Quickstart#

Important

drx is an unofficial, internal-only, experimental library; interfaces will change. We invite and encourage you to use and provide feedback, but ask that you not share externally and be mindful of the experimental, and unoffical nature when choosing to use on a project.

Install#

  1. Confirm you can access the drx repo. Please reach out to Marcus Braun or Marshall Krassenstein for access and be sure to provide your GitHub username.

  2. drx requires python >= 3.7

  3. Install drx from source:

    python -m pip install git+https://github.com/datarobot/drx.git

    drx dependencies will be automatically installed as needed.

Authenticating using HTTPS and the GitHub client

You will not be able to clone or pip install the drx repo without authenticating with GitHub. A simple way to authenticate is using the GitHub command client (if installed). From a terminal, execute the command gh auth login and follow the prompts to authenticate. Once complete you should be able to successfully clone or pip install the repo.

Authenticating using SSH

If you have previously setup SSH keys on GitHub and authorized them with DataRobot, you can install using:

python -m pip install git+ssh://git@github.com/datarobot/drx.git
Downloading and installing from source

Alternatively, you can install by first downloading the drx archived source from the GitHub web interface after logging in:

download source from github

After downloading the source, you can install using

python -m pip install drx-master.zip

Configure#

There are three options for configuring drx:

  1. Configuration with drconfig.yaml

  2. Configuration with environment variables

  3. In-line, programmatic configuration using drx.Context()

When imported, drx will automatically attempt to read and initialize configuration from ~/.config/datarobot/drconfig.yaml and the environment variables DATAROBOT_API_TOKEN and DATAROBOT_ENDPOINT. If parameters conflict, environment variables take precedence.

Optionally, the default prediction environment to use for deployments can be defined by setting the key pred_server_id in drconfig.yaml, or alternatively the environment variable DATAROBOT_PRED_SERVER_ID. If omitted, the first prediction environment returned by GET /api/v2/predictionServers/ will be used by default.

Programmatic configuration#

>>> import datarobotx as drx
>>> drx.Context()  # show current configuration
{'token': '***', 'endpoint': 'https://app.datarobot.com/api/v2', 'pred_server_id': '***'}
>>> drx.Context(token='foo', endpoint='bar')  # update current configuration
{'token': 'foo', 'endpoint': 'bar', 'pred_server_id': '***'}
>>> c = drx.Context()
>>> c.pred_server_id = 'foo_bar'  # alternative approach to update
>>> drx.Context()
{'token': 'foo', 'endpoint': 'bar', 'pred_server_id': 'foo_bar'}
>>> drx.Context(config_path='my_path/my_config.yaml')  # update from file
{'token': 'my_config_token', 'endpoint': 'my_config_endpoint', 'pred_server_id': 'my_config_pred_server_id'}

Use#

Check out some examples of what can be done with drx in the Model, and Consume sections and consider reading our FAQ. You can also check out the example notebook we have in the repo.

Upgrade#

python -m pip install --upgrade git+https://github.com/datarobot/drx.git