Cleaner Agent scan modes

Understand scan types, thresholds, and how to read cleanup recommendations.

Cleaner Agent supports two scan modes.

Choose the mode that matches how strict you want duplicate detection to be.

Use Semantic when you want to find near-duplicates, not just identical wording.

This mode compares intent and similarity across test cases. It works best when different teams describe the same flow in slightly different ways.

Semantic mode also lets you set a similarity threshold.

Lower thresholds return broader matches. Higher thresholds return fewer, stricter matches.

If you set the threshold very high, near-duplicates may no longer appear in results.

Use Exact when you want to find cases that match word for word.

This mode is stricter and usually faster.

It works well when you want to clean up direct copies or repeated imports.

You can scan either:

A journey scan only compares test cases in that scope.

A full scan compares the entire project.

Each duplicate group includes a recommendation for every test case.

Keep marks the best version to retain.

This is usually the most complete or most useful test case in the group.

Delete marks a test case as redundant.

Use this when the case adds no unique value once the primary version is kept.

Review means the test case is similar, but still contains something worth preserving.

In most cases, you should merge that useful detail into the primary version before deleting the original.

Choose Merge when multiple test cases cover the same flow but each one adds useful detail.

This is common when one version has stronger preconditions and another has better step coverage.

Do not delete a case marked Review until you confirm its unique detail is preserved.
Do not expect Exact mode to find similar wording or shared intent.
Do not assume a no-result scan means the repository is clean. The scope or threshold may be too narrow.

Return to Cleaner Agent for the end-to-end workflow.

Last updated 3 months ago

Was this helpful?