TDS Solver - Paste Your Question, Get Your Answer

This project was developed under the Tools in Data Science coursework @ IIT Madras.

It is an agent that exposes an endpoint that receives any question of the Graded Assignment via POST request and sends the answer (after solving) as the response.

The Graded Assignments referred are:

GA 1: https://exam.sanand.workers.dev/tds-2025-01-ga1
GA 2: https://exam.sanand.workers.dev/tds-2025-01-ga2
GA 3: https://exam.sanand.workers.dev/tds-2025-01-ga3
GA 4: https://exam.sanand.workers.dev/tds-2025-01-ga4
GA 5: https://exam.sanand.workers.dev/tds-2025-01-ga5

Core technologies used:

• FastApi
• gpt-4o-mini as the main model
• embeddings for similarity matching
• other tools as needed to solve the given question
• etc.

This project can be used anywhere, all it needs is 2 api keys in the os environment:

• create the os variable AIPROXY_TOKEN and assign the value as the LLM provider's API key
• create the os variable GEMINI_API_KEY and assign the value as the Google's Gemini API key (free tier is enough)
• clone the repository
• globally replace "http://aiproxy.sanand.workers.dev/openai/v1/chat/completions" with your LLM provider's API url
• create the virtual environment using
  • python3 -m venv .venv
• activate the virtual environment
• install the dependencies using
  • pip3 install -r requiremnets.txt
• run the server using
  • python3 server.py
• access the endpoint @ port 8080/api of the host

Here's an example of a sample request:

curl -X POST "localhost:8080/api" \
  -H "Content-Type: multipart/form-data" \
  -F "question=Download and unzip file abcd.zip which has a single extract.csv file inside. What is the value in the "answer" column of the CSV file?" \
  -F "file=@abcd.zip"

and here's a sample response:

{
  "answer": "1234567890"
}

** ensure your local environment has the dependencies listed in needed_dependencies.txt
** A more user-friendly deployable version is in the pipeline. Stay tuned.

Pradeep Mondal
24th April, 2025