Reviewed reuse
Reusable data cleaning recipes need more than a saved prompt
Save the rules from a successful run, then check a profiled target before reuse. A recipe creates a draft plan, not an automatic cleanup. You still review the rules, acknowledge high-risk changes and approve the execution quote.
By the TidyX team · Updated
What a versioned recipe preserves
A recipe version records the successful run rule snapshot, schema fingerprint, column definitions and saved context. It provides a starting point for another project without overwriting an earlier run or its output.
This design keeps reuse explainable. A user should be able to see which version was selected, what changed between versions, and which frozen snapshot was ultimately confirmed for a run.
Schema drift must interrupt automatic reuse
Choose a profiled target project and check compatibility. The service compares the saved schema with the incoming columns and types. Missing or changed fields require review instead of silently assuming that the old rules fit.
An exact match can create a draft for review. Blocking mappings must be resolved before applying the recipe. The resulting rules are validated against the target profile and safeguards before they can be confirmed.
- Exact-match, review-required, blocking and resolved states
- Explicit source-to-target column mapping
- Review categorical mappings and unfamiliar values explicitly
- No automatic execution after a preview match
Reuse still requires a new decision
After a successful cleaning run, use Save as recipe in the report. Open Recipes, select the saved recipe and choose a profiled target project. Check compatibility before creating a draft plan.
Review the new draft and acknowledge enabled high-risk rules. A new cleaning run has its own credit quote; saving or applying a recipe does not bypass confirmation or guarantee that the same choices are suitable for a different dataset.