Run a pipeline
A pipeline run is a single execution of a pipeline with a specific set of parameters. Each run tracks per-step status, logs, metrics, and output artifacts.
Prerequisites
- An uploaded pipeline (see Create a pipeline)
Trigger a run from the SDK
from vantage_sdk import VantageClient
client = VantageClient()
run = client.pipelines.run(
pipeline_name="training-pipeline",
params={"data_path": "/data/raw/dataset.csv"},
experiment="nightly-training",
)
print(f"Run ID: {run.id}")
Monitor a run in the UI
- Click Workbench in the left sidebar, then click Pipelines.
- Click the pipeline name, then select the Runs tab.
- The runs list shows:
| Column | Description |
|---|---|
| Name | Display name of the run |
| Pipeline | Pipeline version |
| Status | Running, Succeeded, Failed, or Skipped |
| Started | Timestamp when the run began |
| Duration | Elapsed time |
- Click a run row to open the detail view.
Inspect run details
The run detail page shows:
- DAG visualization with color-coded step status (green = succeeded, blue = running, red = failed, grey = skipped)
- Step details -- click any step node to view its inputs, outputs, container image, and logs
- Parameters -- the input parameters used for this run
- Metrics -- output metrics produced by the pipeline steps