prompt-executor-bedrock-client/ai.koog.prompt.executor.clients.bedrock/BedrockLLMClient

BedrockLLMClient

class BedrockLLMClient(bedrockClient: BedrockRuntimeClient, moderationGuardrailsSettings: BedrockGuardrailsSettings? = null, clock: Clock = Clock.System) : LLMClient, LLMEmbeddingProvider(source)

Creates a new Bedrock LLM client configured with the specified AWS credentials and settings.

Return

A configured LLMClient instance for Bedrock

Parameters

bedrockClient

The runtime client for interacting with Bedrock, highly configurable

clock

A clock used for time-based operations

moderationGuardrailsSettings

Optional settings of the AWS bedrock Guardrails (see AWS documentation ) that would be used for the LLMClient.moderate request

Constructors

BedrockLLMClient

constructor(identityProvider: IdentityProvider, settings: BedrockClientSettings = BedrockClientSettings(), clock: Clock = Clock.System)

Creates a new Bedrock LLM client configured with the specified identity provider and settings.

constructor(bedrockClient: BedrockRuntimeClient, moderationGuardrailsSettings: BedrockGuardrailsSettings? = null, clock: Clock = Clock.System)

Functions

open override fun close()

embed

open suspend override fun embed(text: String, model: LLModel): List<Double>

execute

open suspend override fun execute(prompt: Prompt, model: LLModel, tools: List<ToolDescriptor>): List<Message.Response>

executeMultipleChoices

open suspend fun executeMultipleChoices(prompt: Prompt, model: LLModel, tools: List<ToolDescriptor>): List<LLMChoice>

executeStreaming

open override fun executeStreaming(prompt: Prompt, model: LLModel, tools: List<ToolDescriptor>): Flow<StreamFrame>

llmProvider

open override fun llmProvider(): LLMProvider

Provides the current language learning model provider utilized by this client.

models

open suspend fun models(): List<String>

moderate

open suspend override fun moderate(prompt: Prompt, model: LLModel): ModerationResult

Moderates the provided prompt using specified moderation guardrails settings. The method evaluates both input and output of the prompt against guardrails and determines if either is harmful, returning a corresponding result.