AgentOptimizationEvaluationConfiguration interface
Reusable quality-measurement configuration.
Properties
| evaluation_model | Model configuration used by model-based evaluators and conversation simulation. |
| evaluators | Evaluator references used to score candidate quality. |
| max_concurrent_agent_runs | Maximum number of target-agent runs executed concurrently during each single-turn evaluation. If omitted, the service defaults to 1. Conversation evaluation supports only 1. |
| training_set | Evaluation set used to guide the optimization search. Inline data supports up to 2,000 test cases when a separate validation set is supplied; otherwise it is also used for validation and is limited to 500. |
| validation_set | Held-out evaluation set used for full candidate evaluation, limited to 500 inline test cases. The training set is reused and subject to the same 500-test-case validation limit when omitted. |
Property Details
evaluation_model
Model configuration used by model-based evaluators and conversation simulation.
evaluation_model: EvaluationModelConfigurationProperty Value
evaluators
Evaluator references used to score candidate quality.
evaluators: AgentOptimizationEvaluator[]Property Value
max_concurrent_agent_runs
Maximum number of target-agent runs executed concurrently during each single-turn evaluation. If omitted, the service defaults to 1. Conversation evaluation supports only 1.
max_concurrent_agent_runs?: numberProperty Value
number
training_set
Evaluation set used to guide the optimization search. Inline data supports up to 2,000 test cases when a separate validation set is supplied; otherwise it is also used for validation and is limited to 500.
training_set: AgentOptimizationEvaluationSetUnionProperty Value
validation_set
Held-out evaluation set used for full candidate evaluation, limited to 500 inline test cases. The training set is reused and subject to the same 500-test-case validation limit when omitted.
validation_set?: AgentOptimizationEvaluationSetUnion