InvokeAgent streamFinalResponse unable to stream

Question

I have a RAG application that uses Bedrock knowledge bases + an agent which is triggered through InvokeAgent within a Lambda. My agent also uses a custom orchestration Lambda to define the agent's workflow.

My issue is that I can't get responses to stream. I've checked the required IAM permissions and ensured the Lambda is set up for streaming. Is this a Bedrock functionality issue? Even within the test area for the agent it gives buffered responses.

Here is a list of the things I have set up to enable streaming:
* Set streamFinalResponse to true
* Given agent + lambda IAM permissions to InvokeModelWithResponseStream
* Wrapped lambda handler in streamifyResponse and implemented using responseStream
* Created a functional URL for the lambda with streaming enabled

Even with all of these enabled it doesn't yield any results, and the final response seems to be outputting in one chunk.

re:Post Agent · Answer

You're facing a common issue with streaming responses from Bedrock Agents. Let me clarify a few points about the current functionality:

First, to use streaming with Bedrock Agents, you need to use the `streamingConfigurations` parameter in the `invoke_agent` call, not just setting `streamFinalResponse` to true. This parameter is available in recent versions of boto3 (check that you're using at least version 1.35.99).

For the IAM permissions, you're on the right track - the agent execution role does need the `bedrock:InvokeModelWithResponseStream` permission for the configured agent model.

When using guardrails with streaming, you can also configure the `applyGuardrailInterval` in the `StreamingConfigurations` to control how often guardrail checks are applied to outgoing response characters.

However, there's an important limitation to be aware of: the InvokeAgent API has internal throttling limits that aren't exposed in the Service Quotas console. Even at low request rates, you might hit per-second or burst limits depending on your agent configuration.

For your Lambda setup, ensure you've also added a resource-based policy statement on the Lambda function that allows Bedrock to invoke it:
1. The principal should be `bedrock.amazonaws.com`
2. The source ARN should be your Bedrock Agent's ARN
3. The action should be `lambda:InvokeFunction`

If you're still experiencing issues after implementing these configurations, consider implementing retries with exponential backoff and jitter to handle potential throttling exceptions.  
**Sources**  
[able to use streamingConfigurations in AgentsforBedrockRuntime invoke_agent boto3 sdk | AWS re:Post](https://repost.aws/questions/QUuzL1XVQ0TVyp5dEPZn5BZw/able-to-use-streamingconfigurations-in-agentsforbedrockruntime-invoke-agent-boto3-sdk)  
[InvokeAgent - Amazon Bedrock](https://docs.aws.amazon.com/bedrock/latest/APIReference/API_agent-runtime_InvokeAgent.html)  
[Bedrock Agent InvokeAgent API throttling despite Service Quota increase | AWS re:Post](https://repost.aws/questions/QUGXuP2lgBSju-hVhWCadsig/bedrock-agent-invokeagent-api-throttling-despite-service-quota-increase)  
[Error when Bedrock agent invoke lambda | AWS re:Post](https://repost.aws/questions/QUXk7QWdzGTh-c5MIWch9NNQ/error-when-bedrock-agent-invoke-lambda)  
[Invoke an agent from your application - Amazon Bedrock](https://docs.aws.amazon.com/bedrock/latest/userguide/agents-invoke-agent.html)

InvokeAgent streamFinalResponse unable to stream

Add your answer

Relevant content