dbt Test Generator

How it works

• Everything runs client-side — no network request, no upload, no server round-trip.
• Uses the same tokenizer as our SQL Formatter/Explainer/Scaffolder to find the SELECT column list, correctly ignoring parentheses depth and stopping at the next clause boundary.
• Columns with an explicit AS alias are named from the alias; bare column references use their own name; anything more complex (an un-aliased expression, or SELECT *) gets a TODO flag since we can't always know the exact output column name.
• Heuristics applied: a column named exactly id suggests unique + not_null. A column ending in _id (a likely foreign key) suggests not_null, plus an optional relationships test you can enable and point at the right target model/column yourself. Columns matching boolean-flag patterns (is_, has_, _flag, active...) suggest accepted_values: [true, false]. Columns matching enum-ish names (status, type, category, tier...) suggest an accepted_values test with a TODO placeholder list — we cannot know the real values without querying your data.
• Every checkbox is editable: uncheck anything that doesn't apply, or check tests we didn't suggest. accepted_values and relationships reveal extra fields when enabled.
• dbt ships four built-in generic tests: unique, not_null, accepted_values, and relationships. As of dbt 1.8 the YAML property is data_tests: (the old tests: key still works as a deprecated alias). As of dbt 1.10.5, test arguments can optionally nest under an arguments: block — older versions need them as top-level properties, which is why "top-level" is the default here. See dbt's data tests docs for the full reference.
• Only the first statement is scaffolded if your input has more than one (separated by ;).