What is a vector database?
A vector database stores data as high-dimensional numeric vectors (embeddings) and finds items by similarity rather than exact match, enabling semantic search over text, images and other unstructured content. It underpins AI features like retrieval-augmented generation by fetching the most relevant context for a query.
Traditional databases match exact values; vector databases match *meaning*. Content is converted into embeddings, numeric vectors that capture semantic meaning, and the database finds the nearest vectors to a query, returning conceptually similar items even when the words differ.
Vector databases power:
- Semantic search results by meaning, not keywords.
- Retrieval-augmented generation (RAG) fetching relevant passages to ground an AI's answer.
- Recommendations and clustering grouping similar items.
They specialise in unstructured data, documents, PDFs, images, where exact-match querying falls short.
iDBQuery uses vector-style semantic retrieval where it fits, especially for documents. When you ingest PDFs or a folder of files (with OCR for scans), iDBQuery indexes their content so you can ask questions and it retrieves the relevant passages to answer, with citations back to the source document. For your structured databases and spreadsheets, iDBQuery instead writes precise SQL against the live data, exact, not approximate. The combination matters: semantic retrieval for unstructured content and exact SQL for structured data, unified in one plain-language experience, so every answer, whether from a document or a table, comes back cited to its source.
Updated 2026-06-22