This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend
I built StudyMate, an AI study partner for a friend who is a real student. She has plenty of study material, but when she gets stuck on a topic she often doesn't know what to do next: re-read, watch a video, practice, or revise.
Most AI tools work like this: Ask → Get answer. Then the student is on their own again.
StudyMate works like this:
Ask → Understand → Practice → Revise → Improve
It looks at what the student asked and offers the right next actions.
The options are adaptive, not fixed. The same app gives different next steps depending on what the student is doing.
StudyMate's chat interface. After each answer, it offers adaptive next actions based on what the student asked.
Personalized AI study partner with contextual learning actions.
This version uses Ollama Cloud by default. The default configurable model is gpt-oss:120b.
cd backend
python -m venv .venv
.venv\\Scripts\\activate
pip install -r requirements.txt
copy .env.example .env
Open backend/.env and add your Ollama Cloud API key:
AI_PROVIDER=ollama_cloud
OLLAMA_HOST=https://ollama.com
OLLAMA_API_KEY=YOUR_KEY_HERE
OLLAMA_MODEL=gpt-oss:120b
Then:
uvicorn app.main:app --reload --port 8000
Check:
http://127.0.0.1:8000/api/health
In a second terminal:
cd frontend
npm install
npm run dev
Open the Vite URL shown in the terminal.
Never commit backend/.env. It contains your private API key.
The React frontend does…
The repo has the React + Vite frontend, the FastAPI backend, and a .env.example with setup instructions.
📧 Contact: chanveersinghdev@gmail.com
Student (question / doubt)
│
▼
React + Vite (UI)
│
▼
FastAPI backend
│
▼
Context / prompt engine
│
▼
Ollama Cloud ─► gpt-oss:120b
│
▼
Answer + adaptive actions
| Part | Tool |
|---|---|
| Frontend | React + Vite |
| Backend | FastAPI |
| AI | gpt-oss:120b (open-weight) via Ollama Cloud |
| Storage | SQL database for user data |
| Built with | Claude Code |
StudyMate is designed around how people actually learn:
Discover → Understand → Practice → Revise → Exam prep
The prompt engine detects the question type and phase, then chooses which actions to show.
The AI is an open-weight model (gpt-oss:120b) running through Ollama Cloud, not a closed proprietary API. The learning logic is separated from the provider and model:
StudyMate learning logic → AI provider → Model
The model is set in .env, so I can switch models or providers, or move to local inference later, without rewriting the product. That means less vendor lock-in and more control over the AI layer.
The Ollama API key stays on the FastAPI backend (.env). The browser never sees it, and only .env.example is committed to GitHub.
StudyMate never calls AI-generated questions "PYQs." They are labelled Practice question. Real previous-year questions will come from a verified database (year + exam + subject + chapter), so students can trust what they study.
I designed the interface like a real product: dark theme, rounded chat and composer, contextual action buttons, hover animations, chat history, and a responsive layout.
The biggest lesson was that a good study tool shouldn't stop at the answer. Knowing what to do next is what turns an answer into learning. StudyMate is the foundation for that, and I want to keep building it with my friend's feedback.