VCodeX
Guardrails & security

PII detection & redaction

Catch and mask sensitive data before it reaches a model.

The PII guardrail scans each prompt with built-in patterns plus custom org rules and takes one of three actions per match.

ActionBehavior
RejectBlock the request
RedactMask the match and continue
AllowPermit the match
  • Rules are org-scoped with tenant isolation, cached briefly per org.
  • Each regex runs under a small timeout budget (configurable) to stay fast.
  • Detection statistics are recorded for review.
prompt:  "email sarah.chen@gmail.com about the outage"
result:  "email [REDACTED] about the outage"   ✓ allowed · 1 redaction

Configure rules in the dashboard under Security → PII Detection, and review matches under Security → PII Audit.

Error logs

Classified, sanitized error logs with 30-day retention.

Secret blocking

Stop API keys and secrets from leaking into prompts.