Agentic retrieval with LangChain and Amazon Bedrock Knowledge Bases
Amazon Web Services Manideep Reddy Gillela
LangChain support for Amazon Bedrock Managed Knowledge Bases showed that single-shot retrieval can return topically relevant chunks that still omit evidence needed for multi-part comparative questions. With numberOfResults set to 5 for a question containing 6 sub-intents, it covered 4 of 6, while setting it to 10 covered all 6 but with duplicated and unused coverage. Using Bedrock’s agentic retrieval adds a planning loop that breaks the query into sub-queries, checks whether it has enough evidence, and runs additional searches when it does not.
Why it matters
Build a Retrieval Augmented Generation (RAG) application on Amazon Bedrock Managed Knowledge Base with LangChain, and see how agentic retrieval handles the multi-part questions that single-shot retrieval answers poorly. Run the same query through both paths, read the trace events, and compare what each retrieval path costs.