Off-topic guard
Keep an assistant on the job you built it for.
Describe what your assistant is for, and each message is checked against that scope. Off-topic requests get a polite deflection instead of a costly, off-brand answer; small talk is recognized as small talk.
Try it live
Real calls to Jev. Pick an example or paste your own.
Pick an example or paste your own input, then run it.
Under the hood
Every question and rule, in the open. Fork it to change any of them.
What it asks
Reads scope and message, then asks 3 questions in one call.
- Q1In scopeYes / no
Is the
messagesomething an assistant whose purpose isscopeshould help with? (Greetings, thanks, and follow-ups about earlier answers count as in scope.) - Q2Asks about competitorsYes / no
Does the
messageask the assistant to recommend or compare competitors' products? - Q3Small talkYes / no
Is the
messagecasual small talk with no real request?
How it decides
Rules run in code, top to bottom. The first one that matches gives the verdict.
- 1If Small talk ≥ 0.7Small talk
- 2If In scope < 0.4Deflect politely
- 3If Asks about competitors ≥ 0.7Deflect politely
- ·DefaultAnswer
Call it from your code
Every template runs as-is with an API key, no setup: POST /api/v1/judges/topic-guard/run.
curl -X POST https://tryevally.com/api/v1/judges/topic-guard/run \ -H "Authorization: Bearer $EVALLY_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "input": { "scope": "Customer support for Acme'\''s project management app.", "message": "Can you write me a cover letter for a marketing job?" }}'More in Guardrails
- Guardrails
Prompt injection guard
Screen user messages for jailbreaks, injected instructions and data exfiltration.
AllowReviewBlock7 questions - Guardrails
LLM reply guard
Check every model reply for policy breaks, unsafe advice and leaked data before it's shown.
AllowReviewBlock+18 questions - Guardrails
PII detector
Find personal data and secrets in text, and quote exactly where they are.
No personal dataPossible PIIContains PII5 questions - Guardrails
Brand voice check
Hold AI-written copy to your brand guidelines, including the words you never use.
On brandReviseOff brand4 questions