Production RAG

The knowledge assistant from Chapter 1 goes live on a Monday. It serves an illustrative 5,000-employee company from 150,000 documents of HR policy, IT runbooks, security standards, product documentation and support macros. Traffic is about 12,000 questions a day, with a peak of six questions a second between nine and ten in the morning. The service-level objective is that 95% of answers start streaming within 1.5 seconds.