From RAG to agents: Build AI pipelines inside Azure HorizonDB

By Abe Omorogbe, Navya Teja Gajula , Binnur Gorer, B Harsha Kashyap, Krishnakumar Ravi (KK) from Microsoft PostgreSQL AI team If you’ve ever shipped a RAG app, this will feel familiar. Your data lives in Postgres. But the pipeline that turns that data into vectors lives somewhere else, s pread across external service s , queue s , and retry logic. And when the embedding API hiccups mid-batch? That’s a 2 a.m. p roduction incident . You didn’t set out to build your own embedding service. You just wanted to search your documents. And RAG is only the beginning. The moment AI works on your data: ex