Creating a custom guardrail
- Select Infor OS > GenAI > Factory > Guardrail.
- Click +Add.
- Specify the values in mandatory fields. Fields with an asterisk (*) are required.
- Name*
- Specify the name of the guardrail.
- Guardrail Type*
-
Specify the type of content the guardrail enforces from the menu options:
- Ban List
- PII Detection
- Restrict To Topic
- Values*
- Specify the text values that the guardrail should actively block.
- Character Variance*
-
Define a character variance percentage between 70% and 100% that controls how much the content varies from the provided values.
Note: This property is only visible when ban list type selected.
- Applied Endpoints*
- Select either GenAI Assistant or Embedded Experience
- Application*
- Select Input and/or Output for when the guardrail should be enforced.
- Select the Invoke When mode to control when the guardrail is applied:
- Role-Based (default): Guardrail is applied based on user role. Select either Apply for all roles or Apply for specific roles (with a multi-select of security roles).
- On-Demand: Guardrail is applied only when triggered by a specific context. Select one of these options (mutually exclusive): Supervisor Only, Selected Agents (with a multi-select to choose specific agents), or Embedded Experience APIs based on request headers.
- Logical-ID-Based: Guardrail is applied only to conversations matching specific logical ID values.Separate sub-sections are available for GenAI Assistant and Embedded Experience, where you can add logical IDs or prefixes (with exact match or prefix match).
- Click Save.