MDAA TS Docs
    Preparing search index...

    RAG data source using Bedrock Knowledge Base for document-based question answering with citation support. Configures the foundation model used to generate answers, the Lambda role that executes the namespace handler, and retrieval-time parameters such as the number of knowledge base results and the prompt and orchestration templates.

    interface BedrockRagDataSourceProps {
        displayInlineCitations?: boolean;
        guardrailId?: string;
        guardrailKmsKeyArn?: string;
        guardrailVersion?: string;
        inferenceMaxTokens?: number;
        inferenceTemperature?: number;
        inferenceTopP?: number;
        kbNumberOfResults?: number;
        lambdaArchitecture?: "ARM_64" | "X86_64";
        lambdaMemorySize?: number;
        lambdaRole: MdaaRoleRef;
        lambdaTimeoutInSeconds?: number;
        modelId: string;
        orchestrationInferenceMaxTokens?: number;
        orchestrationInferenceStopSequences?: string[];
        orchestrationInferenceTemperature?: number;
        orchestrationInferenceTopP?: number;
        orchestrationPerformanceLatency?: string;
        orchestrationPromptTemplate?: string;
        orchestrationQueryTransformationType?: string;
        promptTemplate?: string;
        provisionedConcurrentExecutions?: number;
        pythonRuntime?: string;
        reservedConcurrentExecutions?: number;
    }
    Index

    Properties

    displayInlineCitations?: boolean

    Whether to display inline citations in responses

    guardrailId?: string

    Bedrock Guardrail ID for content filtering

    guardrailKmsKeyArn?: string

    KMS key ARN for Guardrail encryption

    guardrailVersion?: string

    Guardrail version to use

    inferenceMaxTokens?: number

    Maximum tokens for inference

    inferenceTemperature?: number

    Temperature for inference (0.0-1.0)

    inferenceTopP?: number

    Top-p for inference (0.0-1.0)

    kbNumberOfResults?: number

    Number of results to retrieve from knowledge base

    lambdaArchitecture?: "ARM_64" | "X86_64"

    Lambda architecture (ARM64 or x86_64). If undefined, Architecture.X86_64 is used

    lambdaMemorySize?: number

    Memory allocation for the Lambda function in MB. If undefined, 1024MB is used.

    lambdaRole: MdaaRoleRef

    IAM role reference for Lambda execution

    lambdaTimeoutInSeconds?: number

    Timeout for the Lambda function in seconds. If unspecified, 10 minutes is used (600 seconds)

    modelId: string

    Bedrock model ID for generation (e.g., 'anthropic.claude-3-sonnet-20240229-v1:0')

    orchestrationInferenceMaxTokens?: number

    Maximum tokens for orchestration inference

    orchestrationInferenceStopSequences?: string[]

    Stop sequences for orchestration inference

    orchestrationInferenceTemperature?: number

    Temperature for orchestration inference

    orchestrationInferenceTopP?: number

    Top-p for orchestration inference

    orchestrationPerformanceLatency?: string

    Performance vs latency trade-off setting. (e.g. : standard | optimized)

    orchestrationPromptTemplate?: string

    Custom orchestration prompt template

    orchestrationQueryTransformationType?: string

    Query transformation type for orchestration. (e.g. : QUERY_DECOMPOSITION)

    promptTemplate?: string

    Custom prompt template for generation

    provisionedConcurrentExecutions?: number

    Provisioned concurrent executions for Lambda

    pythonRuntime?: string

    Python runtime version. If undefined, Runtime.PYTHON_3_13 is used.

    reservedConcurrentExecutions?: number

    Reserved concurrent executions for Lambda