Upload a pipeline

A pipeline is a compiled workflow definition stored as a YAML file. You create a pipeline in Python with the Kubeflow Pipelines SDK, compile it locally, then upload the resulting file to BullSequana AI. Once uploaded, the pipeline appears in the Pipelines list and you can run it any number of times with different parameters.

Before you start

You need a compiled pipeline YAML file. See the Kubeflow Pipelines SDK documentation for instructions on writing and compiling a pipeline locally.

Upload a pipeline

To upload a pipeline to BullSequana AI, follow these steps:

1. Select Pipelines in the left sidebar.

2. Select + Upload pipeline.

3. In the Upload pipeline dialog, fill in the required fields:

   • Pipeline name: a unique name for this pipeline.
   • Pipeline description: optional description of what the pipeline does.
   • Upload a file: select your compiled .yaml file, or provide a URL to a publicly accessible YAML file instead of uploading from your computer.

4. Select Create.

The pipeline appears in the Pipelines list. Selecting it shows the pipeline graph, where you can inspect each step and its inputs and outputs.

View a pipeline

Select any pipeline in the list to open its detail page. The detail page shows:

• The graph: a visual representation of the steps and their dependencies.
• The Pipeline spec tab: the raw YAML definition.
• The Runs tab: all runs started from this pipeline.

Next steps

• Create an experiment to group runs from this pipeline.
• Start a run to execute the pipeline with specific parameters.