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