Configuration
- Step — pick the step to run. Grouped by system in the picker; internal steps live in the “Internal” group.
- System integration — which tenant’s credentials the step uses. Required for steps that call remote APIs, ignored for local steps.
- Step options — typed form rendered from the step’s declared options. Defaults are materialized the first time you pick the step; cleared fields stay cleared.
- Cache mode —
Defaultreuses results from earlier in the same run when the step supports caching;Forcebusts the flow cache and asks the system to refresh any upstream caches it exposes. - On error — what happens if the step fails:
- Halt flow (default): stop the whole run.
- Halt branch: stop this path only; parallel paths keep going.
- Continue: log the error and keep running. Most useful for best-effort sync actions.
- Use real data in test runs — when on and the node lacks complete test fixtures, test runs execute the step for real, side effects included (writes to external systems, once per iteration inside loops). Complete fixtures always win and skip execution. When off, a test run errors if any declared assign lacks a fixture.
Error port
Wire the Error output port to a subgraph (e.g. an Anomaly node) to handle a failure. On error the subgraph runs as a detour — it records what its nodes emit (anomalies, assigns) — and only then does the On error policy apply: Halt flow aborts the run, Halt branch stops this path, Continue resumes the normal Out path. The error subgraph never runs when the step succeeds. Either way the error is logged and the run finishes aspartial, not ok.
Tips
- Steps declare what they read from
token.optsand what they write totoken.assigns. Downstream Flow nodes read those assigns by name. - Steps that write to an assign show the assign in the Evaluator and Settings panes so you can pick what to persist per run.