Hamilton

Decompose features into individual Python functions.

Use Hamilton Driver to select only needed features.

Run on Ray for parallel execution.

Store outputs in a Feature Store.

Define embedding logic and retrieval logic as separate Hamilton nodes.

Swap LLM models by changing a single module import.

Inject data validation to check prompt quality.

Visualize the retrieval flow for audit.

Map regulatory requirements to specific Hamilton functions.

Auto-generate the calculation graph for auditors.

Use lifecycle hooks to log every intermediate value.

Export lineage as a PDF report.

Encapsulate notebook cells into named functions.

Organize functions into modules.

Test modules in isolation with Pytest.

Execute using the Hamilton Driver in a CI/CD pipeline.

Attach @check_output decorators to functions.

Define Pandera schemas for dataframes.

Configure Hamilton to raise errors on schema violations.

Log validation results to DAGWorks Cloud.

Use @config.when to define experimental logic branches.

Run both versions through the same Driver.

Compare results in the final node.

Switch to the winner via config change.

Write transformations in standard Python/Spark API.

Wrap logic in Hamilton functions.

Use the PySparkGraphAdapter to distribute workload.

Monitor task execution across the cluster.