Create and run custom pipeline steps using the pypeline framework. Use when creating new PipelineStep classes, accessing ExecutionContext (inputs, data_registry, env_vars), or configuring steps in pypeline.yaml. Also use when a user wants to model any sequence of operations as a custom pipeline — PipelineLoader and PipelineConfig are generic (PipelineLoader[T]) so users can define their own step base class, implement concrete steps, and run them with the pipeline machinery (shared context, dependency tracking, YAML config). Covers step configuration, data registry patterns, subprocess execution, custom pipelines, and running pipelines.
2026-07-11