Glossary

Vector Store

A vector store holds embeddings and finds the ones closest to a query. It is the retrieval half of RAG, and Db2 for i has no vector store as of 2026.

A vector store holds embeddings, the numeric representations a model produces for a passage of text, an image or another piece of content, and answers nearest-neighbor queries over them so a system can find the handful of passages most relevant to a question. That retrieval step is the first half of retrieval-augmented generation. The language model only does the second half.

For an IBM i shop this term matters because of an absence. IBM's Db2 for Linux, UNIX and Windows added a native VECTOR data type in Db2 12.1.2 in June 2025, so embeddings can sit in a table beside business data and be searched with SQL distance functions. Db2 for i did not get the equivalent, and the July 2026 Technology Refreshes for IBM i 7.6 and 7.5 added no AI items at all. As of September 2026 there is no vector store in Db2 for i and IBM has not published a date for one.

That puts the index outside the database. IBM's stated direction is OpenSearch running on Power, with Db2 for i remaining the system of record and OpenSearch holding the embedded, searchable copy an AI agent queries. Two practical consequences follow. The vector store is a second system to run, secure and back up. And it holds a readable copy of everything you indexed into it, inheriting none of your Db2 for i object authority automatically.

Related Terms