《使用 Amazon Bedrock AgentCore 构建可扩展的 AI 代理.pdf》由会员分享,可在线阅读,更多相关《使用 Amazon Bedrock AgentCore 构建可扩展的 AI 代理.pdf(53页珍藏版)》请在三个皮匠报告上搜索。
1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.T N C 3 3 0Architecting Scalable AI Agents Using Amazon Bedrock AgentCoreMarilia BritoTechnical Training LeaderAmazon Web Services 2025,Amazon Web Services,Inc.or it
2、s affiliates.All rights reserved.AgendaUsing Amazon Bedrock AgentCore AgentCore Runtime AgentCore Identity AgentCore Gateway AgentCore Memory AgentCore Browser AgentCore Code Interpreter AgentCore Observability 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AI agents in a nutshel
3、lActionsEnvironmentObservationContextTools(and their descriptions)Goals and instructionsAgentLLM prompt 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Implementing an AI agent is complexOpen-source frameworks helps you build agents and create proof of concepts(PoCs)ContextTools a
4、nd tools executionMulti-agent collaborationOrchestrating between actionsStrands AgentsFunctionality available 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Getting agents to production is still too hardUndifferentiated heavy liftingSecurely execute and scale agent code.Discover
5、and connect with custom tools and resources.Understand and audit every interaction.Identity and access controls for all agents and tools.Remember past interactions and learning.Agentic tool use for running complex workflows.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Amazon Be
6、drock AgentCoreShort-term and long-term memoryEnd-to-end visibility to trace,debug and monitor agentsDeploy and operate highly capable agents securely,at scale,using any framework and model.Simplified tools discovery and integrationPurpose-built agent Runtime and IdentityFully managed browser tool a