API Reference

The Lightning SDK is the Python interface for managing Lightning AI resources from scripts, notebooks, services, and CI jobs. Use it when you need the same resource control exposed by the CLI, but want it inside Python code with typed objects, method references, and reusable application logic.

Install

Install or upgrade the package:

pip install lightning-sdk -U

Authenticate

For local development, authenticate once with the CLI:

lightning login

For automation, configure credentials with environment variables:

export LIGHTNING_API_KEY=your-api-key

Core Resources

Studios are persistent development environments. Start with Studio when you need to create or control an interactive workspace, and use Machine when selecting compute.

Jobs run code in managed cloud compute. Use lightning_sdk.Job.run for single-node work and lightning_sdk.MMT.run for multi-machine training.

Deployments turn model servers and applications into managed services. Use Deployment with the deployment helper classes in Deployment for environment variables, secrets, authentication, health checks, and rollout behavior.

Teamspaces, organizations, and users define the account context for SDK calls. Use Teamspace, Organization, and User when you need to resolve or switch the scope of an operation.

Use Organization to query usage/cost activity for an organization, or one or more of its teamspaces, and to download or export that activity as a CSV or JSON report.

Quick Example

from lightning_sdk import Machine, Studio

studio = Studio("my-studio")
studio.start()

job = studio.run("python train.py", machine=Machine.A10G)
job.wait()

studio.stop()

Reference Pages

Was this page helpful?