Search by meaning rather than exact wording needs a different kind of database - that’s what a vector database is built for. It stores, indexes and searches high-dimensional numerical representations of data called embeddings; where a traditional database matches rows on exact values, this one measures the distance between vectors in multi-dimensional space to find records that are semantically similar to a query.
Example
Say a company has hundreds of support articles. Each one gets converted into a vector embedding by an AI model and stored in a vector database like Pinecone, Qdrant or Supabase Vector. Ask a question in plain language, and that question gets converted into a vector too - the database then returns whichever articles are closest in meaning, even if none of the words match exactly.
n8n relevance
In n8n, vector databases usually show up as part of a RAG pipeline. A workflow might generate embeddings for incoming documents with one node, store them in a connected vector database, then later query that database to pull relevant context before handing it to an AI agent or language model node. It’s a common pattern in AI-powered n8n workflows that need to answer questions from a private knowledge base instead of relying only on what the model already knows.
Related concepts
See also RAG (Retrieval-Augmented Generation), AI Agent and Prompt.