Skip to content

Latest commit

 

History

History
86 lines (61 loc) · 4.19 KB

README.md

File metadata and controls

86 lines (61 loc) · 4.19 KB

StudyBuddy (Vortex) Project Summary

1. Overview

StudyBuddy (also referred to as Vortex) is an AI-powered educational platform designed to act as a personalized study companion for students. It leverages Retrieval Augmented Generation (RAG) to provide accurate, textbook-sourced answers and now includes intelligent video recommendations to supplement learning.

2. Technical Architecture

The project follows a Microservices-lite architecture:

  • Frontend: Single Page Application (SPA) built with React & Vite.
  • Backend (Main): Node.js/Express REST API handling auth, data, and chat logic.
  • Backend (Intelligence): Python/FastAPI service for heavy ML tasks (RAG, Embeddings, Recommendations).
  • Database: MongoDB (Primary Data) and Pinecone (Vector Data).
graph TD
    User[User] -->|Interacts| Client[React Client]
    Client -->|REST API| NodeServer[Node.js Server]

    subgraph Data Layer
        NodeServer -->|Reads/Writes| MongoDB[(MongoDB)]
        NodeServer -->|Vectors| Pinecone[(Pinecone)]
    end

    subgraph Intelligence Layer
        NodeServer -->|Relays Requests| PythonService[Python Microservice]
        PythonService -->|Context Search| Pinecone
        PythonService -->|LLM| OpenAI[OpenAI API]
        PythonService -->|Search| YouTube[YouTube API]
    end

    PythonService -->|Returns Analysis| NodeServer
    NodeServer -->|JSON Response| Client
Loading

3. Tech Stack

Component Technology Key Libraries
Frontend React (TypeScript), Vite react-router-dom, tailwindcss, framer-motion, gsap, katex (Math), react-markdown
Backend Node.js, Express mongoose, passport (Auth), node-cron (Automation), multer, pdf-parse
ML Service Python, FastAPI scikit-learn (TF-IDF), pinecone-client, openai, google-api-python-client
Database MongoDB, Pinecone mongoose
Infra - dotenv (Config), nodemon

4. Key Features

🎓 Intelligent Chat (RAG)

  • Contextual Answers: Chatbot answers questions based only on uploaded course materials (textbooks, PDFs), reducing hallucinations.
  • Vector Search: Uses OpenAI embeddings and Pinecone to find relevant textbook chunks.
  • Math Support: Renders complex mathematical formulas using LaTeX/KaTeX.

📺 Video Recommendations (New!)

  • Smart Suggestions: Automatically suggests relevant YouTube tutorials for each syllabus topic.
  • Advanced Search: Uses RAG to extract subtopics from the textbook to refine YouTube search queries (e.g., "Graph Theory" -> "Graph Theory BFS DFS implementation").
  • Caching: Recommendations are cached in MongoDB and refreshed daily (12:00 AM IST) or on startup.

📂 Course & Content Management

  • PDF Parsing: Robust pipeline to parse, chunk, and embed PDFs (including OCR for scanned docs).
  • Topic Extraction: Automatically extracts and organizes topics from course materials.

🔐 Authentication & Security

  • OAuth & Local: Google OAuth integration via Passport.js.
  • Session Management: Secure session handling with MongoDB storage.

5. Recent Core Update: Video Recommendations

Implemented Feb 2026

  • Objective: Supplement text answers with visual learning resources.
  • Implementation:
    • Created recommendationCacheService in Node to manage data.
    • Built Python recommender.py for intelligent query generation.
    • Added automation via node-cron for background updates.
  • Status: Fully functional. Gracefully handles API quota limits.

6. Future Roadmap (Suggested)

  • User Progress Tracking: Analytics on topics studied and videos watched.
  • Interactive Quizzes: Generate quizzes from the RAG context.
  • Mobile Responsiveness: Further optimizations for mobile view.