🛒 DemoKART 2.0

AI-Powered E-Commerce with Neural Embedding Recommendations

🚀 Overview

DemoKART 2.0 is a modern, full-stack E-Commerce platform that implements a production-grade machine learning recommendation engine. Instead of relying on basic keyword searches, DemoKART uses Semantic Search & Neural Embeddings to truly understand what a user is looking for, offering a "smart" shopping experience.

✨ Key Features

• 🧠 4-Layer AI Recommendation Engine:
  • Layer 1 (Content-Based): Uses all-MiniLM-L6-v2 to convert products into 384-dimensional mathematical vectors to find semantically similar items using Cosine Similarity.
  • Layer 2 (Collaborative Profiling): Averages the user's purchase history vectors into a "Personal Taste Vector".
  • Layer 3 (Trending): Tracks global purchase velocity across all users.
  • Layer 4 (Association Rules): Implements Market Basket Analysis to surface complementary goods (e.g. "Frequently Bought Together").
• ⚡ Real-time ML Inference: The PyTorch model is loaded natively into the FastAPI memory pool, updating product embeddings in milliseconds without needing batch processing.
• 🛡️ Secure Admin Portal: Password-protected dashboard with cryptographic SHA-256 hashing to manage inventory and user accounts safely.
• 🌐 Cloud-Native Persistence: Fully stateless architecture persisting data to Neon Postgres.

---

📸 Screenshots

(Replace these placeholders with your actual screenshots by dragging and dropping your images directly into the GitHub editor!)

Home Page & Recommendations

AI-Powered Product Detail

Admin Dashboard

---

🏗 Architecture & Tech Stack

This repository uses a Monorepo structure:

PROJECT_ROOT/
├── src/                  # React Frontend (Vercel)
│   ├── components/
│   ├── engine/           # Similarity Matrix & Rec Engine Math
│   └── pages/
├── api/                  # Python Backend (Render)
│   ├── main.py           # FastAPI Application
│   └── requirements.txt  # ML Dependencies
└── package.json          

• Frontend: React 19, Vite, React Router DOM, Vanilla CSS.
• Backend: Python 3, FastAPI, Uvicorn, psycopg2.
• AI / ML: PyTorch, sentence-transformers (all-MiniLM-L6-v2).
• Database: PostgreSQL (Hosted on Neon).

---

🚀 Quick Start (Local Development)

Prerequisites

• Node.js installed
• Python 3 installed
• A Neon PostgreSQL Database URL

1. Setup the Backend

Navigate to the api folder and install the Python dependencies:

cd api
pip install -r requirements.txt

Create a .env file inside the api/ folder:

DATABASE_URL=postgresql://user:password@neon.tech/neondb

2. Setup the Frontend

Return to the root directory and install Node dependencies:

npm install

3. Start the Application

Start both the React frontend and the FastAPI backend simultaneously:

npm run dev

• Frontend: http://localhost:5173
• Backend Swagger API Docs: http://localhost:8000/docs

(Note: On the very first boot, the FastAPI application will automatically seed your Postgres database with demo products!)

---

🌐 Deployment

This application is designed to be hosted globally using Edge & Web Services.

1. Backend (Render.com)

Deploy the backend as a Web Service on Render.

• Root Directory: api
• Build Command: pip install -r requirements.txt
• Start Command: uvicorn main:app --host 0.0.0.0 --port $PORT
• Don't forget to add your DATABASE_URL as an Environment Variable!

2. Frontend (Vercel)

Deploy the frontend natively on Vercel.

• Import the root folder of this repository.
• Environment Variables: Add VITE_API_URL and set it to your deployed Render URL (e.g. https://demokart-api-xyz.onrender.com).

Built with ❤️ by Manish | Artificial Intelligence Project