This talk aims to show you a working sample of a Retrieval Augmented Generation (RAG) system running on AWS using components like Langchain, Amazon OpenSearch Service, and OpenAI. Let me explain to you why you need such a system.

Visitors to your website are looking for something. Maybe they are looking for something to buy. They can also seek information about delivery terms, opening hours, or advice about a product. Google and other websites see an increasing trend in users that use sentences instead of keywords. Wouldn’t it be great if you could answer the questions instead of pointing them to a page where they need to find the answer themselves?

You get an introduction to question-answering systems, what they do, and how it works. We briefly introduce the components we need: index-based search, vector-based search, LLMs, and Retrieval Augmented generation (RAG).

Next is an extensive demo, plus an explanation of the code. You will see a content pipeline to index content. Then you will see how to utilize different search engines in combination with a Large Language Model to generate answers on the fly. The demo shows you different mechanisms like content extraction and content generation for creating the answer. The results utilizing the different components are compared.

After the presentation, you have a fair understanding of the generic structure of a Retrieval Augmented Generation system to support question-answering on your website or within your application.

Adjustments to run on other platform, use other vector store are possible.