Skip to main content

​
Quickstart

This guide walks you through installing jnc on your Jetson device, creating your first workspace, adding dependencies, and running code.

​
Install jnc

Run the following on your Jetson: 
curl -L https://github.com/gentle-weapons/jnc-releases/releases/latest/download/pbctl-aarch64-unknown-linux-gnu.tar.gz | tar -xz
sudo mv pbctl /usr/local/bin/jnc
Verify the installation: 
jnc --version

​
Verify your hardware

First, confirm jnc can detect your Jetson: 
jnc probe
You should see output like: 
{
  "arch": "aarch64",
  "l4t": "36.4.0",
  "jetpack": "6.1",
  "cuda": "12.2",
  "module": "Jetson Orin Nano",
  "compute_capability": "8.7"
}

​
Create a workspace

jnc init my-jetson-app
cd my-jetson-app
This creates a jnc.toml manifest file that defines your workspace’s dependencies, tasks, and target platform.

​
Add dependencies

Install packages from the conda ecosystem: 
jnc add python numpy opencv
This resolves compatible versions, updates the manifest, and creates a lock file (jnc.lock) for reproducibility.

​
Set system requirements

Specify the L4T version your workspace requires. This ensures dependency resolution picks packages compatible with your Jetson: Edit jnc.toml and add: 
[system-requirements]
l4t = "36"

​
Run commands

Run any command inside the workspace environment: 
jnc run python -c "import cv2; print(f'OpenCV {cv2.__version__}')"

​
Define tasks

Add reusable tasks to your manifest: 
jnc task add check-gpu "python -c 'import torch; print(torch.cuda.is_available())'"
Run them by name: 
jnc run check-gpu

​
Enter the environment shell

Start an interactive shell with all dependencies available: 
jnc shell
python -c "import numpy; print('Ready for development!')"
exit

​
What’s in the manifest?

After these steps, your jnc.toml looks something like: 
[project]
name = "my-jetson-app"
channels = ["conda-forge"]
platforms = ["linux-aarch64"]

[system-requirements]
l4t = "36"

[dependencies]
python = ">=3.12,<4"
numpy = ">=2.2,<3"
opencv = ">=4.10,<5"

[tasks]
check-gpu = "python -c 'import torch; print(torch.cuda.is_available())'"

​
Next steps

  • Basic Usage — common workflows at a glance
  • Environments — manage separate dependency sets for inference, training, etc.
  • Tasks — advanced task configuration
  • CLI Reference — complete command documentation