End-to-End Dataset to Downstream Workflow#

This is a V0.29 recipe page over existing public APIs; it adds no runtime surface.

This recipe is the complete data-readiness path for users starting from an xarray.Dataset.

xarray.Dataset
-> Dataset readiness
-> FieldBatch conversion
-> FieldBatch readiness and residual preflight
-> generator confidence
-> downstream discovery workflow summary

Minimal Recipe#

  1. Build or receive an xarray.Dataset with one scalar data variable such as u.

  2. Run summarize_xarray_dataset_readiness(dataset, data_var="u", metadata=..., expected_equation=...).

  3. Convert with from_xarray_dataset(dataset, data_var="u", metadata=...).

  4. Run summarize_field_batch_readiness(field, residual_evaluator=..., expected_equation=...).

  5. Fit and verify the stable translation generator only inside the supported scalar 1D periodic scope.

  6. Summarize confidence with summarize_generator_confidence(...).

  7. Summarize bridge arrays and backend-neutral discovery results.

  8. Combine evidence with summarize_downstream_discovery_workflow(...).

Tutorial Page#

The rendered notebook 12_dataset_to_downstream_workflow.ipynb shows this recipe with saved outputs and a compact plot. It uses generated Heat data converted into a Dataset so that the example stays self-contained and reproducible.

Boundaries#

This workflow does not add file loaders, Zarr/NetCDF readers, PDEBench/The Well adapters, metadata inference, resampling, multidimensional data, nonuniform grids, or new PDE support.