在生成式人工智能应用中使用工具和代理.pdf

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Using Tools and Agents in Generative AI ApplicationsNavi KaurDelivery Readiness Tech LeadAmazon Web Services 2025,Amazon Web Services,Inc.or its affiliates.All right

2、s reserved.Who Am I Started my cloud journey in 2018Driving positive change by sharing expertise and enabling others to excel 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.When AI becomes a team playerLive console demo aheadIn this demonstration,the presenter shows you how agent

3、s and tools are working together.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AgendaFoundation conceptsFrameworksIntegration protocols 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Foundatio

4、n concepts 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.7The limits of standalone LLMs(1 of 4)Why you need more than just an LLMLLMs have vast knowledge,but its static,based on their training data cutoff 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.8The li

5、mits of standalone LLMs(2 of 4)Why you need more than just an LLMLLMs have vast knowledge,but its static,based on their training data cutoff.LLMs cannot perform real-world actions(run code,query a database,place an order).2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.9The limits

6、 of standalone LLMs(3 of 4)Why you need more than just an LLMLLMs have vast knowledge,but its static,based on their training data cutoff.LLMs lack context about your private,proprietary data.LLMs cannot perform real-world actions(run code,query a database,place an order).2025,Amazon Web Services,Inc