Agentic RAG Stack (LlamaIndex, ChromaDB, LangChain, FastAPI, React) overview

A retrieval-augmented generation architecture where an AI agent dynamically decides what to retrieve, how to chunk and re-rank results, and when to perform multi-hop retrieval across heterogeneous data sources. LlamaIndex provides the data connectors, indexing pipelines, and query engines. ChromaDB (or Qdrant) serves as the vector store for embedding-based similarity search. LangChain handles prompt orchestration, tool integration, and output parsing. FastAPI exposes the RAG pipeline as an async API with streaming support. React powers the chat frontend with real-time token streaming. This stack is best for enterprise knowledge bases, legal document QA, and customer support copilots where static retrieval falls short and the agent must reason over retrieval strategy. The tradeoff is latency — agentic retrieval adds LLM calls per query compared to single-shot RAG.