compose up — Magnolia DXP takes care of the rest.

Button

Get in touch with our team