Skip to main content

Getting Started

This guide walks through installing the Targon CLI and SDK, configuring credentials, and launching your first rental workload.

Prerequisites​

  • Python 3.9 or later (for the Python SDK)
  • Node.js 20 or later (for the TypeScript SDK)
  • Go 1.21 or later (for the Go SDK)
  • Rust toolchain (for the CLI)
  • A Targon account and API key

Install the CLI​

The CLI is a standalone Rust binary. Build it from the targon-sdk repo:

git clone https://github.com/manifold-inc/targon-sdk.git
cd targon-sdk/cli
cargo build --release

See the CLI installation guide for details.

Install the SDK​

The SDK lives under libs/ in the targon-sdk repository. Language packages are published separately.

Python​

pip install targon-sdk

For local development from the repo:

git clone https://github.com/manifold-inc/targon-sdk.git
cd targon-sdk/libs/python
pip install -e ".[dev]"

Import the top-level API:

from targon import Client, Resources
from targon.client import CreateWorkloadRequest, PortConfig

TypeScript​

npm install @targon/sdk

Source: libs/typescript

Go​

go get github.com/manifold-inc/targon-sdk/libs/go

Source: libs/go

Configure Credentials​

Store your API key with the CLI:

targon auth login

Follow the prompt to paste your key. Credentials are stored under ~/.targon/.

The Python SDK also accepts credentials from:

  • The TARGON_API_KEY environment variable
  • The same ~/.targon/ credential files written by the CLI
from targon import Client

client = Client.from_env()  # reads TARGON_API_KEY or ~/.targon/
# or
client = Client(api_key="your-api-key")

Check Available Resources​

List compute that is available right now:

targon inventory --gpu

Or query inventory from Python:

from targon import Client

client = Client.from_env()

for item in client.inventory.capacity(gpu=True):
    print(f"{item.name}: {item.available} available @ ${item.cost_per_hour}/hr")

Use targon inventory to compare GPU tiers, hourly cost, and availability before creating a workload.

Create a Rental​

The fastest way to get started is through the dashboard:

  1. Open the Targon dashboard.
  2. Go to Rentals and click Create Rental.
  3. Pick a GPU or CPU configuration, image, ports, and optional volumes.
  4. Add an SSH key and deploy.

See the Rentals guide for the full walkthrough.

Create a Workload with the SDK​

For programmatic control, create and deploy a rental workload with the Python SDK:

from targon import Client, Resources
from targon.client import CreateWorkloadRequest, PortConfig

client = Client.from_env()

workload = client.workload.create(
    CreateWorkloadRequest(
        name="my-training-job",
        image="pytorch/pytorch:latest",
        resource_name=Resources.H200_SMALL,
        type="RENTAL",
        ports=[
            PortConfig(port=8080, routing="PROXIED"),
            PortConfig(port=2222, routing="DIRECT"),
        ],
    )
)

client.workload.deploy(workload.uid)
client.workload.wait_until_ready(workload.uid)

The SDK exposes service clients on Client for workloads, volumes, SSH keys, inventory, projects, and user account data. See the Workloads API reference for all workload types and fields.

You can also call the REST API directly:

curl -X POST https://api.targon.com/tha/v2/workloads \
  -H "Authorization: Bearer <YOUR_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "my-training-job",
    "image": "pytorch/pytorch:latest",
    "resource_name": "h200-small",
    "type": "RENTAL",
    "ports": [
      {"port": 8080, "protocol": "TCP", "routing": "PROXIED"},
      {"port": 2222, "protocol": "TCP", "routing": "DIRECT"}
    ]
  }'

Then deploy it:

curl -X POST https://api.targon.com/tha/v2/workloads/wrk-<UID>/deploy \
  -H "Authorization: Bearer <YOUR_API_KEY>"

Use targon workload get with the workload UID to confirm it is running.

Next Steps​