Rattler OCI Registry

Repository Path Format Reference

Conda Channels

1. Simple Format

channel/package1/package2/...

Basic format with channel and package names.

Example: Single package from conda-forge

docker pull /conda-forge/boltons

Pulls the boltons package from conda-forge channel.

Example: Multiple packages

docker pull /conda-forge/boltons/numpy

Pulls both boltons and numpy packages.

2. With pkgs Separator

channel/subchannel/pkgs/package1/package2/...

Format with subchannels and explicit pkgs separator.

Example: Subchannel with packages

docker pull /conda-forge/robostack/pkgs/boltons/numpy

Pulls packages from the conda-forge/robostack subchannel.

3. With Platform

channel/subchannel/platform/pkgs/package1/package2/...

Format with explicit platform specification.

Example: Linux 64-bit platform

docker pull /conda-forge/linux-64/pkgs/boltons

Pulls boltons for Linux 64-bit platform.

Example: OSX ARM64 platform

docker pull /conda-forge/osx-arm64/pkgs/numpy

Pulls numpy for macOS ARM64 (Apple Silicon).

GitHub Repositories

4. GitHub Format

github/owner/repo[/platform][/dev]

Format for fetching packages from a GitHub repository's pixi.lock file.

Example: Basic GitHub repository

docker pull /github/prefix-dev/pixi

Fetches packages from pixi.lock in the main branch (production mode).

Example: GitHub with dev mode

docker pull /github/prefix-dev/pixi/dev

Fetches from main branch in development mode.

Example: GitHub with platform

docker pull /github/prefix-dev/pixi/linux-64

Fetches with Linux 64-bit platform specification.

Example: GitHub with platform and dev mode

docker pull /github/prefix-dev/pixi/linux-64/dev

Fetches with platform and development mode enabled.

Note: GitHub format always fetches the pixi.lock file from the main (or master) branch of the repository and uses the "default" environment.

Complete Example: Running Python with NumPy

Here's a complete example showing how to run Python with NumPy in an interactive container.

Run Python with NumPy in interactive mode

docker run -it /conda-forge/python/numpy python

This command pulls Python and NumPy from conda-forge and starts the Python REPL directly.

Inside the Python REPL, import and use NumPy

>>> import numpy as np >>> print(np.array([1, 2, 3])) [1 2 3]

NumPy is ready to use! You can now perform array operations and numerical computations.

Tip: You can add any conda packages to the path. For example, use /conda-forge/python/numpy/pandas/scipy to get multiple data science packages in one container.