Chain SQL steps into a pipeline, set the schedule, and Sprinkle wires the dependencies, runs the tests and alerts you when something breaks. Python when SQL is not enough. And when you would rather describe the transform than write it, the agent drafts it for you to review.
Every transform runs as SQL or Python inside your warehouse. You build the steps, Sprinkle handles dependencies, scheduling, incremental loads, tests and alerts. No orchestrator to run and no YAML to write.
A transform is a SQL step over the tables already in your warehouse. Preview the result, name the output, choose full or incremental materialization.
Sprinkle reads the references between steps and orders the run. Switch to Python when you need pandas, ML or anything SQL cannot express.
"Monthly revenue by region from paid orders" is enough. The agent knows your schema and drafts the SQL, the tests and the schedule. You review the diff before anything runs.
As you add each transform, Sprinkle reads its upstream references and slots the new node into your dependency graph. No manual wiring, no YAML, no broken refs.
Add assertions to any step: uniqueness, nullability, referential integrity, accepted values. Sprinkle proposes them from the schema, you confirm. The pipeline pauses and alerts you if they fail.