Skip to main content

agrag.eval.trajectory.trajectory_case

trajectory_case(question:str, answer:str, trajectory:Trajectory) -> LLMTestCase

Build the test case every trajectory metric scores.

tools_called holds every TOOL step, planner and researcher, as a ToolCall. metadata["trajectory"] holds the serialized trajectory the deterministic metrics read.

Parameters:

  • question (str) – The question the agent answered.
  • answer (str) – The agent's final answer.
  • trajectory (Trajectory) – The run's trajectory from read_trajectory.

Returns:

  • LLMTestCase – The test case with the answer, the tool calls and the trajectory.