【全新上线】使用 Bedrock AgentCore Evaluations 提升生产环境中的代理质量.pdf

1、 2024,Amazon Web Services,Inc.or its affiliates.All rights reserved.2024,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Vivek Singh(he/him)Senior Product Manager-TechnicalAmazon Web ServicesIshan Singh(he/him)Senior Ge

2、nAI Data ScientistAmazon Web ServicesAmanda Lester(she/her)Senior GenAI GTM SpecialistAmazon Web ServicesImprove agent quality in production with Bedrock AgentCore EvaluationsA I M 3 3 4 8 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.1.Overview of Amazon Bedrock AgentCore2.Key

3、challenges with operating&evaluating agents 3.Introducing AgentCore Evaluations4.Demo 5.Best practicesAgenda 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Agentic AIThe future is now 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Amazon Bedrock AgentCoreThe m

4、ost advanced agentic platform:Everything you need for getting agents into productionRuntimeMemoryIdentityGatewayCode InterpreterBrowser PolicyEvaluationsObservability 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Foundational services for running agents,securely at scaleOPEN SOU

5、RCEProtocols:MCP,A2AFrameworks:Strands Agents,CrewAI,Claude Agent SDK,LangGraph,LlamaIndex,OpenAI Agents SDK Enhance with tools&memoryMemoryBrowserCode interpreterGatewayDeploy securely at scaleRuntimeIdentityPolicyObservabilityEvaluationsGain insights into agentic operations 2025,Amazon Web Service

6、s,Inc.or its affiliates.All rights reserved.Execute tasks and jobs without direct supervisionAgents can reason and act autonomously Reason Solve Problems Make DecisionsAgent 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.SupervisorEmployeesEmployeesEmployeesSalesMarketingServiceT