Put the safety layer to work.
This playground calls the live Verity gateway — the same rules and trained classifier that screen real traffic. Try to get an abusive prompt past it, or compare an answer against its source context.
Ready to test
No test run yet
- Decision
— - Deciding layer
— - Matched rules
— - Policy action
— - Policy reason
— - Latency
— - Next hop
—
Live check: this calls the real Verity gateway — rules plus the trained classifier — the same API protecting this site.
Ready to test
No scoring run yet
- Verdict
— - Grounded sentences
— - Method
heuristic fallback
Production path: LLM-as-judge first; this exact content-word heuristic remains the offline fallback.
What this demo is — and is not.
Screening here runs the production guard, classifier and policy code. In mock mode the "model" behind it is simulated, so answers are illustrative — verdicts, signals and latency are real.
If the gateway cannot be reached, the playground says so and shows a local-rules fallback verdict, labeled as such. A fallback is never presented as a live result.
Playground traffic is public demo traffic. Never paste credentials, personal data, or production prompts you would not want in a demo log.
Ready for your own key and dashboard? Start free — or read how Verity works.