Skip to main content
Wardin scans the outbound request body for PII patterns and replaces them with placeholder tokens before forwarding to any LLM provider. The original values never leave your network.

​
Built-in patterns

Wardin redacts a fixed set of four patterns — there’s no custom-pattern configuration: There’s no IP-address pattern and no way to add your own regex today — the set above is the whole list.

​
Audit logging

Redaction shows up on the request’s signed receipt as a GUARDRAIL check — REDACTED when a pattern matched, PASS when nothing did. The check records that a redaction happened, not which pattern matched or the original value. See Receipts for the full check-array shape, and query redacted requests directly with GET /v1/receipts?check=GUARDRAIL.

​
Scope

The gateway parses the outbound request body as JSON and walks every string value in it, applying the four patterns above and re-serializing — so it can’t accidentally miss a JSON-escaped match or corrupt JSON structure. It does not run on response bodies from the provider (your own model output).

​
Enabling redaction

Redaction is a tenant-wide pii policy — it isn’t scoped to individual keys and takes no configuration body:
See the full endpoint reference: POST /v1/policies. There’s at most one pii policy per tenant — creating a second is rejected.