Skip to main content
The Assertion node checks whether a value meets a condition. Pick an operator from the dropdown and the relevant inputs appear: Contains, Equals, GreaterThan, GreaterThanOrEqual, IsEmpty, IsFalse, IsTrue, LessThan, LessThanOrEqual, NotContains, NotEmpty, NotEquals, NotRegex, OneOf, Range, Regex.

​
Fail-fast

By default, assertions record a pass or fail result and the workflow keeps going. Enable the fail-fast checkbox and a failed assertion will abort the entire workflow run immediately. Assertions don’t have an output handle – they’re observers, not gates. They watch data flow past and check whether it meets a condition. With fail-fast off, they quietly record the result. With fail-fast on, they pull the emergency brake on the whole run. If you need to route data based on a condition rather than abort, use an If node.

​
In evaluations

Each assertion carries a weight (0 to 1). The evaluation’s aggregate score is a weighted average of all assertion outcomes – giving you a single number that captures how well your prompts and workflows performed across the dataset. Keep fail-fast off so every row gets evaluated.

​
In regular workflows

With fail-fast enabled, assertions become tripwires: if something has gone wrong, stop everything rather than letting the rest of the workflow act on bad data. Useful for sanity checks like verifying that a generated summary stays under a length threshold, or that a classification result is one of the expected categories.