RAG in Microsoft Agent Framework – Overview

This article provides a technical overview of how Microsoft Agent Framework implements Retrieval-Augmented Generation (RAG). It explains the core components: VectorStore, TextSearchStore, TextSearchProvider, and the Semantic Kernel bridge. It details two retrieval modes: BeforeAIInvoke (automatic injection) and OnDemandFunctionCalling (agent calls search as a tool). It offers recommended defaults for top_k, chunk size, and overlap, and discusses production concerns like chunking, metadata, latency, cost, freshness, security, and telemetry. The article is aimed at developers looking to implement RAG in production assistants using Microsoft's framework.