✨ Cine-AI: The Intelligent Movie Recommender

Cine-AI is a production-ready, full-stack Generative AI application that demonstrates how modern recommendation systems go beyond keywords using semantic search, vector databases, and large language models.

This project showcases end-to-end GenAI system design—from embedding pipelines and vector retrieval to RAG (Retrieval-Augmented Generation), where Google Gemini acts as an AI movie critic to generate explainable recommendations.

🎯 Built to highlight skills in: Full-Stack Development, AI/ML Integration, Vector Databases, System Design, and Cloud Deployment.

---

⭐ Portfolio Highlights

• Designed and deployed a semantic recommendation engine using real-world movie data
• Implemented a complete RAG pipeline with grounding, retrieval, and LLM reasoning
• Integrated Qdrant vector search for millisecond-level similarity matching
• Built a scalable FastAPI backend with async endpoints
• Developed a modern React UI focused on UX, animations, and responsiveness
• Deployed a cloud-native architecture using Vercel, Render, and managed AI services

---

🚀 Features

• 🧠 Semantic Search – Search movies using natural language (e.g., "a mind-bending thriller about dreams" → Inception).
• 🤖 RAG-powered Explanations – AI-generated reasoning for every recommendation using Google Gemini.
• 🔗 Movie-to-Movie Similarity – Discover "soulmate" movies based on plot-level similarity.
• 🔥 Real-time Trending – Fetches the top 10 trending movies globally from TMDB.
• 🎨 Neon Glassmorphism UI – Modern dark-mode UI with smooth animations and responsive design.
• ☁️ Cloud Native – Easily deployable on Vercel, Render, Hugging Face Spaces, and Qdrant Cloud.

---

🏗️ Architecture

Cine-AI follows a microservices-based architecture with clear separation of concerns between frontend, backend, vector storage, and AI services.

graph LR
    A[User / React Frontend] -->|JSON Requests| B[FastAPI Backend]
    B -->|Vector Search| C[Qdrant Cloud]
    B -->|Context + Prompt| D[Google Gemini LLM]
    B -->|Metadata Fetch| E[TMDB API]
    C -->|Top Matches| B
    D -->|AI Explanation| B
    B -->|Final Response| A

Loading

---

🛠️ Tech Stack

Frontend

• Framework: React (Vite)
• Routing: React Router v6
• Styling: Pure CSS3 (Glassmorphism, Grid, Animations)
• Deployment: Vercel

Backend

• Framework: Python FastAPI
• Embedding Model: all-MiniLM-L6-v2 (Sentence Transformers)
• Vector Database: Qdrant (Managed Cloud)
• LLM: Google Gemini 1.5 Flash
• Deployment: Render / Hugging Face Spaces

---

⚙️ Installation & Setup

Prerequisites

• Node.js
• Python 3.9+
• API keys for TMDB, Google Gemini, and Qdrant Cloud

---

1️⃣ Backend Setup (FastAPI)

cd movie-backend

# Create virtual environment
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

Create a .env file:

TMDB_API_KEY=your_tmdb_key
GEMINI_API_KEY=your_gemini_key
QDRANT_URL=https://your-cluster.qdrant.tech
QDRANT_API_KEY=your_qdrant_key

Run the backend server:

uvicorn main:app --reload

---

2️⃣ Database Ingestion (Optional)

If starting with an empty Qdrant cluster, ingest movie embeddings:

python ingest.py

---

3️⃣ Frontend Setup (React)

cd client

# Install dependencies
npm install

# Start development server
npm run dev

---

🔌 API Endpoints

Method	Endpoint	Description
POST	/recommend	Semantic search with AI explanation
POST	/similar	Find movies similar to a given title
GET	/trending	Fetch top 10 trending movies from TMDB
GET	/health	Health check endpoint

---

🌐 Live Demo

🚀 Frontend: https://cine-ai-vert.vercel.app/ ⚙️ Backend API: https://jain-mayukh-movieback.hf.space

⚠️ Note: The backend may take a few seconds to wake up on first request due to free-tier hosting.

---

🧩 RAG Pipeline Explained

Cine-AI uses a Retrieval-Augmented Generation (RAG) pipeline to deliver both accurate and explainable movie recommendations.

1. Query Embedding

• User input is converted into a dense vector using Sentence Transformers (all-MiniLM-L6-v2).
• This captures semantic meaning beyond keywords.

2. Vector Retrieval (Qdrant)

• The query vector is searched against movie plot embeddings stored in Qdrant Cloud.
• Top-K semantically similar movies are retrieved with similarity scores.

3. Context Construction

• Retrieved movie metadata (title, plot, genre, score) is compiled into a structured context.
• This context acts as grounded knowledge for the LLM.

4. LLM Reasoning (Google Gemini)

• Context + user query are sent to Gemini 1.5 Flash.
• Gemini generates a human-like explanation justifying why each movie matches the request.

5. Final Response

• The backend merges similarity scores, metadata, and AI explanations.
• A clean JSON response is sent to the frontend.

---

⚡ Performance & Scaling

Cine-AI is designed for low-latency inference and horizontal scalability.

🚀 Performance Optimizations

• Precomputed Embeddings: All movie plots are embedded offline to avoid runtime overhead.
• Fast Vector Search: Qdrant enables millisecond-level nearest-neighbor search.
• Lightweight LLM Calls: Gemini Flash is used for fast, cost-efficient reasoning.
• Async FastAPI: Non-blocking request handling for high concurrency.

📈 Scalability Strategy

• Stateless Backend: Easily scalable across multiple instances.

• Managed Vector DB: Qdrant Cloud auto-scales with dataset size.

• Frontend CDN: Deployed on Vercel with global edge caching.

• Future Enhancements:

  • Redis caching for repeated queries
  • Batch embedding updates
  • User-personalized recommendation vectors

---

📸 Screenshots

Add screenshots of the following views:

• Home Page

• Search Results

• Trending Movies

---

🤝 Contributing

Contributions are welcome! 🚀

1. Fork the repository
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

---

📄 License

This project is distributed under the MIT License. See the LICENSE file for more information.

---

👤 Author

Mayukh Jain 🔗 LinkedIn | 💻 GitHub