A full-stack project exploring how student data can help educators identify academic risk signals and plan timely support.
Repository · React + Express + MongoDB + Python ML service
Warning signs can be spread across attendance, grades, assignments and class participation. StayOnTrack brings student information and risk alerts into one application workflow so educators can review signals and consider timely support.
The goal is to support earlier conversations and interventions — not to treat a model's output as a final judgement about a student.
- Student data & risk alerts — API routes for student records and current alerts.
- Authentication — a login endpoint for user authentication.
- Risk recalculation integration — the backend can call a separately configured ML service.
- Web client — React interface built with Vite, with routing, API calls and charting dependencies.
- Deployment configuration — Render Blueprint configuration and a Dockerfile are present in the repository.
The real-world quality of predictions and educational outcomes have not been independently validated here. Check the current implementation and connected services before relying on any risk output.
| Layer | Stack |
|---|---|
| Frontend | React, Vite |
| Styling | Tailwind CSS |
| API / backend | Node.js, Express |
| Database | MongoDB, Mongoose |
| ML integration | Python, scikit-learn (risk recalculation) |
| Client libraries | Axios, React Router, Recharts |
| Deployment config | Docker, Render; Firebase configuration is also present |
Student / Educator
│
▼
React + Vite client
│ HTTP API
▼
Node.js + Express API ───────► MongoDB
│
▼
Configured ML service
(risk recalculation)
- Node.js and npm
- MongoDB local instance or MongoDB Atlas
- Python, if you are running the ML service locally
git clone https://github.com/theoudar03/student-risk-predictor.git
cd student-risk-predictorUse the root .env.example as a reference. Create the environment files for your local setup.
server/.env
MONGO_URI=your_mongodb_connection_string
JWT_SECRET=replace_with_a_long_random_secret
ML_SERVICE_URL=http://localhost:8001client/.env
VITE_API_BASE_URL=http://localhost:5000Set ML_SERVICE_URL to the actual URL of your ML service. Never commit real credentials or secret keys.
cd server
npm install
npm run devThe backend's development script uses nodemon. Keep this terminal running.
Open a second terminal from the repository root:
cd client
npm install
npm run devOpen the local URL printed by Vite.
cd client
npm run buildML service note: the backend expects an ML service URL for risk recalculation. Configure and start that service separately before testing the related flow; confirm its current source files for the exact startup command and API contract.
| Method | Endpoint | Purpose |
|---|---|---|
GET |
/health |
Health check |
POST |
/api/auth/login |
User authentication |
GET |
/api/students |
Retrieve student records |
GET |
/api/alerts |
Retrieve risk alerts |
POST |
/api/admin/risk-recalculate |
Trigger risk recalculation |
These routes are based on the existing project documentation. Check the backend source for request formats, response schemas and authentication requirements.
This is an evolving project. Before using it with real student records, validate model performance and limitations, secure endpoints, protect sensitive data, and test the complete frontend–API–database–ML flow. Risk scores should support human review, never replace it.
- Add current UI screenshots and a short demo walkthrough.
- Document the ML service setup and input/output schema.
- Add API examples and error-response documentation.
- Add automated tests for key API and risk-recalculation flows.
- Document model evaluation and known limitations.
made to help people notice the signal — and support the student. ✦