AWS Bedrock - Knowledge Base - Demo

Quick Step by Step Instruction to Play with AWS Bedrock - Knowledge Base - Demo with Cricket Statistics

Nov 29, 2023

Enhanced Responses with RAG: Amazon Bedrock's Knowledge Bases empower Facility Managers (FMs) and agents by integrating Retrieval Augmented Generation (RAG), ensuring more relevant and accurate responses by leveraging private data sources.
End-to-End Workflow Support: The fully managed capability of Knowledge Bases for Amazon Bedrock covers the entire RAG workflow, simplifying processes from data ingestion to retrieval and prompt augmentation. No custom integrations are needed, and session context management supports seamless multi-turn conversations.
Effortless Data Connection: Securely connecting FMs and agents to data sources is streamlined. By pointing to the data's location in Amazon S3, the system automatically fetches, processes, and stores information in a vector database, creating a hassle-free experience.
Flexible Vector Store Options: Whether using Amazon OpenSearch Serverless, Pinecone, or Redis Enterprise Cloud, the system allows for easy specification of an existing vector store. If none exists, Amazon Bedrock creates an Amazon OpenSearch Serverless vector store for efficient data storage.
Transparent Information Retrieval: To boost transparency and credibility, all information retrieved from Knowledge Bases for Amazon Bedrock comes with citations. This ensures proper source attribution, contributing to trustworthy and reliable interactions.

Use Bedrock - Claude2 Model without Knowledge Base

You won’t see the results in text generation due to "knowledge cutoff". It refers to the point in time at which the information used to train a machine learning model ends.

Prompt

Who won the Match Against India vs Pakistan in Worldcup 2023 - Also give a Quick Summary

Enable Knowledge Base

Go to —> AWS Bedrock - Orchestration —> Knowledge

Create Knowledge Base

I used below dataset from 2023 Cricket Worldcup till Semifinal

https://github.com/jayyanar/serverless-rag-demo/blob/main/worldcup-upto-semifinal.txt

Verify the Index in Opensearch Serveless —> Vector Store —> Index

Now you can “Test Knowledge” from “Knowledge Base”

You can find the Output Retrieved from Knowledge Base and also Source of Text Completion