Vector Database

A vector database is a specialized database designed to store, index, and efficiently search high-dimensional vectors (embeddings). It's the search engine behind RAG systems, enabling AI agents to find semantically relevant information from their knowledge bases in milliseconds.

How vector databases work: documents are converted to embedding vectors (arrays of numbers capturing semantic meaning), vectors are stored with metadata (source, timestamp, category), an index structure enables fast similarity search, and queries are also converted to vectors and compared against stored vectors using distance metrics (cosine similarity, euclidean distance).

Popular vector databases include: Pinecone (managed, easy to use), Weaviate (open-source, feature-rich), Qdrant (open-source, high performance), Chroma (lightweight, developer-friendly), pgvector (PostgreSQL extension — add vectors to existing databases), and Milvus (enterprise-scale, open-source).

Key features to evaluate: query speed (how fast can it search?), scale (how many vectors can it handle?), filtering (can you combine vector search with metadata filters?), updates (how easy is it to add/remove/update vectors?), and managed vs. self-hosted (operational complexity).

For AI agent builders, the vector database is a critical but usually invisible component. When a user asks a question, the knowledge base search happens through the vector database — finding the most relevant content chunks in milliseconds. Platforms like Chipp manage the vector database infrastructure, so builders simply upload content and the search works automatically.