leptonai/leptonai

What it solves

Lepton AI provides a streamlined way to manage and operate the NVIDIA DGX Cloud Lepton platform. It removes the complexity of cloud infrastructure management by providing a unified Python library and command-line interface (CLI) to deploy and interact with AI workloads.

How it works

The project consists of two primary interfaces:

1. The lep CLI: A command-line tool used to create and manage endpoints, batch jobs, dev pods, storage, secrets, and clusters (Ray and Slurm).
2. The Client Python library: A tool that reads an endpoint's OpenAPI schema and allows developers to call deployed cloud endpoints as if they were native Python functions.

Additionally, the project includes "skills" that allow AI agents (like Claude Code or Codex) to operate the Lepton platform using natural language commands via the lep CLI.

Who it’s for

Developers and ML engineers who are using the NVIDIA DGX Cloud Lepton platform to deploy, scale, and manage their AI models and training jobs.

Highlights

• Unified Management: Manage endpoints, batch jobs, and dev pods from a single CLI.
• Dynamic Client: The Python client automatically discovers available paths and documentation from the endpoint's OpenAPI schema.
• Agentic Operation: Includes plugins for AI agents to manage workloads through natural language.
• Flexible Authentication: Supports API tokens and secure endpoint defaults for controlled access.