Basic understanding of Retrieval-Augmented Generation (RAG) in AI space

-

 Retrieval-Augmented Generation (RAG): In simple terms, RAG is a technique that uses AI to improve the accuracy of text generation. RAG is a framework that combines large language models like GPT-3 with external data sources. It works by retrieving relevant information from databases or docs and then using it to generate better responses. RAG applications use an LLM to generate responses and retrieve information from a knowledge base to improve the accuracy of the answers. Most companies building RAG applications for customer support, education, content recommendations, and entertainment etc.

For example, imagine you asked a RAG-powered chatbot, “What’s the capital of India?” The chatbot would first quickly search an online database to find that “The capital of India is New Delhi.” Then, using that retrieved information, the chatbot would respond, “The capital of India is New Delhi.”

So, in a nutshell, a RAG system: 

  • Retrieve the relevant data
  • Use that data to augment their responses



















Picture Credit: Cookbook OpenAI

RAG working functionality can be described in generally 5 steps:
  • Retrieval: RAG starts by searching through a database or a collection of documents to find information relevant to the input question or prompt. It uses techniques like keyword matching, semantic similarity, or other methods to find the most relevant documents.
  • Selection: Once it retrieves a set of documents, RAG selects the most relevant ones based on certain criteria. This could involve scoring each document based on its relevance to the input and choosing the top-ranked ones. 
  • Integration: After selecting the relevant documents, RAG integrates the information from these documents into its model. It might preprocess the text, extract key information, or represent the documents in a way that the model can understand. 
  • Generation: With the integrated information, RAG generates an answer or response to the input question or prompt. It uses techniques from language generation models like GPT to produce human-like text based on the integrated information. 
  • Refinement: Finally, RAG might refine its generated response based on feedback or additional context. It could adjust the answer to make it more accurate, coherent, or relevant.
Biz benefits of RAG:
1) With RAG, customer reps can ask the company LLM a product question and receive a comprehensive answer that is automatically sourced from relevant product manuals, FAQs, customer reviews, sizing charts, inventory data, and other documents. This could take seconds. 
2) RAG empowers users to quickly retrieve relevant information without having to sift through large volumes of data. 
3) Access to accurate and up-to-date information allows users to make more informed decisions faster.

Some of the usecases for building RAG:
  • Chatbots
  • KB Search
  • Medical Drug Discovery
  • Fraud Detection
To know more about RAG, google it or visit the below sites mentioned in references.

References:

Popular posts from this blog

Connecting Claude to Pega Infinity 25.1.3 via MCP — Step-by-Step

-
Image
Enable AI-powered access to your Pega Investigative Case Management data directly from Claude prompts using the Model Context Protocol (MCP) and OAuth 2.0. What is MCP and why connect Pega to Claude? The Model Context Protocol (MCP) is an open standard that lets AI assistants like Claude securely call tools and read data from external systems — all from a natural-language prompt. Think of it as a universal adapter between your enterprise systems and AI. Pega Infinity V25.1.3 ships with a first-class MCP service built on its existing REST framework. Once enabled, Claude can: Query and summarize investigative cases without switching apps Retrieve assignment details and case status on demand Trigger Pega workflows from conversational prompts Version requirement: This guide is based on Pega Infinity V25.1.3 (Community Edition) with the Rule-Service-MCP rule type. Earlier versions may not include native MCP support. Architecture overview The connection uses three layers: Claude as the AI c...
Read more

itextpdf API to generate PDF doc from an image file using Pega PE

-
Image
Probably some of the techies might have an idea on an opensource called " itextpdf " (earlier name is lowagie ) to convert an image file(png,jpeg,jpg,tiff,gif etc.) to a PDF file. It is very simple to use this API in Pega function to build the logic using java code. Moreover, this API comes in one single jar file and we need to import it in Pega before we start using it.  I see the conversion comes good in PDF file but not sure if any limitations of this API in terms of conversion is like formatting, font, images etc. So far so good and may be we can keep this API as one of our solutions while working on the requirements accordingly.  Heard that same API can be used to convert the text file to PDF but I didn't explore on it. Those who are interested to work on text -> PDF can take a look at the documentation. Don't hesitate to contact me offline for any further discussion in this context :) :) Prerequisites for doing this exercise: 1) itextpdf 5.4.0 jar file 2) Peg...
Read more

Understanding of Hugging Face platform for AI/ML platform

-
Image
Hugging Face is a platform that provides tools and libraries for natural language processing (NLP), machine learning, and artificial intelligence. It is known for its Transformers   library, which offers pre-trained models for tasks such as text classification, translation, question answering, and more. Hugging Face also hosts a model hub where users can share and discover pre-trained models. Hugging Face is useful in various domains, including: Natural language processing (NLP): Provides tools and pre-trained models for tasks like text classification, translation, sentiment analysis, and question answering. Machine learning research: Offers state-of-the-art models and datasets for researchers to experiment and innovate. Data science and analytics: Helps data scientists analyze and interpret large volumes of text data. AI application development: Enables developers to build AI-powered applications with ease using pre-trained models and APIs. A very simple exercise using the google...
Read more