jarvislabs 0.2.0b2

CLI and SDK for JarvisLabs.ai GPU cloud

jarvislabs

CLI and Python SDK for managing JarvisLabs GPU instances.

Detailed docs are being revised. For now, use jl --help, jl run --help, and the examples below.

Installation

From source:

uv pip install -e .

As a package:

uv pip install jarvislabs

Authentication

jl login

Or set an env var:

export JL_API_KEY="<your_api_key>"

CLI Quick Start

Show top-level help and command groups:

jl --help
jl instance --help
jl scripts --help
jl filesystem --help

Common commands:

jl status
jl gpus
jl templates
jl instance list
jl scripts list
jl filesystem list

Instance lifecycle:

jl instance create --gpu A100 --storage 40 --name my-instance
jl instance pause <machine_id>
jl instance resume <machine_id>
jl instance destroy <machine_id>

Script and filesystem integration:

jl scripts add ./startup.sh --name setup-script
jl filesystem create --name data --storage 120
jl instance create --gpu A100 --script-id <script_id> --fs-id <fs_id>

SDK Quick Start

from jarvislabs import Client

with Client() as client:
    instances = client.instances.list()
    print([i.machine_id for i in instances])

Create an instance:

from jarvislabs import Client

with Client() as client:
    inst = client.instances.create(
        gpu_type="A100",
        num_gpus=1,
        template="pytorch",
        storage=40,
        name="my-instance",
    )
    print(inst.machine_id, inst.status)

Current Behavior Notes

• Region is automatically resolved by the SDK/CLI.
• create/resume are reserved-only (is_reserved=True).
• CLI command naming uses list consistently (instance list, scripts list, filesystem list).

Development

uv run ruff format .
uv run ruff check --fix .
uv run pytest