Deployment
==========

Containerization
----------------

All processes (including models) in the MISTK ecosystem run as Docker containers
and require a **Dockerfile** to configure the runtime image.  The **Dockerfile**
below runs the Logistic Regression model example.
It includes the essentials for running within MISTK that can be used for any model,
however models are free to add any customizations to their runtime image:

.. literalinclude:: sample_code_files/Dockerfile
    :linenos:
    :language: guess

The final CMD line starts the model and sets up its endpoints.  The third
argument (`scikit_learn.logistic_regression`) is the package/module of our model
and the fourth argument (`ScikitLearnLogisticRegressionModel`) is the model class name.

To update this Dockerfile for a new model, a model developer would update the install path
on lines 14-16 and the model path and name on line 21. The MISTK package
version denoted by <VERSION> must also be updated on line 6.

The setup.py script referenced above is a standard python setup install script
which lists the python package dependencies for the model:

.. literalinclude:: sample_code_files/setup.py
    :linenos:
    :language: python