【全新上线】在亚马逊基岩知识库中构建多模式 RAG 体验.pdf

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.A I M 3 3 2 6-RBuild multimodal RAG experiences in Amazon Bedrock Knowledge BasesJean-Pierre DodelPrincipal Product ManagerAmazon Bedrock Knowledge BasesHammad Mirza

2、Senior Software Development EngineerAmazon Bedrock Knowledge Bases 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AgendaKnowledge ModalitiesCustomer ProblemBedrock Knowledge BasesBuilding RAG applications with Knowledge BasesHandling Multimodal ContentCreating Knowledge Bases tha

3、t support multimodalityDemoWhiteboarding and Q&A 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Amazon Confidential and Trademark.DataVideoImageDocumentTextAudioSpeechGenerative AI applicationsKnowledge ModalitiesIDC indicates 90%of enterprise data is unstructured and growingIDC

4、link4 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Amazon Confidential and Trademark.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Amazon Confidential and Trademark.5Customer ChallengesUnlocking value from their growing repositories of unstructured dataSear

5、ching and reasoning across different content typesImproving the accuracy and price-performance of search at scale 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Amazon Confidential and Trademark.2024,Amazon Web Services,Inc.or its affiliates.All rights reserved.Amazon Confidentia

6、l and Trademark.Vector search has transformed how we find and retrieve data6A M A Z O NB E D R O C KK N O W L E D G EB A S E SKnowledge bases for Amazon BedrockNATIVE SUPPORT FOR RETRIEVAL AUGMENTED GENERATION(RAG)7MODELAnthropicClaudeMetaLlamaAmazonNovaUSER QUERYAUGMENTED PROMPTANSWERVECTORSTORE321