Skip to content

AgentTarget

junjo.evaluation.AgentTarget

Kind: Class

SDK version: 0.69.0

Documentation channel: Stable release

AgentTarget(*, key: str, name: str, input_version: int, input_type: TypeForm[InputT], factory: Callable[[InputT, EvaluationContext, ResourcesT], AgentInvocation[AgentInputT, AgentOutputT, DependenciesT, AgentStateT, StoreT] | Awaitable[AgentInvocation[AgentInputT, AgentOutputT, DependenciesT, AgentStateT, StoreT]]], projector: Callable[[AgentExecutionResult[AgentOutputT, AgentStateT], InputT, EvaluationContext, ResourcesT], object | Awaitable[object]])

View source

Execute a real application Agent through Agent.execute().

key is the stable evaluation dispatch identity. name is the human-readable name of the real Agent stored with Studio cases and shown in evaluation results.

kind = kind

View source

Public attribute.

key = key

View source

Public attribute.

name = name

View source

Public attribute.

input_version = input_version

View source

Public attribute.

input_schema: dict[str, object]

View source

Public attribute.