Scope and Safety Tuning

Learn how to customize GenerativeAgent’s scope and safety guardrails

GenerativeAgent includes default safety and scope guardrails to keep conversations aligned with business needs and to ensure GenerativeAgent engages only in appropriate topics.

These tools allow you to:

• Define custom categories for what’s considered “in-scope”
• Configure input safety categories for allowed customer messages
• Maintain default safety protections while adding organization-specific allowances

Customizing Scope and Safety Settings

Scope and safety controls are available in the main Settings page of GenerativeAgent. After making any changes to these settings, be sure to test them using the Previewer before deploying to production.

• In-Scope Categories
• Input Safety Categories

To define valid topics for GenerativeAgent:

1. Navigate to Settings > In-Scope Categories
2. Click “Add Category”
3. Enter a category name
4. Provide specific examples of acceptable topics/requests
5. Save your changes

The default safety and scope guardrails remain active even when adding custom categories. Your configurations help customize permissible interactions while maintaining core safety features.

If scope settings seem too restrictive, you can add new categories or expand existing ones. Always test changes in the Previewer before deployment.

To define valid topics for GenerativeAgent:

1. Navigate to Settings > In-Scope Categories
2. Click “Add Category”
3. Enter a category name
4. Provide specific examples of acceptable topics/requests
5. Save your changes

The default safety and scope guardrails remain active even when adding custom categories. Your configurations help customize permissible interactions while maintaining core safety features.

If scope settings seem too restrictive, you can add new categories or expand existing ones. Always test changes in the Previewer before deployment.

To specify allowed customer message types:

1. Go to Settings > Input Safety Categories
2. Click “Add Category”
3. Define the category name
4. Add example messages that should be allowed
5. Include context explaining why these inputs are safe
6. Save your configuration

Safety Context

Input safety categories require explanations to provide context for why certain inputs are deemed safe. This helps GenerativeAgent accurately apply exceptions relevant to your specific needs while maintaining overall safety standards.

Next Steps

After configuring scope and safety settings, you may want to explore:

Previewer

Safety and Troubleshooting

Deploying to GenerativeAgent