Skip to main content

​
Environments

A jnc workspace can contain multiple environments, each composed of one or more features. This system lets you maintain separate dependency sets for different purposes (development, testing, production) within a single workspace.

​
Default environment

Every workspace has a default environment that includes the default feature. When you run jnc add python, the package is added to the default feature and is available in the default environment. 
[dependencies]
python = ">=3.12,<4"
numpy = ">=2.0,<3"

​
Features

A feature is a named set of dependencies, tasks, and configuration. Features are the building blocks of environments. 
[feature.test.dependencies]
pytest = ">=8.0,<9"

[feature.docs.dependencies]
sphinx = ">=7.0,<8"
Add packages to a feature: 
jnc add --feature test pytest
jnc add --feature docs sphinx

​
Composing environments

Environments are composed by combining features. Define them in the manifest: 
[environments]
test = ["default", "test"]
docs = ["default", "docs"]
The test environment includes everything from both default and test features.

​
Using environments

Target a specific environment with most commands: 
jnc run --environment test pytest
jnc shell --environment docs
jnc list --environment test
jnc install --environment test

​
Environment installation

When you run jnc install, all environments are solved and installed. Each environment gets its own isolated set of packages. Packages shared across environments are deduplicated on disk using hard links or reflinks.

​
Platform support

Each environment inherits the workspace’s target platforms. You can also specify platform-specific dependencies within features: 
[target.linux-aarch64.dependencies]
cuda-toolkit = ">=12.0,<13"

[feature.gpu.target.linux-aarch64.dependencies]
pytorch-cuda = ">=2.0,<3"