Software Environments¶
A crucial part of doing your work is making sure you have the software packages you need. Coiled helps you manage software environments by building Docker images from pip and conda environment files for you. You can then use these environments locally, remotely in a Dask cluster, and can share them with your friends and colleagues.
Supported software specifications¶
Coiled supports publicly accessible conda packages, pip packages, and/or Docker images for creating software environments. You can also compose these steps by, for example, conda installing packages into a custom Docker image.
Design¶
Coiled uses packaging conventions you’re already familiar with. You can point Coiled at a list of conda or pip pacakges:
import coiled
coiled.create_software_environment(
name="my-software-env",
conda=["dask", "xarray=0.15.1", "numba"],
)
or to a local conda environment.yml or pip requirements.txt file:
coiled.create_software_environment(
name="my-software-env",
conda="environment.yml",
)
to have custom Docker images built and stored for later use. Note that you do not need to have Docker installed for Coiled to build Docker images for you!
Usage¶
Coiled software environments can be used both locally by installing the software environment on your machine and on remote Dask clusters (e.g. running on AWS):
import coiled
# Create a cluster that uses the custom "my-software-env" software environment
cluster = coiled.Cluster(software="my-software-env")
You can also collaborate with your friends and colleagues by easily sharing software environments.