Next — AI-Powered Movie Discovery Platform

Next is a full-stack movie discovery and recommendation platform.
It combines modern web technologies, recommendation systems, and large language models to deliver personalized movie recommendations, conversational search, and intelligent movie discovery experiences.

The project is organized as a monorepo containing two submodules:

• frontend/ — React + Vite client application
• backend/ — Flask API, recommendation engine, and AI services

---

Features

Personalized Recommendations

• Behavior Sequence Transformer (BST-style) recommender trained on user rating history
• SVD collaborative filtering recommender
• Background recommendation computation with APScheduler
• Personalized recommendation feeds stored per user

AI Movie Finder

Describe a movie in natural language:

“space movie with a black hole”

The backend uses an LLM to extract candidate movies and validates them against TMDB metadata.

Conversational Chatbot

An AI-powered movie chatbot capable of:

• Random movie recommendations
• Genre-based recommendations
• Director-based recommendations
• Actor-based recommendations
• Similar-movie recommendations
• Context-aware multi-turn conversations

Movie Discovery

• Trending, popular, and top-rated movies
• TV show browsing
• Movie details pages
• Cast and trailers
• Similar and recommended titles
• Full-text search

User Features

• Firebase authentication
• Email verification and password reset
• Watchlists
• Star ratings
• Activity tracking

---

Project Architecture

Layer	Technology
Frontend	React 18, Vite, Redux Toolkit, MUI
Backend	Flask 3.x, Python 3.11
Database/Auth	Firebase Auth + Firestore
Recommendation Models	TensorFlow/Keras BST + SVD
AI Services	Groq LLM API
Movie Data	TMDB API

---

Repository Structure

./
│
├── frontend/     # React + Vite frontend
├── backend/      # Flask backend + AI services
└── README.md

---

Frontend

The frontend is a React-based movie discovery interface that integrates both TMDB APIs and the Flask backend.

Frontend Tech Stack

• React 18
• Vite 4
• Redux Toolkit
• React Router v6
• MUI v5 + Emotion
• SCSS
• Axios

Frontend Features

• Trending / popular / top-rated browsing
• Movie and TV search
• Detailed movie pages
• Watchlists and ratings
• Personalized recommendations
• AI movie finder
• Floating AI chatbot

Frontend Setup

Prerequisites

• Node.js 18+
• Backend server running locally

Installation

cd frontend

npm install

Environment Variables

Create a .env file:

VITE_APP_TMDB_TOKEN=<your-tmdb-v4-read-access-token>
VITE_FLASK_API_URL=http://127.0.0.1:5000

Run Development Server

npm run dev

Frontend runs on:

http://127.0.0.1:5173

---

Backend

The backend powers authentication, recommendation models, AI integrations, and movie-related APIs.

Backend Architecture

Module	Responsibility
user/	Authentication, ratings, watchlists, history
movie_finder/	Natural-language movie search
movie_chatbot/	Conversational movie recommendations
movie_recommender/	BST + collaborative filtering
core/	Shared services and utilities
config/	Typed configuration system

Backend Tech Stack

• Python 3.11
• Flask 3.x
• TensorFlow 2.15 / Keras 2.15
• Firebase Admin SDK
• Groq SDK
• TMDB API
• APScheduler

---

Backend Setup

Prerequisites

• Python 3.11
• Firebase project
• TMDB API key
• Groq API key
• SMTP email account

Installation

cd backend

python -m venv venv

Activate virtual environment:

# Windows
venv\Scripts\activate

# Linux / macOS
source venv/bin/activate

Install dependencies:

pip install --upgrade pip
pip install -r requirements.txt

---

Backend Configuration

Create the following files inside backend/config/.

FlaskAppConfig.json

{
  "SECRET_KEY": "<secret>",
  "SECURITY_PASSWORD_SALT": "<salt>",
  "JWT_BLACKLIST_TOKEN_CHECKS": ["access", "refresh"],
  "JWT_BLACKLIST_ENABLED": true,
  "MAIL_SERVER": "smtp.gmail.com",
  "MAIL_PORT": 465,
  "MAIL_USE_SSL": true,
  "MAIL_USE_TLS": false,
  "MAIL_USERNAME": "<email>",
  "MAIL_PASSWORD": "<password>"
}

FirebaseConfig.json

{
  "apiKey": "...",
  "authDomain": "<project>.firebaseapp.com",
  "databaseURL": "https://<project>.firebaseio.com",
  "storageBucket": "<project>.appspot.com"
}

GroqConfig.json

{
  "api_key": "<groq-api-key>",
  "model": "llama-3.3-70b-versatile"
}

TMDBConfig.json

{
  "api_key": "<tmdb-api-key>",
  "language": "en"
}

key.json

Firebase service-account private key downloaded from Firebase Console.

---

Run Backend

python app.py

Backend runs on:

http://127.0.0.1:5000

---

Recommendation System

The recommendation engine combines two approaches:

1. Behavior Sequence Transformer (BST)

A Transformer-based recommendation model trained on user interaction sequences.

Features:

• Sequential recommendation
• Learns user behavior patterns
• Personalized predictions

2. Collaborative Filtering (SVD)

Matrix-factorization collaborative filtering using user ratings.

Features:

• Learns latent user/movie embeddings
• Captures collaborative preferences
• Complements sequence modeling

Recommendations are periodically recomputed using APScheduler and stored in Firestore.

---

AI Components

Movie Finder

Uses a Groq-hosted LLM to:

1. Parse natural-language descriptions
2. Generate candidate movie titles
3. Validate matches against TMDB

Conversational Chatbot

The chatbot:

• Detects recommendation intent
• Extracts filters
• Maintains conversational memory
• Returns contextual recommendations

Supported intents:

• Random recommendations
• By genre
• By director
• By actor
• Similar to another movie

---

API Overview

Authentication

POST /auth/signup
POST /auth/login
POST /auth/password_reset

Recommendations

GET /movie_recommender/
GET /movie_recommender/content-based
GET /movie_recommender/cf

AI Features

POST /movie_finder/
POST /chatbot/

User Actions

POST /ratings
POST /watchlist
GET /watchlist
POST /activity

---

License

This project is licensed under the MIT License.