RAG Chatbot Portal

This project is a full-stack web application that provides an admin and user portal for a Retrieval-Augmented Generation (RAG) chatbot. It consists of a FastAPI backend for API services and a choice of a Streamlit or Gradio frontend for the user interface.

The entire application is designed to be containerized with Docker and deployed to cloud platforms like Google Cloud Platform (GCP).

Project Structure

.
├── backend/
│   ├── Dockerfile
│   ├── README.md
│   └── ... (FastAPI source code)
│
├── frontend/ (or frontend-gradio/)
│   ├── Dockerfile
│   ├── README.md
│   └── ... (Streamlit/Gradio source code)
│
├── tests/
│   ├── locustfile.py
│   ├── setup_test_environment.py
│   └── ... (Load testing scripts)
│
└── docker-compose.yml

Features

• Dual User Roles: Separate interfaces and permissions for regular Users and Admins.
• User Management: Admins can add, delete, and manage user credentials.
• Data Management: Users and Admins can upload and manage PDF documents.
• VectorDB Integration: Uploaded documents can be ingested into a vector database for RAG.
• Interactive Chat: Users can chat with the RAG model, which utilizes the ingested documents.
• Containerized: Fully containerized with Docker for easy setup and deployment.

Deployment to GCP with Cloud Run

This is the recommended method for deploying the application to production.

Step 1: Build and Push Docker Images

1. Authenticate with GCP: Make sure you have the gcloud CLI installed and authenticated.

   gcloud auth login
   gcloud auth configure-docker

2. Enable Services: Enable Artifact Registry and Cloud Run API in your GCP project.

3. Build the Images: From the root directory, build both images using Docker Compose.

   docker-compose build

4. Tag and Push Images: Tag the built images and push them to Google Artifact Registry. Replace [PROJECT-ID] with your GCP Project ID.

   # Tag the backend image
   docker tag rag-backend gcr.io/[PROJECT-ID]/rag-backend:latest
   # Push the backend image
   docker push gcr.io/[PROJECT-ID]/rag-backend:latest
   
   # Tag the frontend image
   docker tag rag-frontend gcr.io/[PROJECT-ID]/rag-frontend:latest
   # Push the frontend image
   docker push gcr.io/[PROJECT-ID]/rag-frontend:latest

Step 2: Deploy the Backend Service

1. Deploy the backend image to Cloud Run.

   gcloud run deploy rag-backend \
     --image gcr.io/[PROJECT-ID]/rag-backend:latest \
     --platform managed \
     --region [YOUR_REGION] \
     --allow-unauthenticated \
     --set-env-vars="ADMIN_USERNAME=admin,ADMIN_PASSWORD=[CHOOSE_A_SECURE_PASSWORD]"

2. After deployment, Cloud Run will provide a public URL for your backend service. Copy this URL.

Step 3: Deploy the Frontend Service

1. Deploy the frontend image to Cloud Run, making sure to provide the backend's public URL as an environment variable.

   gcloud run deploy rag-frontend \
     --image gcr.io/[PROJECT-ID]/rag-frontend:latest \
     --platform managed \
     --region [YOUR_REGION] \
     --allow-unauthenticated \
     --set-env-vars="BACKEND_URL=[PASTE_THE_BACKEND_URL_HERE]"

2. Your application is now live! Access the frontend using the URL provided by Cloud Run.

Local Development using Docker Compose

To run the entire stack locally for development and testing:

1. Build and Start Containers: From the root directory, run:

   docker-compose up --build

2. Access Services:
   • Frontend: http://localhost:8501 (for Streamlit) or http://localhost:7860 (for Gradio)
   • Backend: http://localhost:8000

Load Testing

The tests/ directory contains scripts for load testing the application using Locust. See the tests/README.md for detailed instructions on how to prepare the environment and run the tests.