Python Dependencies via Pip
Last updated 20 June 2018
Table of Contents
This guide outlines how to fully utilize Heroku’s support for specifying dependencies for your Python application via
pip support is very transparent. Any requirements that install locally with the following command will behave as expected on Heroku:
$ pip install -r requirements.txt
Packages with unsupported C dependencies that do not provide “manylinux” wheels will not install on Heroku. Learn more.
To specify Python package dependencies on Heroku via pip, add a pip requirements file named
requirements.txt to the root of your repository.
Flask==0.8 Jinja2==2.6 Werkzeug==0.8.3 certifi==0.0.8 chardet==1.0.1 distribute==0.6.24 gunicorn==0.14.2 requests==0.11.1
If you follow these simple recommendations, your application builds will be deterministic:
- All package versions should be explicitly specified.
- All secondary dependencies should be explicitly specified.
This will ensure consistent build behavior when newer package versions are released.
Anything that works with a standard pip requirements file will work as expected on Heroku.
Thanks to pip’s Git support, you can install a Python package that is hosted on a remote Git repository.
If your package is hosted in a private Git repository, you can use HTTP Basic Authentication:
You can also specify any Git reference (e.g. branch, tag, or commit) by appending an
@ to your URL:
Optionally, you can install a dependency in “editable” mode, which will link to a full clone of the repository. This is recommended for Git-backed distributions that rely on upstream changes, as well as larger repositories.
egg fragment is only valid with editable requirements.
Remote file-backed distributions
You can also install packages from remote archives.
This can be useful in many situations. For example, you can utilize GitHub’s tarball generation for repositories with large histories:
Local file-backed distributions
Pip can also install a dependency from your local codebase. This is useful with making custom tweaks to an existing package.
You can use Git Submodules to maintain separate repositories for your File-backed dependencies. Git modules will automatically be resolved when you push your code to Heroku.
To add a local dependency in
requirements.txt, specify the relative path to the directory containing
If you make changes to the library without bumping the required version number, however, the changes will not be updated at runtime. You can get around this by installing the package in editable mode:
In order to minimize points of failure, it is considered best practice within the Python community for development shops to host their own instances of the “Cheeseshop” containing their dependencies.
To point to a custom Cheeseshop’s index, you can add the following to the top of your requirements file:
All dependencies specified in that requirements file will resolve against that index.
Cascading requirements files
If you would like to utilize multiple requirements files in your codebase, you can include the contents of another requirements file with pip:
We do not recommend this approach, but it is available to you.
Heroku also supports traditional Python package distribution, powered by
If your Python application contains a
setup.py file but excludes a
python setup.py develop will be used to install your package and resolve your dependencies.
This works best with setuptools. Projects that use distutils directly will be installed, but not linked. The module won’t get updated until there’s a version bump.
If you already have a requirements file, but would like to utilize this feature, you can add the following to your requirements file:
This causes pip to run
python setup.py develop on your application.
Scientific Python Users
If your application utilizes obscure dependencies (scipy, scikit-learn, etc), you can use this example application as an easy starting place to deploy your application to Heroku today, utilizing the power of our beta Docker support and Continuum’s powerful Miniconda package manager: