Automatic AI Assistance
Some AI features are not something you invoke. They run as part of a screen you are already using and put their results in front of you. There is nothing to switch on and no prompt to write.
Schema checking on a trace
When you inspect a step's output after a real run, the platform compares what the step actually returned against the schema it declared, and surfaces the discrepancies.
Where: the workflow trace, when viewing a step's output after a run.
- Go to Metrics > Traces.
- Select the run you want to inspect.
- Open the step whose output you want to check.
This matters because a schema that drifts from reality fails quietly. The workflow keeps running, and the mismatch only surfaces later in something downstream that depended on the declared shape.
What to prioritize
Not every discrepancy is worth acting on. A field that is present in the data but missing from the schema is usually the one to fix first, because anything mapping from it is working without a contract.
Activity metadata and mappable fields
When you create activities from an OpenAPI spec, the platform generates the activity's metadata and its mappable fields together, in a single pass for every endpoint you select.
Where: activity creation from an OpenAPI spec, at the Add Details step.
These two always run together — you never trigger one without the other.
Work in small batches
| Practice | Why it helps |
|---|---|
| Start with a few endpoints | Confirms the results look right before you commit to a large batch |
| Check that list fields expanded properly | A list should show the structure of its items, not just that it is a list |
How you know it worked
Schema discrepancies appear alongside the step output in the trace. Generated activity metadata and mappable fields appear in the Add Details step, ready for you to review before continuing through the wizard.
Next steps
- Creating Activities from an OpenAPI Spec — the full activity creation reference.
- Operating and Monitoring — working with traces and metrics.