Single-shot and agentic retrieval with Amazon Bedrock Managed Knowledge Base

Amazon Bedrock Managed Knowledge Base now supports single-shot RAG and retrieval-only requests through AgenticRetrieveStream, enabling applications to choose the right retrieval mode for each query.

• Single-shot RAG (maxAgentIteration=1): One retrieval plus one answer generation, no planning rounds
• Retrieval-only (maxAgentIteration=0): One retrieval with no LLM calls or answer generation
• Agentic retrieval (maxAgentIteration=2-5): Multiple planning rounds and searches for complex questions
• All modes support multiple knowledge bases and can be selected per-request without API changes
• Single-shot RAG reduces latency and model usage compared to agentic retrieval for straightforward questions
• Retrieval-only mode useful for employees who need source passages rather than generated answers

Applications can now use one API to handle both simple lookups and complex multi-part questions, balancing speed and answer quality without maintaining separate code paths.