Paper Key : IRJ************296
Author: Sumit,Aman Raj,Rohit Sah
Date Published: 07 Jul 2024
Abstract
In modern world, data has become incredibly valuable, sometimes it leading to apprehension about sharing personal information with online AI models or websites. One of the most widely used AI functionalities is Retrieval Augmented Generation (RAG), which serves as a question & answer chatbot for documents. With RAG, users can submit various document formats like PDFs, Excel sheets and ask questions related to the content and get the relevant answer. Rather than providing the same answer, which is already present in document, RAG generates customized answers using Large Language Models (LLMs) based on the user's queries.RAG falls under the category of Generative AI, uses Large Language Models (LLMs) to generate responses based on their vast knowledge base. This capability enables RAG to provide answers about those queries which do not even specify in the document. Given the critical importance of data privacy, our project aims to develop a secure and locally hosted solution to protect user confidentiality.Our proposed solution involves using a local LLM model that operates on the user's Personal Computer (PC), which ensures the data privacy and security. More important, the project is designed to function independently without relying on an internet connection. Additionally, for users whose PCs may not support the LLM model or have low hardware configuration, we offer an alternative solutiona user-friendly Application Programming Interface (API) provided by Open AI or Hugging Face. Through this API, users can seamlessly integrate their own private API key to run the RAG application on their system, ensuring accessibility and convenience while prioritizing data privacy.