Skip to main content

Guardrails

Coming Soon

wl-guardrails is under active development and will be available in an upcoming release.

wl-guardrails is a real-time content safety engine that enforces guardrail policies on agent inputs and outputs using policy-as-code.

Planned Capabilities​

  • Rego/OPA policies — define safety rules using the Open Policy Agent policy language (via Regorus)
  • Content filtering — real-time safety checks on agent inputs and outputs
  • YAML definitions — declarative guardrail policy configuration
  • Prometheus metrics — observability and alerting integration

Architecture​

wl-guardrails will integrate into the governance proxy pipeline alongside wl-apdp policy evaluation:

Agent request → wl-proxy
  ├─ Cedar policy eval (wl-apdp)     — authorization
  ├─ Guardrail eval (wl-guardrails)  — content safety
  ├─ Credential injection (wl-secrets-broker)
  └─ Forward to upstream

Stay Updated​

Contact your Watchlight AI representative or visit the Partner Portal for updates on availability.