Question

A development team is building a customer support chatbot using AWS Lambda that calls Amazon Bedrock. Requirements: the frontend must display tokens progressively as they are generated to reduce perceived latency, the code must remain portable across different Anthropic and Amazon Nova models without rewriting the request payload structure, and the team wants to minimize custom parsing logic for the streamed chunks. Which Bedrock API approach should the team use in the Lambda function?

Accepted Answer

Use the ConverseStream API, which provides a unified request/response structure across models and emits standardized streaming events.

Answer

Use the InvokeModel API with a model-specific JSON body and poll the response until the full completion is returned.

Answer

Use the InvokeModelWithResponseStream API with per-model body formatting and manually parse each provider's proprietary chunk schema.

Answer

Use the Converse API (non-streaming) and split the returned text into artificial chunks on the client side to simulate streaming.