Pipelines

Preview

The Pipelines tab is in early access while the underlying orchestrator is finalized. List and detail views are read-only; you can't yet author or run pipelines from the UI. See troubleshooting for the known limitations.

A pipeline is a named DAG of steps. Each step is a containerized task — a Python function, a script, a shell command — wired with inputs, outputs, and conditions. Pipelines turn one-off scripts into reproducible, schedulable workflows.

Pipeline concepts

• Experiments — logical groupings of pipeline runs for comparison and tracking.
• Runs — individual pipeline executions with parameters, logs, and step-level status.
• Recurring Runs — cron-style triggers that kick off runs on a schedule.
• Artifacts — input and output data produced by pipeline steps.
• Executions — low-level step executions within a run.

See Anatomy for the full step model.

Next steps

• Anatomy — steps, inputs, outputs, and conditions
• Experiments — grouping runs
• Runs — individual executions
• Recurring Runs — scheduled triggers
• Artifacts — pipeline inputs and outputs
• Executions — step-level executions