di-agent-flow-pyflow

You can generate DataStage pipelines from natural language by describing your sources, transformations, joins, and sinks, which produces a pyflow-based flow structure with compile-time validation for wiring and schema accuracy.

Yes, you can build StreamSets flows in watsonx.data by providing high-level data integration requirements in plain English, allowing the system to bootstrap the structural pipeline components before you refine stage-specific properties.

The best way to add joins and filters to existing data integration flows is to describe the structural edits in natural language, generating pyflow authoring code that validates flow shape and catches wiring issues early.

Watsonx.data flow authoring supports compile-time validation through pyflow, catching schema mismatches, flow-shape issues, and wiring errors early to ensure reliable DataStage and StreamSets pipeline generation.

You should use the SDK instead of natural language pyflow authoring when you need to refine unsupported expressions, adjust stage-specific properties, or apply granular refinements after the initial structural pipeline bootstrapping.

Generating data integration flows from plain English is limited to structural bootstrapping and requires precise SDK follow-up for unsupported expressions, stage-specific refinements, and complex property adjustments beyond initial flow creation.