/conda-buildpack

Primary LanguageShellOtherNOASSERTION

Conda Environment Buildpack

This is the Heroku Buildpack for Conda using it's new environment spec. Anything you can install with conda install can be installed using this in the post_containerize script.

Usage

To control what gets installed, create an environment.yml file in the root of your repository. For example, if you wanted to install Flask, you would add this:

name: myproject  # overridden at install, so this is for your use with conda env
dependencies:
  - flask

To run a server on your ship create a Procfile that is one line and looks like:

web: source activate /home/ubuntu/app/.heroku/miniconda/envs/heroku-env; gunicorn server:app --bind 0.0.0.0:$PORT --timeout 90;

to use your custom code change only the gunicorn part, the activate has to be done to use the conda env you built from your environment.yml (this buildpack will name it heroku-env)

Using a package from different GLG private git repository

In environment.yml

...
  - zlib=1.2.11
  - pip:
    - "-e git+git@github.com:/glg/ai-cortex.git#egg=ai_cortex==0.5.0"
    - click==7.0
 ...

For staging/production

Using python package from private github repo devship/production environment

Modify orders and add after_containerize file. Here is an example.

  • In orders: 
    # AI Team conda buildpack:
export BUILDPACK_URL=git@github.com:glg/ai-team-conda-buildpack.git

# Required if using python/js packages from internal private repos:
export GIT_COMMAND="sudo starphleet-git"

# Required if using python packages from internal private repos to
# avoid a possible hang-causing confirmation:
export PIP_EXISTS_ACTION=w

# More orders stuff
# ...

Sample Project: