Ducky.Study

AI coding tutor, designed for university use. Custom-tailored for MSU classes, with in-depth context and responses designed to teach important concepts and designed to not provide code answers.

Comment

Inspiration

We all love generative AI, but sometimes it is a little too easy to cheat with. What if there was a tool that allowed students to use generative AI, but without the cheating? That's what we wanted to do with Ducky Study; an AI tool tailored to specific university classes, that emphasizes learning in a way students can understand and benefit from. We're using it anyways, so why not make it more beneficial?

What it does

Ducky Study uses the Google Gemini API, along with custom prompts based on CSE 331 as an example, to take user code and pass it to Gemini for analysis. The specific prompt used depends on what the user's input is and which assignment their code is for. Gemini will then provide feedback on the user's code that emphasizes important concepts and learning, including resources from the specific assignment and from the class.

How we built it

We used a Flask app, running on Google Cloud Run with DNS from Cloudflare. Our AI calls are done through OpenRouter, using Gemini Flash 1.5. The styling for the website is just HTML, CSS, and JS. Our backend is written in Python.

Challenges we ran into

There were a few hurdles we had to overcome for this project.

  • Hosting site: it took a while to figure out how to host the site, and we ended up using a Docker container on Google Cloud Run. After getting that set up, there were issues with the DNS configuration that took a good amount of time to fix, but we got there!
  • External factors: we have a lot of homework that also happens to be due this weekend, so managing our time between that and Spartahack has definitely been a challenge.
  • New experience: since we're all pretty new to hackathons, there was a bit of a learning curve involved as well. ## Accomplishments that we're proud of Our first hackathon experience; we're all very proud of the work we were able to put in! We're super happy to have a website hosted and working. ## What we learned We learned:
  • How to quickly prototype using generative AI and deployment tools like Google Cloud Run
  • Far too much about troubleshooting DNS issues
  • How to interact with AI through API calls for free, effective prompts ## What's next for Ducky Study We have a lot of future features we would love to implement. While we are not sure we will have the time to keep working on this project (for now) here's some of our ideas!
  • User login + per-user class selections
  • Chat histories
  • More classes, assignments
  • Customizable levels of help in response (e.g. for 231, have the response be quite specific, but for 425, have the response be more vague)
  • Work with MSU / CSE professors to add security and extra features

Built With

Share this project:

Updates