Commands reference

Assumes the cluster and project have been successfully configured. Refer to Installation for details.

Running kfc command

Notice: Ensure your environment (e.g., Jupyter notebook that comes preinstalled with Kubeflow) has been appropriately configured to access Kubeflow Pipelines. Refer to details in Kubeflow documentation.

Once installed with pip install kfc, run kfc --help to print full command reference.

It is recommended to install a specific version of kfc: pip install kfc==<version>, the <version> should match the Kfops version installed in the cluster.

Running chatops commands

Notice: In-cluster (Argo) workflow that executes chatops commands has been already configured to access kKubeflow Pipelines.

Executed in the context of Pull Request.

Complete list with explanations:

  • /build - Compiles Kubeflow Pipeline using code in PR. Optionally, if was configured, builds and pushes to container registry images built.

  • /run - Executed Kubeflow Pipeline.

  • /build_run - /build and /run in single command.

  • /deploy - Deploys the model. Requires the pipeline to be already run in the same PR (otherwise it will report an error). If the run was executed more than once, /deploy will deploy the model from the last pipeline run. If you want to deploy a specific (Kubeflow Pipelines) run ID, use /deploy --run-id=<RUN-ID> where <RUN-ID> will be reported in PR after successfull pipeline execution.

  • /staging_deploy - Similar to /deploy but deploys ML model to the staging environment.