FirebaseAI
@available(iOS 15.0, macOS 12.0, tvOS 15.0, watchOS 8.0, *)
public final class FirebaseAI : SendableThe Firebase AI SDK provides access to Gemini models directly from your app.
-
Creates an instance of
FirebaseAI.Declaration
Swift
public static func firebaseAI(app: FirebaseApp? = nil, backend: Backend = .googleAI(), useLimitedUseAppCheckTokens: Bool = false) -> FirebaseAIParameters
appA custom
FirebaseAppused for initialization; if not specified, uses the defaultFirebaseApp.backendThe backend API for the Firebase AI SDK; if not specified, uses the default
googleAI()(Gemini Developer API).useLimitedUseAppCheckTokensWhen sending tokens to the backend, this option enables the usage of App Check’s limited-use tokens instead of the standard cached tokens. Learn more about limited-use tokens, including their nuances, when to use them, and best practices for integrating them into your app.
Return Value
A
FirebaseAIinstance, configured with the customFirebaseApp. -
generativeModel(modelName:generationConfig:safetySettings:tools:toolConfig:systemInstruction:requestOptions:)
Initializes a generative model with the given parameters.
Note
Refer to Gemini models for guidance on choosing an appropriate model for your use case.
Declaration
Swift
public func generativeModel(modelName: String, generationConfig: GenerationConfig? = nil, safetySettings: [SafetySetting]? = nil, tools: [Tool]? = nil, toolConfig: ToolConfig? = nil, systemInstruction: ModelContent? = nil, requestOptions: RequestOptions = RequestOptions()) -> GenerativeModelParameters
modelNameThe name of the model to use; see available model names for a list of supported model names.
generationConfigThe content generation parameters your model should use.
safetySettingsA value describing what types of harmful content your model should allow.
toolsA list of
Toolobjects that the model may use to generate the next response.toolConfigTool configuration for any
Toolspecified in the request.systemInstructionInstructions that direct the model to behave a certain way; currently only text content is supported.
requestOptionsConfiguration parameters for sending requests to the backend.
-
Initializes an
ImagenModelwith the given parameters.Note
Refer to Imagen models for guidance on choosing an appropriate model for your use case.
Declaration
Swift
public func imagenModel(modelName: String, generationConfig: ImagenGenerationConfig? = nil, safetySettings: ImagenSafetySettings? = nil, requestOptions: RequestOptions = RequestOptions()) -> ImagenModelParameters
modelNameThe name of the Imagen 3 model to use.
generationConfigConfiguration options for generating images with Imagen.
safetySettingsSettings describing what types of potentially harmful content your model should allow.
requestOptionsConfiguration parameters for sending requests to the backend.
-
Initializes a new
TemplateGenerativeModel.Declaration
Swift
public func templateGenerativeModel() -> TemplateGenerativeModelReturn Value
A new
TemplateGenerativeModelinstance. -
Initializes a new
TemplateImagenModel.Declaration
Swift
public func templateImagenModel() -> TemplateImagenModelReturn Value
A new
TemplateImagenModelinstance. -
[Public Preview] Initializes a
LiveGenerativeModelwith the given parameters.Note
Refer to the Firebase docs on the Live API for guidance on choosing an appropriate model for your use case.Warning
Using the Firebase AI Logic SDKs with the Gemini Live API is in Public Preview, which means that the feature is not subject to any SLA or deprecation policy and could change in backwards-incompatible ways.
Declaration
Swift
@available(iOS 15.0, macOS 12.0, tvOS 15.0, *) @available(watchOS, unavailable) public func liveModel(modelName: String, generationConfig: LiveGenerationConfig? = nil, tools: [Tool]? = nil, toolConfig: ToolConfig? = nil, systemInstruction: ModelContent? = nil, requestOptions: RequestOptions = RequestOptions()) -> LiveGenerativeModelParameters
modelNameThe name of the model to use.
generationConfigThe content generation parameters your model should use.
toolsA list of
Toolobjects that the model may use to generate the next response.toolConfigTool configuration for any
Toolspecified in the request.systemInstructionInstructions that direct the model to behave a certain way; currently only text content is supported.
requestOptionsConfiguration parameters for sending requests to the backend.