Glossary · AI & ML

Vector Search over Schema

Vector search over schema is a technique that finds the database tables and columns most semantically relevant to a user's question by comparing embeddings, rather than matching exact names.

Users say "customer churn"; the database says "subscription_cancellations". Keyword matching fails here, but vector search succeeds: every table and column is converted into an embedding that captures its meaning, the question is embedded the same way, and the system retrieves the schema elements closest in meaning–regardless of naming. This is what lets an AI query tool handle vocabulary mismatch, the most common reason natural language queries fail on real databases. It is also how large schemas get filtered down to fit a model's context window.

Fig — every answer ships with the tables, rows and SQL behind it.