UniScout AI

A professional college information assistant that provides accurate data by scouting official university websites and verified third-party aggregators in real-time.

Core Capabilities

• Official Source Detection: Automatically identifies official university domains while filtering out unreliable sources.
• Dynamic Data Acquisition: Scrapes and analyzes college data on-demand to ensure the most current information.
• RAG-Enhanced Intelligence: Utilizes Retrieval-Augmented Generation (RAG) with Google Gemini or Groq (Llama 3) for grounded, factual responses.
• Targeted Information Retrieval: Specific focus on admissions, fee structures, placement statistics, and campus facilities.

Technical Architecture

• Frontend: React 19, Tailwind CSS 4, Vite
• Backend: FastAPI (Python 3.12)
• Database: PostgreSQL / SQLite (SQLAlchemy)
• Intelligence: Google Gemini / Groq (Llama 3.3 70B)
• Automation: Playwright (Dynamic Scraping), Search Services

Setup and Installation

Prerequisites

• Node.js 18.0 or higher
• Python 3.12 or higher
• Playwright dependencies

1. Backend Configuration

Navigate to the backend directory and set up the Python environment:

cd backend
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install -r requirements.txt
playwright install chromium

Create a .env file in the backend/ directory:

DATABASE_URL=sqlite:///./college_bot.db
GEMINI_API_KEY=your_gemini_api_key
GROQ_API_KEY=your_groq_api_key (optional)
USE_GROQ=True (optional)

Start the backend server:

uvicorn app.main:app --reload

2. Frontend Configuration

Navigate to the frontend directory and install dependencies:

cd ../frontend
npm install

Start the development server:

npm run dev

The application will be accessible at http://localhost:5173.

System Workflow

1. Discovery: Identify the correct university website from the user's query.
2. Analysis: Classify the intent (e.g., Placements, Fees) to target specific data points.
3. Scouting: Execute a dynamic scrape using Playwright for high-fidelity data extraction.
4. Synthesis: Combine official data with aggregator insights to provide a comprehensive analysis.
5. Response: Deliver an AI-generated answer grounded strictly in the acquired evidence.

Maintenance

To clear the local knowledge base, delete the college_bot.db file in the backend directory. The system will automatically rebuild it on the next query.