Glossary
Vector Search
Vector search finds content by meaning rather than by matching words. Text is converted into numeric representations, and results are the passages whose representations sit closest to the query's.
Glossary
Vector search finds content by meaning rather than by matching words. Text is converted into numeric representations, and results are the passages whose representations sit closest to the query's.
Vector search compares meaning, so conceptually similar passages are found even when they share no vocabulary.

Search then becomes a proximity problem in that space.
Vector search is weak exactly where precision matters: product codes, names, part numbers and any query where the literal string is the point. Hybrid retrieval covers both cases.
Documents are split before embedding, and if a split lands mid-argument, the retrieved passage answers half a question. How content is divided usually affects answer quality more than which embedding model was chosen.
No. It handles meaning well and exact strings badly. Product codes, names and reference numbers are found more reliably by keyword matching, which is why hybrid retrieval is the usual production answer.
A numeric representation of a passage, produced by a model, positioned so that passages with similar meaning are close together. Comparing embeddings is what makes search by meaning possible.
Vector search finds content by meaning rather than by matching words. Text is converted into numeric representations, and results are the passages whose representations sit closest to the query's.