Redact Document Smart Service

What the Redact Document smart service does

The Redact Document smart service permanently removes sensitive information from a PDF document using the rules configured in a Document Redaction AI skill. The Redact Document smart service takes a PDF as its input, detects sensitive text using the skill's keyword, regular expression, and semantic rules, and returns an image-only PDF with the detected text permanently removed.

Redact Document permissions

The user who starts the process does not need access to the AI skill called by the Redact Document smart service. Without those permissions, some smart service configuration options are masked.

To view the redacted document, the user who starts the process must have permission to the input document and to the folder where the redacted document is saved.

Redact Document properties

• Category: AI Skills
• Icon:

• Assignment Options: Unattended

• Compatible with autoscale: No

Redact Document configuration options

Setup tab

The Setup tab of the Redact Document smart service lets you select an AI skill:

• Select AI Skill: Search or browse for a Document Redaction AI skill to use with this smart service.

Data tab

Node inputs

The Redact Document node accepts the following inputs. The Semantic Instructions, Keywords, and Regex Patterns inputs override the corresponding rules configured in the AI skill at runtime.

Name	Data Type	Multiple	Required	Description
Document	Document	N	Y	The PDF document to redact.
Folder	Folder	N	Y	The Appian folder where the redacted document is saved.
Usage Groups	List of Group	Y	N	Groups to associate with this AI execution for usage attribution. If left empty, the execution is logged as unattributed.
Semantic Instructions	Text	N	N	Overrides the semantic instructions configured in the AI skill. Used only for semantic matching.
Keywords	List of Text	Y	N	Overrides the keyword list configured in the AI skill.
Regex Patterns	List of Text	Y	N	Overrides the regular expression patterns configured in the AI skill.

Node outputs

The Redact Document node returns the following outputs.

Output	Data Type	Multiple	Description
Redacted Document	Document	N	The redacted Appian document, named redacted_<original-name>.pdf and saved to the specified folder. The output is an image-only PDF with the detected text permanently removed.
Matches	Document	N	A document containing the details of each redaction match in JSON format, including the method that detected it, the matched keyword or pattern, the bounding boxes, and the page number.
Success	Boolean	N	Returns true if the smart service executed successfully.
Error Response	Text	N	An informational message returned when an error occurs during execution.
AI Actions Consumed	Number (Integer)	N	The number of AI Actions consumed by the execution.
Model	Text	N	The generative AI model used for semantic matching.

Redact Document usage considerations

Usage limits

The Document Redaction AI skill and the Redact Document smart service enforce the following guardrails:

Guardrail	Limit
Maximum input file size	10 MB
Maximum pages	25
Maximum output file size	25 MB
Maximum keywords	100
Maximum keyword length	200 characters
Maximum regular expression patterns	100
Maximum regular expression pattern length	200 characters
Maximum semantic instruction length	150,000 characters

Attribute AI usage to groups

You can attribute AI usage from the Redact Document smart service to specific Appian groups for internal chargeback and cost allocation. To associate one or more groups with the execution, configure the Usage Groups input on the Data tab. The Usage Groups input accepts a List of Group and is configured on the smart service node, not the AI skill itself. Configuring Usage Groups on the node means different process models, or different branches within the same process, can call the same AI skill and attribute those calls to different groups depending on the business context.

If the Usage Groups input is left empty, the execution is logged as unattributed.

AI Actions consumption

Every Redact Document execution consumes AI Actions because the smart service runs text recognition on every page of the document. When semantic matching is active, the execution also runs a generative AI model, which consumes additional AI Actions. An execution that uses semantic matching consumes more AI Actions than an execution that uses only keyword or regular expression matching.

The redacted document is image-only

The Redact Document smart service returns an image-only PDF. The output document has no text layer, so the redacted text cannot be recovered in any of the following ways:

• Copy and paste: Selecting and copying text from the output document returns nothing.
• Search: Searching the output document with Ctrl+F finds no text.
• Text extraction: Running text extraction tools against the output document returns no text.
• Metadata inspection: Inspecting the output document's metadata reveals no redacted text.

Because the output document is image-based, its file size can differ from the input document.

Redact Document runtime behavior

When the process reaches the node, the Redact Document smart service calls the Document Redaction AI skill configured in the Setup tab. All redaction methods active in the skill run, and the smart service redacts the union of the results.

The Redact Document smart service redacts a document up to 25 pages and 10 MB, and returns a redacted document up to 25 MB.

Redact Document error scenarios

The Redact Document node fails if the input document exceeds the supported size or page limits. If the node fails for this reason, send a PDF within the supported limits to the smart service.

Feedback