Invisible work is most of what a good DBA does, and the better it goes, the less anyone sees it. AI can now produce a polished report in seconds, so polish no longer proves anything. This post shows how to keep a real record, log the near misses and earn the trust clients are paying for.
Letting an LLM Query SQL Server Safely: A Read-Only Login and Guardrails
Let an LLM query SQL Server through a restricted login, approved views, enforced row filtering, execution limits, and query auditing.
VECTOR_SEARCH in SQL Server 2025: Finding Similar Rows With a Vector Index
VECTOR_SEARCH returns the rows whose vectors sit closest to a vector you give it, and a vector index makes that search approximate and fast. This post builds a six-row test with hand-made numbers on SQL Server 2025, with no AI model, and lists the limits the test showed.
AI_GENERATE_CHUNKS in SQL Server 2025: Splitting Long Text Into Pieces
AI_GENERATE_CHUNKS cuts one long piece of text into smaller pieces inside a query. This post tests the preview function on SQL Server 2025, shows what its columns hold and how overlap works, and explains how chunk size changes the number of pieces.