课程: Responsible AI on AWS: Bedrock Guardrails, Amazon Q Security, and SageMaker Clarify

免费学习该课程！

今天就开通帐号，24,700 门业界名师课程任您挑！

Amazon Bedrock Guardrails: Handling edge cases

Amazon Bedrock Guardrails: Handling edge cases

课程: Responsible AI on AWS: Bedrock Guardrails, Amazon Q Security, and SageMaker Clarify

开始一个月试用 购买团队方案

Amazon Bedrock Guardrails: Handling edge cases

“

- [Instructor] This technical diagram demonstrates the Bedrock Guardrail Edge Case Handling architecture for content filtering, and also the validation process. Let's take a look at the input first. In this input processing, the requests first pass through this Input Validation Layer, and they contain Metadata, Content, and also the Guardrail configurations themself. This is an initial sanitation that applies to the standard validation rules. Next step here, we have the Edge Case Detection, and this is pattern-matching engine that identifies anomalous input, so you could look for things like malformed content structure, boundary conditions, input size extremes, character and coding issues, invalid metadata combinations, and then in terms of the decision logic, this is where the rules engine would evaluate detection results, and you would look at things like a critical detection would require immediate rejection. A moderate might need some kind of specialized handling, and a minor…

内容