Here is a 149-word summary of the blog post:

In late 2023, AWS launched Knowledge Bases for Amazon Bedrock, allowing secure connections between machine learning models and company data for Retrieval Augmented Generation (RAG). RAG incorporates custom data into pretrained models without exposing it publicly, improving AI responses using provided data unavailable otherwise. For example, giving financial records to generate reports. RAG allows leveraging large language models with private data and more personalized, contextually relevant responses.

The demo uses Bedrock’s Knowledge Bases to build a chat interface where users ask questions about documents. The knowledge base contains the 2023-2024 Finnish IT Service Industry Agreement. The AI generates responses using information extracted from documents. Built on API Gateway, Lambda, Bedrock, S3 and OpenSearch, a S3 bucket stores documents vectorized into a knowledge base. The frontend uses ReactJS and Cloudscape. Users' prompts go to a Lambda function calling Bedrock’s RetrieveAndGenerate API, retrieving relevant documents for the model to generate Finnish or English responses.

Perfect for advanced internal document search, users can ask where specific information is found across organizations' documents. RAG brings AI efficiency to any private data given access securely through Bedrock, unavailable to public models.