Inspiration

Inspiration

I kept losing SAT points on the same concepts, and I had trouble using existing AI models for my prep. General chatbots solve the question for you, so I never learned why I got it wrong. I wanted a tutor that makes me try first, finds the exact concept behind my mistake, and then gives me practice on it.

What I learned

  • How a web page is put together: the page you see, the code behind it, and how they talk to each other
  • How to connect an app to an AI service and make it answer the way I want
  • How to save data, like chats and my SAT mistakes
  • How to keep secret keys safe and put a project on GitHub
  • How to test my work and fix problems

How I built it

The backend uses Python and FastAPI. main.py holds the API key and talks to the Google Gemini API, so the key never goes in the web page. The frontend is plain HTML, CSS and JavaScript. Chats and my mistake log are saved in a SQLite file.

The tutor streams its answers and remembers the conversation. It asks what I chose first, then gives one hint at a time before the full solution. A stronger Gemini model is used only when a photo is in the chat. I deployed it online on Render.

I built this with the help of AI tools, and I tested and improved it myself.

Challenges I faced

  • Stopping it from just giving answers. I had to carefully guide the AI to ask first, give one hint at a time, and never flatter me.
  • Reading photos. The AI can misread an SAT question from a camera photo, so the app shows what it read and lets me correct it before continuing.
  • Keeping the API key safe. I moved all AI calls to the backend so the key is never exposed.
  • Free-limit and server errors. Gemini sometimes hits the free limit or is busy, so I added friendly error messages for those cases and for a bad key. ## What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for SAT Mistake Tutor

Built With

Share this project:

Updates

Submission history