EvaluatorReference

class aws_cdk.aws_bedrock_agentcore_alpha.EvaluatorReference(*args: Any, **kwargs)

Bases: object

(experimental) Represents a reference to an evaluator for online evaluation.

Use the static factory methods to create evaluator references:

EvaluatorReference.builtin() for built-in evaluators
EvaluatorReference.custom() for custom evaluators

Stability:: experimental

Example:

# Using custom evaluators
# my_custom_evaluator: agentcore.IEvaluator
# Using built-in evaluators
helpfulness = agentcore.EvaluatorReference.builtin(agentcore.BuiltinEvaluator.HELPFULNESS)
custom = agentcore.EvaluatorReference.custom(my_custom_evaluator)

Methods

bind()

(experimental) Binds the evaluator reference to produce the L1 property.

Stability:: experimental
Return type:: EvaluatorReferenceBindResult

Attributes

evaluator_id

(experimental) The evaluator identifier.

Stability:: experimental

Static Methods

classmethod builtin(evaluator)

(experimental) Creates a reference to a built-in evaluator.

Built-in evaluators are provided by Amazon Bedrock AgentCore and assess different aspects of agent performance at various levels (session, trace, or tool call).

Parameters:

evaluator (BuiltinEvaluator) –

The built-in evaluator to reference.

Return type:

EvaluatorReference

Returns:

An EvaluatorReference instance

Stability:

experimental

Example:

helpfulness = agentcore.EvaluatorReference.builtin(agentcore.BuiltinEvaluator.HELPFULNESS)
goal_success = agentcore.EvaluatorReference.builtin(agentcore.BuiltinEvaluator.GOAL_SUCCESS_RATE)

classmethod custom(evaluator)

(experimental) Creates a reference to a custom evaluator.

Custom evaluators are created using the Evaluator construct and can be LLM-as-a-Judge or code-based (Lambda) evaluators.

Parameters:

evaluator (IEvaluator) –

The custom evaluator construct to reference.

Return type:

EvaluatorReference

Returns:

An EvaluatorReference instance

Stability:

experimental

Example:

# my_custom_evaluator: agentcore.IEvaluator

ref = agentcore.EvaluatorReference.custom(my_custom_evaluator)