Vector Database

A store that finds text similar to a question; the usual engine behind RAG.

Definition

A vector database stores text as numbers so an AI can find content that looks similar to a question. It's the usual engine behind RAG. It finds what looks alike but doesn't know relationships like eligibility, credit or substitutes, which is what a knowledge graph adds.

Key points

<ul><li>Example: a search for "something for a sore throat" can find "throat lozenges", even though the wording is different.</li><li>It works well for searching large amounts of text, such as manuals, FAQs and policies.</li><li>It's weaker on exact details. "16 amp" and "6 amp" look similar as text but are different products, so commerce agents also need exact lookups.</li></ul>

‍