exls 0.6.0

The exalsius CLI

exalsius CLI - exls

The exalsius CLI - exls- is a command-line tool for orchestrating and managing distributed AI training workloads on ephemeral clusters. As a core component of the exalsius stack, it enables deployment and management of AI training jobs across multiple cloud providers and on-premise hardware.

Features

• Cluster Management: Deploy, manage, and scale ephemeral clusters for your AI workloads.
• Node Management: Easily add and manage nodes from different cloud providers or your own hardware.
• Cost-Aware GPU Selection: Scan for GPU prices across cloud providers to find the most cost-effective options.
• Workspace Orchestration: Create and manage various types of workspaces, such as Jupyter, DevPod, and more, on your clusters.
• Extensible and Cloud-Agnostic: Designed to be flexible and work with a variety of cloud providers.

Getting Started

Prerequisites

• Python 3.12 or newer.

Installation

It is recommended to install exls in a virtual environment.

python -m venv .venv
source .venv/bin/activate

You can install the CLI using pip or uv:

With pip:

pip install git+https://github.com/exalsius/exalsius-cli.git@main

With uv:

uv pip install git+https://github.com/exalsius/exalsius-cli.git@main

To verify the installation, run:

exls --help

Usage

Login

Before you can use the CLI, you need to authenticate with your exalsius account:

exls login

Core Commands

Here are some of the core commands to get you started:

• exls offers: Find the best GPU deals.
  • exls offers list --gpu-type "H100": List all offers for H100 GPUs.
• exls nodes: Manage the nodes in your node pool.
  • exls nodes import-offer <offer-id>: Import a node from a cloud provider offer.
  • exls nodes list: List all available nodes.
• exls clusters: Manage your clusters.
  • exls clusters deploy --interactive: Interactively create a new cluster.
  • exls clusters list: List all your clusters.
  • exls clusters get <cluster-id>: Get details for a specific cluster.
• exls workspaces: Manage workspaces on your clusters.
  • exls workspaces deploy jupyter <cluster-id>: Deploy a Jupyter workspace on a cluster.
  • exls workspaces list <cluster-id>: List workspaces on a cluster.
• exls get-token: Print your access (bearer) token for use in scripts that call the exalsius API. The token is refreshed automatically if expired.
  • curl -H "Authorization: Bearer $(exls get-token)" https://api.exalsius.ai/...
  • exls get-token --format json: Emit {"access_token": "<token>"} instead of the bare token.

For more details on each command, you can use the --help flag, for example exls clusters --help.

Documentation

For more in-depth information and advanced usage, please refer to our official documentation.

• Full Documentation
• API Reference

Community & Support

• Join our Discord server to chat with us.
• If you encounter any bugs or have feature requests, please open an issue on our GitHub repository.

License

This project is licensed under the Apache 2.0 License. See the LICENSE file for details.