Vid-Quiz-Gen

A backend system that recommends LeetCode problems based on YouTube DSA tutorial content using semantic similarity matching.

Overview

Vid-Quiz-Gen extracts transcripts from YouTube DSA tutorials, processes them using AI-powered summarization, and matches them against a database of LeetCode problems using vector embeddings to suggest relevant practice problems.

Features

• Transcript Extraction: Scrapes YouTube video captions using Colly
• AI Summarization: Summarizes transcripts using Gemini 2.0 Flash
• Semantic Search: Generates 768D embeddings with Gemini API and queries using pgvector
• Problem Matching: Recommends top LeetCode problems based on video content
• gRPC Service: Python microservice for scraping LeetCode problem statements

Tech Stack

• Backend: Golang, Python
• Database: PostgreSQL with pgvector extension
• AI/ML: Gemini API (text-embedding-004, gemini-2.0-flash)
• Communication: gRPC, Protocol Buffers
• Web Scraping: Colly (Go), leetscrape (Python)

Setup

Prerequisites

• Go 1.21+
• Python 3.9+
• PostgreSQL with pgvector extension
• Gemini API key

Installation

1. Clone the repository

git clone https://github.com/Sayan-995/vidquizgen.git
cd vidquizgen

2. Install Go dependencies

go mod download

3. Install Python dependencies

cd lcscrape
pip install -r requirements.txt

4. Set environment variables

export GEMINI_API_KEY=your_api_key
export DATABASE_URL=your_postgres_connection_string

5. Run the services

# Start Python gRPC server
cd lcscrape && python lcscrape.py

# Start Go backend (in another terminal)
go run main.go

API

POST /api/quiz

Generate LeetCode problem recommendations from a YouTube video.

Request:

{
  "url": "https://www.youtube.com/watch?v=VIDEO_ID"
}

Response:

{
  "questions": [
    {
      "title": "Two Sum",
      "url": "https://leetcode.com/problems/two-sum"
    }
  ]
}

License

MIT