com.google.mlkit.genai.prompt
Interfaces
Caches |
Provides an interface to explicitly manage caches for |
GenerativeModel |
Provides an interface for performing content generation. |
Classes
CachedContext |
Represents an identifier of a cached context that can be used in inference. |
Candidate |
A piece of a response from the model. |
Content |
Represents a single "turn" or message in a conversation. |
Content.Builder |
Builder for |
CountTokensResponse |
Data class holding token count information. |
CreateCachedContextRequest |
A request to create a cached context. |
CreateCachedContextRequest.Builder |
Builder for |
FunctionCallPart |
A data class representing a function call part of a prompt. |
FunctionResponsePart |
A data class representing a function response part of a prompt. |
GenerateContentRequest |
A request to generate content. |
GenerateContentRequest.Builder |
Builder for |
GenerateContentResponse |
Represents a response from the model. |
GenerateTypedContentRequest |
A request to generate typed content. |
GenerateTypedContentRequest.Builder |
Builder for |
GenerateTypedContentResponse |
Represents a typed response from the model. |
GenerationConfig |
Configuration parameters to use for content generation. |
GenerationConfig.Builder |
Builder for |
ImagePart |
A data class representing an image part of a prompt. |
ModelConfig |
Specifies the desired model for content generation. |
ModelConfig.Builder |
Builder for |
Part |
A sealed class representing a part of a multimodal prompt. |
PromptPrefix |
Represents a prompt prefix for on-device prompt caching. |
SystemInstruction |
A data class representing a system text part of a prompt. |
TextPart |
A data class representing a text part of a prompt. |
TypedCandidate |
A piece of a response from the model. |
Objects
Generation |
Entry class to get a |
Annotations
Candidate.FinishReason |
The reason why this |
Content.Role |
Role associated with |
Feature |
Features (capabilities) that may be supported by a generative model. |
ModelPreference |
Defines the developer's preference to guide model selection, balancing characteristics like speed and capability. |
ModelReleaseStage |
Defines the release stage preference for model selection. |
TypedCandidate.TypedFinishReason |
The reason why this |
Top-level functions summary
Top-level functions
content
fun content(init: Content.Builder.() -> Unit): Content
Helper function for DSL-like syntax to construct Content with default Content.Role.USER role.
content
fun content(@Content.Role role: Int, init: Content.Builder.() -> Unit): Content
Helper function for DSL-like syntax to construct Content with specified Content.Role.
createCachedContextRequest
fun createCachedContextRequest(name: String, promptPrefix: PromptPrefix): CreateCachedContextRequest
Helper method to construct a CreateCachedContextRequest.
Example Usage:
createCachedContextRequest("my_cache", PromptPrefix("Repeat 3 times: "))
generateContentRequest
fun generateContentRequest(content: Content, init: GenerateContentRequest.Builder.() -> Unit = {}): GenerateContentRequest
Helper method to construct a GenerateContentRequest in a DSL-like manner.
Example Usage:
val request = generateContentRequest(
content {
image(bitmap1)
text("Identify this object.")
text("Does this sound belong to it?")
image(bitmap2)
text("How about this one?")
}
) {
temperature = 0.5f
}
)
generateContentRequest
fun generateContentRequest(text: TextPart, init: GenerateContentRequest.Builder.() -> Unit): GenerateContentRequest
Helper method to construct a GenerateContentRequest in a DSL-like manner.
Example Usage:
// Generation Request with text only:
generateContentRequest(TextPart("Hello World")) {
temperature = 0.5f
seed = 123
topK = 10
candidateCount = 5
maxOutputTokens = 200
}generateContentRequest
fun generateContentRequest(image: ImagePart, text: TextPart, init: GenerateContentRequest.Builder.() -> Unit): GenerateContentRequest
Helper method to construct a GenerateContentRequest in a DSL-like manner.
Example Usage:
// Generation Request with image and text:
generateContentRequest(ImagePart(bitmap), TextPart("Hello World")) {
temperature = 0.5f
seed = 123
topK = 10
candidateCount = 5
maxOutputTokens = 200
}generateContentRequest
fun generateContentRequest(
systemInstruction: SystemInstruction,
content: Content,
init: GenerateContentRequest.Builder.() -> Unit
): GenerateContentRequest
Helper method to construct a GenerateContentRequest in a DSL-like manner.
Example Usage:
// Generation Request with system instruction and content:
generateContentRequest(SystemInstruction("You are a helpful assistant."),
content {
image(bitmap1)
text("Identify this object.")
text("Does this sound belong to it?")
image(bitmap2)
text("How about this one?")
}
) {
temperature = 0.5f
}generateContentRequest
fun generateContentRequest(
systemInstruction: SystemInstruction,
text: TextPart,
init: GenerateContentRequest.Builder.() -> Unit
): GenerateContentRequest
Helper method to construct a GenerateContentRequest in a DSL-like manner.
Example Usage:
// Generation Request with system instruction and text:
generateContentRequest(SystemInstruction("You are a helpful assistant."), TextPart("Hello World")) {
temperature = 0.5f
seed = 123
topK = 10
candidateCount = 5
maxOutputTokens = 200
enableThinking = true
}generateContentRequest
fun generateContentRequest(
systemInstruction: SystemInstruction,
image: ImagePart,
text: TextPart,
init: GenerateContentRequest.Builder.() -> Unit
): GenerateContentRequest
Helper method to construct a GenerateContentRequest in a DSL-like manner.
Example Usage:
// Generation Request with system instruction, image, and text:
generateContentRequest(
SystemInstruction("You are a helpful assistant."),
ImagePart(bitmap),
TextPart("Hello World"),
) {
temperature = 0.5f
seed = 123
topK = 10
candidateCount = 5
maxOutputTokens = 200
enableThinking = true
}generateTypedContentRequest
fun <T : Any> generateTypedContentRequest(
generateContentRequest: GenerateContentRequest,
outputClass: KClass<T>
): GenerateTypedContentRequest<T>
Helper method to construct a GenerateTypedContentRequest in a DSL-like manner.
This is an experimental feature.
Example Usage:
// Generation Request for typed content:
generateTypedContentRequest(request, MyClass::class) {
}generateTypedContentRequest
fun <T : Any> generateTypedContentRequest(
generateContentRequest: GenerateContentRequest,
outputClass: KClass<T>,
includeSchemaInPrompt: Boolean
): GenerateTypedContentRequest<T>
Helper method to construct a GenerateTypedContentRequest with the specified includeSchemaInPrompt flag.
This is an experimental feature.
Example Usage:
// Generation Request for typed content:
generateTypedContentRequest(request, MyClass::class, false)generationConfig
fun generationConfig(init: GenerationConfig.Builder.() -> Unit): GenerationConfig
Helper method to construct a GenerationConfig in a DSL-like manner.
Example usage:
// Default generation config - uses ModelConfig.DEFAULT and a default background thread pool.
val config1 = generationConfig {}
// Customized model spec
val config2 = generationConfig {
workerExecutor = myExecutor
modelConfig = modelConfig {
releaseStage = ModelReleaseStage.PREVIEW
preference = ModelPreference.FAST
}
}
// Using prebuilt default ModelConfig with a custom executor
val config3 = generationConfig {
workerExecutor = myExecutor
modelConfig = ModelConfig.DEFAULT
}modelConfig
fun modelConfig(init: ModelConfig.Builder.() -> Unit): ModelConfig
Helper function for Kotlin DSL to build ModelConfig.