Inspiration

As a student, I kept hitting the same wall every semester: notes scattered across apps, deadlines buried in syllabi I'd forget to reread, and don't understand what to read before exams. I wanted a tool that didn't just store notes, but actually did something with them and help me and others learn.

What it does

Study Buddy turns your notes and syllabus into an active study system. Upload notes or ask questions, and it auto-extracts exam and assignment dates into a calendar, generates a day-by-day study plan working backward from your deadline, and creates practice quizzes, all from your actual material. A note-grounded Q&A answers questions with source citations, while Ask AI handles general questions across any subject when you don't have notes yet. Any answer can expand into a deeper explanation with one click.

How we built it

FastAPI + SQLite backend, React frontend. Codex scaffolded every backend service and frontend component from detailed prompts specifying exact data models, error handling, and conventions. GPT-5.6 (Terra) powers every AI feature via the Responses API. I reviewed every generated file before integrating, then deployed backend to Render and frontend to Vercel.

Challenges we ran into

Windows/OneDrive file locking broke my venv setup repeatedly. A UTF-16 encoded file silently broke uploads until Codex traced it to a narrow decoding assumption. CORS and a Vercel SPA routing issue caused confusing failures post-deploy. I also hit OpenAI billing/credit issues mid-build and debugged those alongside real feature work.

Accomplishments that we're proud of

proud of finishing this project by myself under 6 days ,though there were struggles i never gave up

What we learned

Specific prompts with exact signatures and conventions produce far better Codex output than vague requests. Also relearned that deployment and persistence deserve attention as early as the features themselves.