Skip to main content

Transform (Proprietary)

Transform is Custodian Labs' proprietary de-identification algorithm within Guardian Layer. It replaces detected sensitive values with safer alternatives while keeping the sentence readable.

Use this mode when you want protected output that still looks natural enough for review, demos, testing, AI workflows, or downstream applications.

When to Use Transform​

Use transform when you want to:

  • Keep the original meaning and structure of the text.
  • Replace sensitive values instead of showing obvious redaction marks.
  • Generate readable sample data from sensitive input.
  • Send de-identified text into another workflow, model, or application.

If you need a stricter output that clearly hides sensitive values, use redact instead.

Endpoint​

POST /api/v1/deidentify/text/proprietary/outputs

This endpoint returns de-identified text outputs for a single text request.

Request​

Set masking_type to transform.

{
  "text": "Dr. Emily Chen called Alex at 555-0100 about the appointment in Auckland.",
  "domain": "General",
  "pca_dimension": 2,
  "masking_type": "transform",
  "pii_entities": ["PERSON", "PHONE_NUMBER", "LOCATION"]
}

Parameters​

ParameterTypeRequiredDescription
textstringYesThe input text you want to de-identify.
domainstringNoDetection domain. Supported values are General, Medical, Finance, and Custom. Defaults to General.
pca_dimensionintegerNoControls the proprietary detection space. Supported values are 1, 2, and 3. Defaults to 2.
masking_typestringNoUse transform for proprietary replacement. This is the default value.
pii_entitiesarrayNoLimits which entity categories are transformed. Omit this field, pass an empty list, or pass ["ALL"] to process all supported entities.

For the full entity list, see Supported Entities.

Response​

The response includes the original text, one or more transformed outputs, a mode label, and metadata.

{
  "original_text": "Dr. Emily Chen called Alex at 555-0100 about the appointment in Auckland.",
  "outputs": [
    {
      "id": 1,
      "text": "Dr. Sarah Miller called Jordan at 555-0188 about the appointment in Wellington."
    },
    {
      "id": 2,
      "text": "Dr. Laura Adams called Casey at 555-0142 about the appointment in Christchurch."
    }
  ],
  "mode": "PROPRIETARY_OUTPUTS",
  "meta": {
    "masking_type": "transform",
    "domain": "General",
    "pca_dimension": 2
  }
}

How to Read the Outputs​

outputs contains transformed alternatives. In most product flows, you can use the first output by default.

If your workflow includes human review, you can show multiple outputs and let the reviewer choose the version that best fits the use case.

Transform and Redact​

Transform and redact both belong to Guardian Layer, but they are not the same type of capability.

ModeWhat it doesBest for
transformUses Custodian Labs' proprietary algorithm to replace sensitive values with readable alternatives.Natural-looking de-identified text, demos, AI workflows, sample data.
redactUses Guardian Layer masking to hide sensitive values with masking marks.Strict hiding, compliance review, audit-friendly output.

If you want visible redaction behavior, see Redact / Masking / Placeholder.

Notes​

  • Transform changes sensitive values; it does not simply delete them.
  • The output may preserve the structure of the original sentence.
  • Entity filtering is optional, but useful when you only want to transform certain categories.
  • The API key used for the request must have enough remaining usage credits.
  • For file uploads, use the file de-identification endpoints instead of this text endpoint.

Next Steps​