Connecting to MLOps

For interacting with MLOps we need to access the clients.

To interact with the MLOps platform, you will need to access the provided clients.

MLOps Codex offers three primary clients:

• mlops_codex.training.MLOpsTrainingClient Used for accessing and managing the training of Machine Learning models.

• mlops_codex.model.MLOpsModelClient Designed for handling model-related operations, including deployment and monitoring.

• mlops_codex.datasources.MLOpsDataSourceClient Enables integration and management of data sources for your ML workflows.

• mlops_codex.preprocessing.MLOpsPreprocessingClient Provides tools for managing and automating data preprocessing tasks in Machine Learning workflows.

• mlops_codex.external_monitoring.MLOpsExternalMonitoringClient Enables monitoring of deployed Machine Learning models trained on your own machine.

You need the server URL, an email and a password to access the MLOps. The best way to do it is using a .env file with the following env variables

MLOPS_URL='https://neomaril.datarisk.net'
MLOPS_USER='email@email.com'
MLOPS_PASSWORD='password@123'

If you create the .env file in the same directory where you are running your code, it will be automatically imported.

# Import the client
from mlops_codex.model import MLOpsModelClient
from mlops_codex.training import MLOpsTrainingClient
from mlops_codex.datasources import MLOpsDataSourceClient
from mlops_codex.preprocessing import MLOpsPreprocessingClient
from mlops_codex.external_monitoring import MLOpsExternalMonitoringClient

# Start the client via model client
model_client = MLOpsModelClient()

# Start the client via training client
training_client = MLOpsTrainingClient()

# Start the client via data source client
datasource_client = MLOpsDataSourceClient()

# Start the client via preprocessing client
preprocessing_client = MLOpsPreprocessingClient()

# Start the client via external monitoring client
external_monitoring_client = MLOpsExternalMonitoringClient()

Creating a group

Groups provide a way to organize training experiments and models that may have different end-users or purposes, enabling the creation of isolated environments for each group. When a group is created, a unique token is generated, which is used to run models and enhance platform security. This token will expire after one year. Every resource you create in MLOps should belong to a group, making the creation of a group the first step in your workflow. You can create a group using any of the available clients, simply by providing its name and a description to the group for better clarity and organization.

# Import the client
from mlops_codex.training import MLOpsTrainingClient

training_client = MLOpsTrainingClient()

training_client.create_group(
    name='nb_demo',
    description='Group for the demo'
)

# This token has a 1 year expiration date, to generate a new one use the refresh method

model_client.refresh_group_token(
    name='nb_demo', # Group name
    force=True # To force creating a new token even if the old is valid
)

Add your group token to the .env file:

MLOPS_GROUP_TOKEN='YOUR_GROUP_TOKEN'