Welcome to dstack-factory’s documentation!

Contents:

dstack-factory consists of two components: (1) factory and (2) runtime.

factory is a service that builds (or downloads) python wheels from a pip requirements file or from a wheel package uploaded to the wheelhouse. If a python package is already available as wheel package, it is cached. Otherwise the source is downloaded from pip, git or any other source supported by pip and compiled/packaged as a wheel file.

Examples of python packages that currently still need to be build are:
  • psycopg2
  • lxml
  • pycrypto
  • pillow
  • weasyprint (and it’s dependencies)
  • anything hosted on GitHub, e.g. forks of python packages or unreleased versions

runtime is a base docker image that contains all the necessary libraries to run most popular python packages with external dependencies. For example pandas, matplotlib, weasyprint, etc. all require non-python libraries. In addition to the base runtime docker image, dstack-factory also provides three entry point Dockerfiles to support different workflows for building and deploying docker and python based applications.

Indices and tables