Video-Recommendation-System

A powerful backend system that delivers personalized and category-based motivational video recommendations using deep learning and hybrid recommendation models.

🚀 Project Features

• Personalized content recommendation based on user interactions.
• Handles cold-start problem using mood/category-based filtering.
• Hybrid approach using collaborative filtering + content-based embeddings.
• SBERT embeddings for user and post representation.
• Daily caching of embeddings and interaction data.
• Integrates with Empowerverse and Socialverse APIs.

🧰 Tech Stack

• Backend: FastAPI
• ML: SentenceTransformers (SBERT), FAISS, ALS (implicit)
• Data: Pandas, Scikit-learn
• API Testing: Postman

---

🧱️ System Architecture

• User Embeddings: Text features from bio, role, and user type encoded using SBERT; numerical features scaled with MinMaxScaler.

• Post Embeddings: Metadata (title, topic, slug, and post summary) embedded using SBERT.

• Hybrid Model:

  • Collaborative Filtering: Using Implicit ALS on user-post interactions.
  • Content-Based Filtering: Using SBERT and FAISS for semantic similarity.
  • Blended Score: Combines both methods with weighted fusion to rank posts.

• Caching:

  • Daily caching of embeddings and interactions in cache/ using pickle.
  • Avoids recomputation unless the day changes.

• Data Fetching:

  • Fetches views, likes, inspires, ratings, users, and posts from the Socialverse API.

• Serving Layer:

  • FastAPI serves endpoints and resolves usernames to user_ids.
  • Embeddings are used to generate feed responses in real time.

---

📁 Folder Structure

video-recommendation-engine/
├── app/
│   ├── routes/
│   │   └── feed.py           # FastAPI endpoint logic
│   └── main.py               # FastAPI entry point
├── model.py                  # Hybrid recommendation model + embedding logic
├── cache/                    # Stores pickled embeddings, DataFrames, and cache date
├── requirements.txt          # Python dependencies
├── .env                      # Environment variables (not committed)
├── README.md
└── empowerverse_postman_collection.json   # Postman demo requests

---

⚙️ Setup Instructions

🔐 Prerequisites

• Python 3.8+
• Virtual environment (recommended)

✅ Installation

git clone https://github.com/Tim-Alpha/video-recommendation-assignment.git
cd video-recommendation-engine
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install -r requirements.txt

🔐 Configure Environment Variables

Create a .env file in the root directory:

FLIC_TOKEN=your_flic_token
API_BASE_URL=https://api.socialverseapp.com

▶️ Run the Server

uvicorn app.main:app --reload

Then open: http://127.0.0.1:8000/docs

---

📡 API Endpoints

🎯 Recommendation APIs

Get Personalized Feed GET /feed?username=<username>

Get Category-based Feed GET /feed?username=<username>&project_code=<project_code>

🔐 Authorization Header

All requests to Socialverse API use:

"Flic-Token": "your_flic_token"

---