Vector Databases Explained

Vector databases store embeddings and make similarity search fast. In AI Engineering, they are often used to retrieve relevant context for RAG systems.

Embeddings

An embedding is a numeric representation of text, image, audio, or another input. Items with related meaning tend to sit near each other in vector space.

type EmbeddedChunk = {
  id: string;
  text: string;
  embedding: number[];
  metadata: {
    source: string;
    tenantId: string;
    updatedAt: string;
  };
};

Similarity Search

When a user asks a question, the system embeds the query and searches for nearby vectors. The result is a ranked candidate list.

Metadata Filters

Metadata filters keep retrieval relevant and safe. Use them for tenant boundaries, permissions, source type, date ranges, product areas, and language.

Hybrid Search

Hybrid search combines semantic vector search with keyword search. This is useful when exact strings matter, such as product SKUs, API names, legal terms, or error codes.

Next Step

Take the vector database quiz, then inspect how your source documents should be chunked and filtered.