AgentPromptEvaluation

A human evaluation of one agent run, attributable to a prompt version.

type AgentPromptEvaluation {
  id: ID!
  ai_agent_run_id: ID!
  verdict: String!
  verdict_label: String!
  notes: String!
  created_by: ID
  created_by_name: String
  created_at: DateTime
  prompt_version_id: ID
  prompt_version: Int
  related_entity_type: String
  related_entity_id: ID
  run_status: String
  run_output: String
}

Fields

AgentPromptEvaluation.id ● ID! non-null scalar

AgentPromptEvaluation.ai_agent_run_id ● ID! non-null scalar

AgentPromptEvaluation.verdict ● String! non-null scalar

AgentRunEvaluationVerdict case name: INCORRECT | UNSURE | CORRECT.

AgentPromptEvaluation.verdict_label ● String! non-null scalar

AgentPromptEvaluation.notes ● String! non-null scalar

AgentPromptEvaluation.created_by ● ID scalar

AgentPromptEvaluation.created_by_name ● String scalar

AgentPromptEvaluation.created_at ● DateTime scalar

AgentPromptEvaluation.prompt_version_id ● ID scalar

The prompt version that produced the evaluated run (for grouping).

AgentPromptEvaluation.prompt_version ● Int scalar

Human per-prompt version number (v1/v2/…) of the run's prompt version.

AgentPromptEvaluation.related_entity_type ● String scalar

AgentPromptEvaluation.related_entity_id ● ID scalar

AgentPromptEvaluation.run_status ● String scalar

AgentPromptEvaluation.run_output ● String scalar

The run's output_payload as a pretty JSON string.

Returned By

agentPromptEvaluations query ● recordAgentRunEvaluation mutation