Classification

Not yet checked on the product

This chapter was written from the product's source code and has not yet been checked by a person on a running Cogneris. If what you see on screen differs, the screen is right — please tell your Cogneris contact so the chapter can be corrected.

Classification identifies the type of each document of the process. It sends the documents to the Cogneris Vision service, which returns a document type, a confidence and, when one of your company's templates covers that type, the template.

Palette	Nodes › Classification
Outputs	Normal and Alternative. Alternative is taken when a document is below the acceptance threshold or was not classified — only when it is connected. See Outputs.
Content status	From source; not observed (workflow-fe f88c5c1, workflow-be 6703373, 2026-10-22)

When to use it

• Before Extraction, so that extraction follows the fields your template declares instead of letting the AI name them.
• Before a Manual Review whose queues are organised by document type: classification records each document's type and template, and the review uses it.
• Before a Conditional that routes on the document type.

If you only need the type to route right after extraction, the Detect document type option of Extraction does it without a separate phase.

Classification accepts only one incoming link. The editor does not draw a second one.

What happens in a run

1. The phase reads the documents from the output of the phase that leads into it (from Start, these are the process documents). Documents that already failed in an earlier phase are left out and carried forward as skipped.
2. It sends all remaining documents to the Vision service in one call.
3. For each document it records the result: the document type, the confidence, the pages and the template of your company that covers that type. The template is empty when no template covers the type. Results are matched to documents by file name, or by order when the service returns no name.
4. It stamps the type and the template on the document of the run. Manual review uses this to send each document to the queue of its template type.
5. The output is one result row per document (document, classification, and skipped for a document left out) plus the list of documents, unchanged. The next phase works on the same files.

A later phase or condition reads these values at results[N].classification.documentType, results[N].classification.confidence and results[N].classification.templateId, where N is the document's position starting at 0.

When every document had already failed, the phase completes without calling the Vision service and every row is marked skipped.

Failures

Classification is a single call for all documents, so an error fails the phase as a whole — never a single document.

• A temporary failure of the Vision service (server error, timeout, connection error, too many requests) is retried up to 3 times inside the phase. When the service asks to wait at most 10 seconds, the phase waits that long. If it still fails, the phase is run again, up to 2 more times.
• An error answered by the Vision service, or a permanent refusal, fails the phase and the process at once.

Outputs

The node has two outputs, Normal and Alternative. When Classification completes, the run takes one of them:

Result	Output taken
Every classified document has a confidence at or above the Classification Acceptance Threshold	Normal
At least one document is below the threshold, or the classifier returned nothing for it	Alternative
Alternative is not connected	Normal, whatever the confidence — the threshold only routes when there is somewhere to go

The phases on the output not taken are skipped. A link drawn without choosing an output counts as Normal. Documents that already failed before this phase do not count.

What changed. Before this version, the service did not read the threshold and ran the phases on both outputs. A published workflow whose Classification has both outputs connected now runs only one of them.

To route on the document type, or on anything other than confidence, connect Classification to a Conditional and branch on its True and False outputs.

Configuration

Select the node to open Phase Configuration. Its header shows the phase type as CLASSIFICATION.

Field	What it does	Required	Format / values	Default	Validation messages
Description	The name shown on the node.	No	Free text. Left empty, the node keeps its previous name.	Classification	—
Classification Acceptance Threshold	The minimum confidence for a document to go on through Normal. The hint reads Minimum confidence (0–100%; 80 if left untouched). A document below it, or not classified, takes the Alternative output — only when Alternative is connected; otherwise the run continues through Normal. It does not change the classification itself, only the output taken.	No	Whole or decimal number. A value below 0 is saved as 0, above 100 as 100. Left empty, the field is removed.	80 — also when the field was never changed and nothing is saved.	—

There is no template or document type to choose here. The Vision service decides the type and returns the template of your company that covers it.

Common problems

Message	Cause	What to do
(none — the link is not drawn)	You tried to connect a second phase into Classification. It accepts one incoming link.	Remove the existing link first, or place the phases in sequence.
Auth token missing for run; cannot call the document service. (service text, run error)	The process carries no credential for the Vision service.	Start the process again. If it persists, contact Cogneris support.
No documents found for classification. (service text, run error)	The phase received no documents.	Check that the phase leading into Classification passes documents on, for example that it is linked from Start.
Classifier call failed: … (service text, run error)	The Vision service answered an error, or stayed unavailable after the retries.	Check that your company has templates for the documents you receive. If the text names a temporary failure, start the process again.
A document has an empty template	No template of your company covers the type that was detected.	Create or adjust a template for that document type. Without it, Extraction names the fields itself.
Every process goes to Alternative	The threshold is above the confidence the classifier gives your documents (80 unless you changed it).	Lower Classification Acceptance Threshold, or check the classifier's training.
A low-confidence document went on through Normal	Alternative is not connected, so the threshold does not route.	Connect Alternative to the phase that should handle those documents.

Related

• Extraction — uses the template found here
• Zero Shot — type and content without a template
• Conditional — route on the type or the confidence
• Organize human review
• Design and publish a workflow

---

Last reviewed 2026-10-22 · Version 1.0.0 · Owner: product-documentation