使用检索增强生成+(RAG)+改进生成式+AI+应用程序的响应.pdf

1、 2024,Amazon Web Services,Inc.或其附属公司。保留所有权利。2024,Amazon Web Services,Inc.或其附属公司。保留所有权利。?(RAG)(RAG)?AI AI?2024,Amazon Web Services,Inc.或其附属公司。保留所有权利。?(RAG)?RAG?Amazon Bedrock?Amazon Bedrock?2024,Amazon Web Services,Inc.或其附属公司。保留所有权利。?务?务?2024,Amazon Web Services,Inc.或其附属公司。保留所有权利。?(FM)(FM)?(RAG)?FM?练

2、?2024,Amazon Web Services,Inc.或其附属公司。保留所有权利。?VS VS?RAG?FAQ?API?Amazon Bedrock?Amazon Bedrock?Amazon Bedrock-?Amazon Bedrock-FM 2024,Amazon Web Services,Inc.或其附属公司。保留所有权利。?务?聊?务?2024,Amazon Web Services,Inc.或其附属公司。保留所有权利。RAG RAG?研?研?研?API?务?研 2024,Amazon Web Services,Inc.或其附属公司。保留所有权利。?API 务?All_orde

3、rs 务?order=?XYZ?Embedding?-0.020.89-0.38-0.530.950.17 2024,Amazon Web Services,Inc.或其附属公司。保留所有权利。Embedding Embedding?Embedding?Embedding?Embedding?Embedding?0.027-0.011-0.0230.025-0.009-0.025?Embedding?-0.0110.0210.013?-0.0090.0190.015?2024,Amazon Web Services,Inc.或其附属公司。保留所有权利。?Embedding Embedding?

4、RAG RAG?Embedding?(LLM)?(RAG)?务?2024,Amazon Web Services,Inc.或其附属公司。保留所有权利。Titan Text EmbeddingsTitan Text Embeddings V2V2?Token?8,000?256?512?1024?(?务?)?ID?amazon.titan-embed-g1-text-02?Embedding?Embedding?Amazon Titan Text Embeddings V2?RAG?务?100?2024,Amazon Web Services,Inc.或其附属公司。保留所有权利。?Embeddi

5、ng?Embedding?Embedding?-0.020.89-0.38-0.530.950.17?RAG RAG?2024,Amazon Web Services,Inc.或其附属公司。保留所有权利。?RAG RAG?Embedding?的?2024,Amazon Web Services,Inc.或其附属公司。保留所有权利。?FM?务?(RAG)?FM?(Chat with your document)?模?RAG?Amazon Bedrock Amazon Bedrock?2024,Amazon Web Services,Inc.或其附属公司。保留所有权利。?Amazon bedroc

6、k?Embedding?模?研?研?(Amazon S3)?研?MetadataCloudFormation/CDK Support?300 Token?Embedding?Amazon TitanCohere Embed EnglishCohere Embed Multilingual?OpenSearch Serverless(public and private)Aurora PostgreSQL (pgvector)MongoDBPineconeRedis 2024,Amazon Web Services,Inc.或其附属公司。保留所有权利。模?模?Amazon bedrock?Emb