Netflix Recommendations RAG Project

A prototype of an AI-powered recommendation system that uses Retrieval-Augmented Generation (RAG) techniques to recommend Netflix titles and generate dynamic descriptions using an LLM (e.g., OpenAI GPT). The system includes a FastAPI backend and a React frontend.

Table of Contents

• Project Overview
• Project Structure
• Setup & Installation
• Running the Backend
• Running the Frontend
• Usage
• RAG Workflow Explanation
• Future Improvements

Project Overview

This prototype demonstrates how to build a recommendation system using a CSV dataset of Netflix titles. The system does the following:

• Retrieves Netflix titles based on user preferences (e.g., genre, rating, release year).
• Augments the original descriptions with a GPT model to produce short, marketing-style summaries.
• The final recommendations are then displayed on the React frontend, allowing users to interact with the recommended movies.

Project Structure

Your project folder looks like this:

NetflixRecommendationSystem/
├── backend/
│   ├── data/
│   │   └── netflix_titles.csv
│   ├── venv/               
│   ├── app/
│   │   ├── pydantic_models.py
│   │   ├── routes/
│   │   │   └── recommend_routes.py
│   │   ├── services/
│   │   │   ├── rag_service.py
│   │   │   └── recommendation.py
│   │   └── __init__.py
│   ├── main.py
│   ├── requirements.txt
│   └── README.md           
├── frontend/
│   ├── src/
│   │   ├── App.js
│   │   ├── index.js
│   │   └── ...
│   ├── index.css
│   ├── package.json
│   └── ...
└── ...

Key Directories & Files

• backend/data/: Contains the netflix_titles.csv file used as the knowledge base.
• backend/app/routes/: Holds FastAPI route: recommend_routes.py.
• backend/app/services/: Core logic for recommendation (recommendation.py) and GPT augmentation (rag_service.py).
• backend/main.py: Entry point for FastAPI.
• backend/requirements.txt: Python dependencies.
• frontend/: React application for the UI.

Setup & Installation

Clone the Repository

git clone <https://github.com/abkan2/Netflix-Recommendations>
cd NetflixRecommendationSystem

Create a Virtual Environment

cd backend
python -m venv venv
source venv/bin/activate  # On macOS/Linux
venv\Scripts\activate     # On Windows

Install Backend Dependencies

pip install -r requirements.txt

Install Frontend Dependencies

cd ../frontend
npm install

Running the Backend

Navigate to the Backend Directory

cd ../backend

Set Up Environment Variables

If you're using GPT for augmented descriptions, set your OpenAI API key. For example:

export OPENAI_API_KEY=<your-key>

or create a .env file (if you have a loader for environment variables).

Run the FastAPI App

uvicorn main:app --reload

This should start the backend server at http://localhost:8000.

Running the Frontend

Navigate to the Frontend Directory

cd ../frontend

Start the React App

npm run dev

This should start the frontend on http://localhost:3000.

Usage

1. Open the App

   • Go to http://localhost:3000 in your browser.

2. Make Your Selections

   • Choose a genre (e.g., "Comedy").
   • Choose a rating (e.g., "TV-MA").
   • Optionally, provide a release year.

3. Get Recommendations

   • Click the "Get Recommendations" button. The frontend will send a POST request to the backend with your filters.

4. View & Interact

   • A list of recommended movies will appear.
   • Click on any recommendation to see its augmented description (generated via GPT).

RAG Workflow Explanation

Retrieve

• The system loads the Netflix CSV data into a Pandas DataFrame.
• When a user provides filters (genre, rating, release year), the backend filters the DataFrame to find matching titles.

Augment

• For each recommended movie, the backend calls the OpenAI GPT API (or another LLM).
• GPT is provided with a prompt that includes the movie's title, director, cast, and original description.
• GPT returns a short, enticing summary that is appended to the movie's data.

Response

• The backend sends these enriched movie entries back to the frontend as JSON.
• The frontend displays them in a list or table.
• The user can click a title to see its GPT-generated "augmented description."