生成式人工智能、智能体、MCP 以及人工智能驱动软件开发的未来.pdf

《生成式人工智能、智能体、MCP 以及人工智能驱动软件开发的未来.pdf》由会员分享，可在线阅读，更多相关《生成式人工智能、智能体、MCP 以及人工智能驱动软件开发的未来.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.D V T 2 1 7Generative AI Agents,MCP,and the future of AI-powered software developmentGeetha RamachandranEnterprise Productivity Engineering,FINRAPavitra KrishnanDeve

2、loper Agents and Experiences,AWSDerek ChilesDeveloper Agents and Experiences,AWS 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AgendaIntroductionAgent CompositionKiro and Custom AgentsNext Gen SDLC at FINRARecap 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.

3、2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Agent Composition 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.An LLM agent runs tools in a loop to achieve a goal.-Simon Willison 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.An LLM agent

4、runs tools in a loop,while building context,to achieve a goal.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.InputTool SelectionTool ExecutionReasoning(LLM)ResponseAn LLM agent runs tools in a loop,while building context,to achieve a goal.Context 2025,Amazon Web Services,Inc.or i

5、ts affiliates.All rights reserved.InputTool SelectionTool ExecutionReasoning(LLM)ResponseContextAn LLM agent runs tools in a loop,while building context,to achieve a goal.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.An LLMProvides semantic relationships encoded into model weigh

6、ts to map language to compressed representations in vector spaceProvides reasoningPerforms multi-step inference by autoregressively predicting the next token through learned attention patternsProvides knowledgeProvides semantic relationships encoded into model weights to map language to compressed r