race-cli

Command-line interface for the Race Engineering GPU cloud platform.

Install

pip install race-ai

Quick start

# Generate an API key first, via the Race Engineering web dashboard
# (Account Settings -> API Keys -> Create Key)

race login --key rk_xxxxxxxxxxxxxxxxxxxxxxxx
race deploy --gpu <slice_id> --template <template_id>
race status <pod_id>

Commands

Run race --help or race <command> --help for full option details at any time.

Configuration

By default the CLI talks to the production API. To point it elsewhere (e.g. a self-hosted or staging instance):

race login --key <key> --base-url https://your-instance.example.com

The base URL is remembered for future commands — you don't need to pass it every time.

Authentication & security

• API keys are generated through the web dashboard, not this CLI — race login only stores a key you already have.
• Keys are stored using your operating system's credential store (macOS Keychain, Windows Credential Manager, Linux Secret Service) via the keyring package — never written to a plaintext file on disk.
• race login validates the key against the API before storing it, so a mistyped or invalid key is never silently saved.

Retries and idempotency

race deploy automatically retries once on a network-layer failure (timeout, dropped connection) using the same idempotency key for every retry of that attempt — this guarantees a retried deploy can never accidentally create a second pod. A definitive error response from the server (e.g. insufficient balance, invalid template) is not retried.

Requirements

• Python 3.9+
• A Race Engineering account and API key

License

MIT — see LICENSE.