Retrieval-Augmented Generation (RAG) System for LLMs

Overview

This repository contains the implementation of a Retrieval-Augmented Generation (RAG) system designed to enhance the responses of Large Language Models (LLMs). By integrating cloud-deployed RAG with LLMs, our solution improves accuracy, contextual relevance, and informativeness of the model outputs in real-world applications.

Features

• Modular Architecture: Combines LLMs with dynamic external data retrieval.
• Enhanced Responses: Increases the precision and relevance of model outputs.
• Scalable Infrastructure: Utilizes Docker and Kubernetes for robust deployment.

Documentation

For more detailed information, refer to the report.

To Run

Follow these steps to deploy and access the application on Google Kubernetes Engine:

1. Apply the Deployment Configuration

• Open your terminal.
• Run the following command to apply the deployment.yaml configuration:
• kubectl apply -f deployment.yaml

2. Retrieve the External IP and Port

• To find out the external IP address and port, execute: kubectl get svc
• Note down the External IP and the port number provided.

3. Access the Application

• Open a web browser of your choice.
• Enter the URL in the address bar based on the model you are using:
• For non-RAG LLM:

  http://<External IP>:<Port>/search_simple?query=<query>
  

• For RAG LLM:

  http://<External IP>:<Port>/search?query=<query>
  

• Replace <External IP> and <Port> with the actual IP address and port number from step 2. Replace <query> with your search query.

Ensure you replace placeholders with actual values when executing commands. This guide assumes that your deployment.yaml is correctly configured and that you have the necessary permissions to deploy to GKE.