Alternative to Conda Installers#
Anaconda Distribution,
Miniconda, and Miniforge are the familiar ways many users get conda. cx is
not a demand to move away from them. It is an alternative bootstrap path for
users who want a native executable, a locked first install, and an opinionated
base prefix.
What stays familiar#
After bootstrap, cx delegates ordinary commands to the installed conda
executable:
cx create -n analysis python=3.12 numpy pandas
cx install -n analysis scipy matplotlib
cx list -n analysis
cx env list
The packages still come from conda channels. The default channel is
conda-forge, and the installed prefix is a standard conda prefix on disk.
What changes#
The base prefix is managed more strictly than a typical Anaconda Distribution, Miniconda, or Miniforge base:
cxbootstraps into~/.conda/expressby default.The base prefix is frozen after bootstrap, so day-to-day work should happen in named environments.
Activation can use
cx shell ENV, powered by conda-spawn, without requiringconda init.
Try cx side by side#
You can try cx without changing an existing Anaconda Distribution,
Miniconda, or Miniforge installation:
cx bootstrap
cx create -n cx-test python=3.12
cx shell cx-test
python --version
exit
cx manages its own prefix. Do not point it at an existing Anaconda
Distribution, Miniconda, or Miniforge base prefix.
Recreate an existing environment#
If you have an environment.yml, pass the command through cx:
cx env create -f environment.yml
cx shell my-environment
If the file does not name the environment, choose one explicitly:
cx env create -n analysis -f environment.yml
cx shell analysis
When to keep using an installer distribution#
Keep using a traditional installer when that is the approved path for your
organization, classroom, cluster image, or vendor workflow. cx is most useful
when you want the conda CLI with a smaller bootstrap artifact, a locked
first-run setup, a frozen base prefix, or an offline cxz binary.