Skip to main content
The forward output hands part of a document off to another workflow. Each classified segment becomes a new document in the target workflow, with its own run. It is a terminal node: nothing continues after it here, and the target workflow runs independently. Use it to split a mixed bundle: classify the pages, then forward each segment to the workflow that specializes in it.
To reuse another workflow’s steps on the document you already have, use a Call workflow action instead. Forwarding creates a new document and a separate run; calling keeps one document and one run. See Reusable sections.
The target workflow must contain a Forward trigger to receive a forwarded document. Workflows without one do not appear in the picker, and a forward whose target loses its trigger later fails with a message naming that workflow. It never falls back to another trigger.

​
Configuration

Only workflows with a forward trigger appear, in the same environment, and never the workflow you are editing. Two forward nodes in one workflow may not target the same workflow. One classification producing two documents in the same place is almost always a mistake, so Ingestly rejects it when you save. (A Call workflow action has no such rule: reuse is its purpose.)

​
What gets forwarded

A forward is fed by a classification: connect it to one of a Classify node’s outputs, and it sends the pages that took that output to the workflow it names. Each group of pages is its own document, so a bundle whose pages classify as two separate invoices produces two documents, each carrying only its own pages. The source is split per segment, which requires a PDF. A non-PDF can still be forwarded whole, when the classified pages cover the entire document; otherwise the step fails and says so rather than silently forwarding the wrong bytes. If no pages reached this forward, the step is skipped: that is a real decision, not a failure. If there is no readable classification to act on at all, the step fails, because a forward with nothing to read has dropped data.

​
How forwarding works

  1. This workflow reaches the forward output node
  2. One document is created in the target workflow, carrying the pages that reached this forward
  3. Each target workflow starts its own run through its forward trigger
  4. The runs are recorded separately. Open each from the Runs page
  5. If a target run fails, or cannot start, the failure is recorded back onto this forward step
Because each dispatch is its own document and run, a retry of this step reuses the documents it already created rather than duplicating them.

​
Chain depth

Forward chains are capped at a depth of 10, so a loop assembled across workflows (A forwards to B, B forwards back to A) stops instead of creating documents forever.

​
Inputs and outputs

Allowed inputs: any action node. In practice a Classify action supplies the segments. Output: terminal node, no downstream connections.

Forward trigger

Receive a forwarded document

Classify action

Classify pages into the segments a forward dispatches

Call workflow action

Reuse another workflow’s steps without handing the document over

Callback output

Send results to an external webhook instead