TDS Virtual Teaching Assistant

This project develops a Virtual Teaching Assistant API tailored for the Tools in Data Science (TDS) course at IIT Madras. The API intelligently responds to student queries by leveraging course materials and historical discussions from the Discourse forum, enabling quick, accurate, and automated support for learners.

Features

• ✅ FastAPI REST endpoint at /api/ accepting POST requests
• ✅ JSON request/response format with question and optional base64 image
• ✅ Structured response with answer and relevant links
• ✅ Course content integration with TDS-specific knowledge
• ✅ Discourse post simulation with realistic Q&A data

Setup

1. Install dependencies:

pip install -r requirements.txt

2. Set up environment variables:

cp .env.example .env
# Edit .env with your configuration

3. Scrape Discourse data (optional):

python scraper/scraper.py

4. Run the application:

python app.py

API Usage

Endpoint

POST /api/

Request Format

{
  "question": "Your question here",
  "image": "base64_encoded_image_data_optional"
}

Response Format

{
  "answer": "Detailed answer to the question",
  "links": [
    {
      "url": "https://discourse.onlinedegree.iitm.ac.in/t/...",
      "text": "Relevant link description"
    }
  ]
}

Example Usage

curl "https://your-app-url.com/api/" \
  -H "Content-Type: application/json" \
  -d '{"question": "Should I use gpt-4o-mini which AI proxy supports, or gpt3.5 turbo?"}'

Project Structure

├── app.py                 # Main FastAPI application
├── scraper/
│   ├── scraper.py  # Discourse and Course Content data scraper
├── data/
│   ├── discourse_posts.json # Scraped Discourse data
│   └── course_content.json  # Course content data
├── models/
│   ├── request_models.py     # Pydantic request models
│   └── response_models.py    # Pydantic response models
├── services/
│   ├── question_processor.py # Question processing logic
│   └── answer_generator.py   # Answer generation service
├── requirements.txt       # Python dependencies
├── LICENSE               # MIT License
└── README.md            # Project documentation

Evaluation

To test the application with the provided evaluation:

1. Update project-tds-virtual-ta-promptfoo.yaml with your API URL
2. Run: npx -y promptfoo eval --config project-tds-virtual-ta-promptfoo.yaml

License

MIT License - see LICENSE file for details.