Getting Started
Deploy your first model in under 60 seconds with the Upbox CLI.
1. Install the CLI
Run `npm install -g upbox` or `brew install upbox` to get the CLI. Docker images are also available at `upbox/cli:latest`.
2. Authenticate
Generate an API key from the Upbox dashboard and run `upbox login` to authenticate. Your credentials are stored securely in `~/.upbox/config`.
3. Deploy your first model
Run `upbox deploy ./model.pt` to push your model to production. Upbox auto-detects the framework (PyTorch, TensorFlow, ONNX) and configures the runtime.
4. Get your endpoint
Within seconds, you'll receive a live inference endpoint like `https://api.upbox.io/v1/models/your-model`. Start sending requests immediately.
5. Monitor in real-time
Run `upbox logs --follow` to stream inference logs, or visit the Upbox dashboard for latency graphs, request volumes, and error rates.
6. Scale automatically
Upbox handles scaling out of the box. As traffic increases, new instances spin up in milliseconds. When traffic drops, you scale to zero and stop paying.
7. Update with zero downtime
Push updates with `upbox deploy ./model-v2.pt`. Hot-swap technology means your endpoint never goes down - traffic shifts seamlessly to the new version.
8. Roll back instantly
Something wrong? Run `upbox rollback` to revert to any previous version in under a second. All version history is preserved automatically.