Vector databases and semantic search

Vector search stores numeric representations of text, images or other content, called embeddings, and finds items with similar meaning. It powers semantic search, recommendations and the retrieval step in RAG systems.

1. 1

   Embed

   Content is converted into vectors using an embedding model.

2. 2

   Store and index

   Vectors are stored in a dedicated vector database or an extension such as pgvector.

3. 3

   Query

   Questions are embedded the same way and matched to the closest content, often combined with keyword search.

• Semantic search across documents and products
• Recommendations and similar item suggestions
• Retrieval for AI assistants

We often start with vector support inside an existing database before adding a separate system, to keep the architecture simple.