Embeddings Recommendation Engine

Overview

This is a simplified demo of a lesson recommender system for language learning apps, showcasing AI-driven personalization. Users can explore lessons, book up to three, and receive recommendations based on semantic title similarities. The demo uses precalculated embeddings, cosine similarities, and weighted attributes (60% text, 30% language, 10% CEFR). For a real-world, robust solution with advanced methods and models, hiring a professional is recommended!

Features

• Explore Lessons: View a list of lessons with title, language, and CEFR level (A1 to C2).
• Book Lessons: Add up to three lessons to your bookings (client-side, reload to reset).
• AI Recommendations: Get personalized lesson recommendations based on your three most recent bookings.
• Svelte UI: A clean, responsive interface with a collapsible guide for users.
• Deployment: Deployed to a VPS using GitHub Actions.

Tech Stack

• Frontend: Svelte (UI), Native JavaScript (logic), Native CSS (styling)
• Backend: Spring Boot, Java, SQL
• AI: TensorFlow (embeddings, cosine similarities)
• Build: Maven
• Deployment: GitHub Actions to VPS

Setup

Prerequisites

• Java 21+
• Maven
• MariaDB 11.7
• Node.js (for Svelte build)

Steps

1. Clone the Repository:

   git clone https://github.com/ferderer/embeddings-recommendation-engine
   cd embeddings-recommendation-engine
   

2. Backend Setup:

• Configure MariaDB and update application.properties with your database credentials.
• Build and run the Spring Boot app:

  mvn spring-boot:run -P dev
  

3. Frontend Setup:

• Navigate to the Svelte app directory (if separate, e.g., frontend/):

  cd ui-svelte
  npm install
  npm run build
  

• Copy the built files (e.g., dist/) to src/main/resources/static/ in the Spring Boot project.

4. Run the App:

• Access the app at http://localhost:8080/.

5. Deployment:

• The project uses GitHub Actions to deploy to a VPS. Update the workflow in .github/workflows/ with your VPS credentials.

Usage Guide

• Explore Lessons: View all lessons in the table (title, language, CEFR).
• Book a Lesson: Click "Book" to add a lesson (up to three). A checkmark and "Booked" will appear.
• View Booked Lessons: See your bookings with timestamps (latest on top). Reload the page to start over.
• Get Recommendations: After booking, view AI-driven recommendations based on your recent bookings.

Note: This demo uses a simple recommendation engine with precalculated embeddings and cosine similarities. Real-world engines are more complex, supporting various methods (collaborative filtering, content-based) and models (deep learning, matrix factorization). For a professional solution, hire me!

Hiring a Professional

This demo is a simplified prototype. To build a robust, scalable recommender system tailored to your needs, with advanced methods and models, contact me! I’m a freelance senior developer with 25+ years of experience in AI solutions and fullstack development. DM me on LinkedIn [your LinkedIn profile link] to discuss your project!

License

© 2025 Vadim Ferderer. All rights reserved for commercial use.