Overview
Topic Adherence guardrail controls which topics AI agents can discuss. It ensures conversations stay within defined boundaries by allowing or denying specific topics.Configuration Options
Mode
- Allow: Only allow conversations about specified topics
- Deny: Block conversations about specified topics (allow everything else)
Topic Lists
- Topics Allowed: List of topics that are permitted (used in allow mode)
- Topics Denied: List of topics that are forbidden (used in deny mode)
Use Cases
- Customer Support: Keep agents focused on product-related topics
- Educational Bots: Maintain focus on specific subjects
- Compliance: Prevent discussions of regulated or sensitive topics
- Brand Safety: Avoid topics that could damage reputation
Best Practices
- Use clear, specific topic descriptions
- Test topic boundaries thoroughly
- Consider using allow mode for stricter control
- Regularly review topic lists based on usage patterns