Skip to main content
POST
cURL

Authorizations

Authorization
string
header
required

API key for programmatic access (Authorization: Bearer sk-<api_key>)

Body

application/json

Privacy redaction endpoint

Sends each input text through the privacy-filter model and returns it rewritten with realistic-looking dummies in place of detected PII (e.g. redacted1@example.com, +1-555-0100, Redacted001). Useful as a one-shot sanitizer ahead of calling a third-party LLM. The original PII is never echoed back; only the redacted form is returned.

Internally this is a /v1/privacy/classify call plus local span-apply — same billing (one input_tokens charge per input), same concurrency limits, same model routing.

model
string
required

ID of the model to use for redaction (typically openai/privacy-filter).

input
any

Text or list of texts to redact. Either a string or array of strings.

threshold
number<double> | null

Optional minimum confidence score for redacted spans (0.0–1.0).

Response

Privacy redaction completed successfully

data
object[]
required

Per-input redaction results.

model
string

Model identifier that produced the redaction.