Amazon Bedrock introduces single-shot and retrieval-only modes for knowledge bases
In short: Amazon has expanded the capabilities of the AgenticRetrieveStream API, allowing developers to specify the iteration level for knowledge base requests. You can now choose between zero, one, or multiple retrieval steps to balance speed, cost, and complexity per request.
This summary was generated automatically by AI from AWS Machine Learning's publication. It is our own text, not a copy of the original — facts, figures and quotes belong to the source, linked above and below.
What changed?
- 1New maxAgentIteration range: 0 to 5.
- 2maxAgentIteration=0: Returns passages only with zero LLM calls.
- 3maxAgentIteration=1: Enables single-shot RAG (one retrieval, one response generation).
- 4maxAgentIteration=2-5: Retains existing agentic planning behavior for complex queries.
| Parameter | Before | Now |
|---|---|---|
| maxAgentIteration range | 2 to 5 | 0 to 5 |
Why it matters
Previously, handling both simple factual lookups and complex multi-part questions required separate implementations. Now, you can use one API to optimize performance: low-latency single-shot for simple queries and multi-step agentic planning for complex reasoning.
What it means for AI agents and contact centers
When integrating RAG into voice agents, you can now optimize cost and latency by defaulting to single-shot retrieval for direct user intents. Use multi-step agentic retrieval only for complex, multi-part inquiries where accuracy is the primary constraint. Compare performance by testing how these modes affect the time-to-first-token in your voice pipelines.
🧪 Worth Testing
It simplifies the backend infrastructure for contact center agents by consolidating retrieval strategies into a single API endpoint.
Sources
- AWS Machine LearningOfficialPrimary sourceOriginal article →„Single-shot and agentic retrieval with Amazon Bedrock Managed Knowledge Base“9 Nov 2026, 22:13
- Published by source
- 9 Nov 2026, 22:13
- Found by our system
- 9 Nov 2026, 22:22
- Summary generated
- 9 Nov 2026, 22:23
This article was written by AI from the original source. Facts, numbers and prices come from the source; missing values are marked “Not specified”. Legal notice, copyright and privacy