What it does
StudySphere AI takes your textbooks, lecture slides and class recordings and puts them in one place. You can ask it questions, and every answer comes with its source, like the textbook page, the slide number or the time in the video. If a question is not in your course material, it tells you honestly that it does not know instead of guessing.
It also makes practice quizzes based on what you are weak in, tracks how well you know each topic, finds the real reason you are struggling, and plans your revision before you forget things.
How we built it
The website is built with React and TypeScript, and the backend uses Python with FastAPI. We started the project in Google AI Studio and used the Gemini API for the AI part. Every piece of course content is saved with its exact location, like the page number, slide number or timestamp. When you ask a question, the system finds the best matching pieces and shows them as sources. If it cannot find a good match, it says so.
Challenges we ran into
Reading textbooks, slides and recordings was harder than we thought, because each one is different and we had to keep the exact page, slide and time for every piece. Getting the tutor to answer when it can and refuse when it cannot also took a lot of testing. We also found a funny bug where the correct answer in every quiz question was option 2, so we fixed that by shuffling the options. And as we added features, the menu became crowded with repeated pages, so we merged and cleaned it up.
Accomplishments that we’re proud of
We are proud that our tutor shows its sources and is honest when it does not know. We are also proud that the whole study journey is in one app: upload your material, ask questions, take quizzes, see your progress and get a revision plan. We also built an evaluation section so we can show that the system actually works.
What we learned
A good AI tutor depends more on finding the right sources and being honest than on having a fancy model. Saying “I don’t know” is a good feature, not a weakness. Even a small bug can make users lose trust. We also learned that testing from the beginning keeps the team focused.
What’s next for StudySphere AI
Next we want to add Google login and a proper database so each student’s data is saved to their account. We also want the tutor to search trusted websites when the course material cannot answer, and clearly say the answer came from outside the course. We plan to combine the study schedule and quizzes into one planner, improve reading of scanned notes and noisy recordings, and put the app online so any class can use it.
Built With
- bash
- css3
- fastapi
- framer-motion
- gemini-api
- google-ai-studio
- google-fonts
- html5
- javascript
- localstorage
- lucide-react
- node.js
- openapi
- powershell
- python
- rag
- ragas
- react
- react-router
- rest-api
- tailwindcss
- typescript
- uvicorn
- vite
Log in or sign up for Devpost to join the conversation.