Skip to content
FlowArray help

The Data tab: rehearse a run for $0

See exactly what every item will send, with your values already filled in, before a single provider call is made — and the real outputs in the same place afterwards.

In this article (5)

The single cheapest habit in FlowArray. A batch of two hundred generations is real money; looking at what it is about to send costs nothing.

Where it is

In Map view, click any card — the settings panel on the right has Set up and Data. Open Data and step through your items.

What it shows before a run

  • The resolved input for each item: your instruction with every placeholder already substituted, exactly as the provider will receive it.
  • The images that item will send, as thumbnails.
  • For a tool step, the arguments it will be called with.
  • For a Decide or Repeat card, which route each item will take.

It is not an approximation. The preview runs the same substitution code the engine executes, so what you read is what will be sent.

What it shows after a run

The same tab fills with real outputs per item — text rendered as formatted text (with an Aa / </> toggle for the raw markdown), images as thumbnails, tool results as structured data. Model, image, video and tool cards also carry their own output on the card face, so a finished map reads like a contact sheet.

What to look for

  1. Literal braces. {prodcut_name} in the resolved text means a typo — the item's real value is not going in.
  2. Empty where you expected content. Usually a column that exists on the sheet but not on the batch card, or the other way round.
  3. The wrong images. Check whether the step should be using the item's originals or the previous step's output.
  4. Routes on a Decide card. If every item is heading down the otherwise path, the condition is not matching what you think it is.

The related free rehearsal

If you have no keys in yet — or want to test a chain's shape without paying for it — run the flow through the built-in free test mode instead: it produces placeholder output at $0. It proves the wiring, not the quality. See mock runs.