Documents Classify Bundle

Sends a file with several documents (a bundle) whose split is not trustworthy to the classification station, along with the segments IDP Classify proposed. Instead of extracting blindly, a person confirms where each document starts and what type it is.

Use it when IDP Classify returns splitNeedsReview = true.

Documents Classify Bundle   path      = = file
                            reference = = file
                            segments  = = doc["segments"]
                            pages     = = doc["pages"]
                            dpi       = 200

When to use it

A split is uncertain when there are unrecognized pages, or when the file was split into several documents with no printed marker backing it (only the model’s judgment). IDP Classify flags it with splitNeedsReview. The typical pattern:

If   condition = = doc["splitNeedsReview"] == true
     then:
       Documents Classify Bundle   path = = file, reference = = file,
                                    segments = = doc["segments"], pages = = doc["pages"], dpi = 200
       Continue

Parameters

Parameter	Type	What it is
path	text	The bundle file’s path.
reference	text	Stable key for the bundle (e.g. the file name).
segments	list	The segments proposed by IDP Classify — doc["segments"].
pages	list	The per-page dimensions — doc["pages"] — so the station can render the pages.
dpi	int	The dpi the pages were read at.
retentionDays	int	How many days to keep the bundle. Empty = the environment default.

What happens next

The bundle appears under Documents → Classify in the environment. A person adjusts the boundaries/types and confirms. From there, Documents Take Confirmed picks it up to extract each segment.

Related activities

• IDP Classify — classifies per page and proposes the segments.
• Documents Take Confirmed — picks up the already-confirmed bundles.

The OCR cache

Also pass document (the full object IDP Classify returned) and the reading — words, pages, text — is stored next to the bundle as an OCR cache. When a person confirms the split and Take Confirmed Bundles picks the bundle up, that reading comes back with it and each segment is extracted without re-reading the file or paying Azure again. Without document everything still works — extraction just re-reads.