Kubernetes-based Jupyter Notebook instance management with automatic lifecycle control.
- One-click notebook provisioning with email-based instance tracking
- AMD GPU support with automatic resource allocation
- Admin panel for instance management
- Auto-cleanup: 10min idle timeout, 6h max lifetime
| Method | Endpoint | Description |
|---|---|---|
| GET | / |
User interface |
| POST | /api/notebook/request |
Request notebook (body: {"email": "xxx"}) |
| GET | /api/notebook/status?email=xxx |
Check status |
| GET | /admin |
Admin panel (user: admin) |
| GET | /api/admin/instances |
List all instances |
| DELETE | /api/admin/instance/{email} |
Destroy instance |
| DELETE | /api/admin/instances/all |
Destroy all |
| GET | /health |
Health check |
# Deploy to K8s
kubectl apply -f k8s-deployment.yaml
# Access
# User: http://<NODE_IP>:30080/
# Admin: http://<NODE_IP>:30080/admin (admin / admin123)pip install -r requirements.txt
uvicorn app.main:app --reload --port 8000Environment variables in k8s-deployment.yaml:
| Variable | Description |
|---|---|
SERVICE_HOST |
External IP for notebook URLs |
DEFAULT_IMAGE |
Default notebook container image |
GPU_LIMIT |
GPUs per notebook |
IDLE_TIMEOUT_MINUTES |
Idle timeout before cleanup |
MAX_LIFETIME_HOURS |
Maximum notebook lifetime |
ADMIN_PASSWORD |
Admin panel password |