Guardrails
struct Guardrails : Sendable, EquatableSettings for controlling how potentially harmful content is blocked or flagged by the model.
Guardrails are roughly equivalent to SafetySettings for Gemini models. For more details,
see the Apple
documentation.
-
The default guardrail settings for the on-device model.
For more details, see the Apple documentation.
Declaration
Swift
public static let `default`: FirebaseAI.SystemLanguageModel.Guardrails -
Guardrail settings that are less restrictive for content transformation prompts.
Content transformation includes tasks such as summarizing or rewriting text. For more details, see the Apple documentation.
Declaration
Swift
public static let permissiveContentTransformations: FirebaseAI.SystemLanguageModel.Guardrails