WorkflowEvaluationExtensions.EvaluateAsync Method
Definition
Important
Some information relates to prerelease product that may be substantially modified before it’s released. Microsoft makes no warranties, express or implied, with respect to the information provided here.
Evaluates a completed workflow run.
public static System.Threading.Tasks.Task<Microsoft.Agents.AI.AgentEvaluationResults> EvaluateAsync(this Microsoft.Agents.AI.Workflows.Run run, Microsoft.Agents.AI.IAgentEvaluator evaluator, bool includeOverall = true, bool includePerAgent = true, string evalName = "Workflow Eval", Microsoft.Agents.AI.IConversationSplitter? splitter = default, string? expectedOutput = default, System.Threading.CancellationToken cancellationToken = default);static member EvaluateAsync : Microsoft.Agents.AI.Workflows.Run * Microsoft.Agents.AI.IAgentEvaluator * bool * bool * string * Microsoft.Agents.AI.IConversationSplitter * string * System.Threading.CancellationToken -> System.Threading.Tasks.Task<Microsoft.Agents.AI.AgentEvaluationResults><Extension()>
Public Function EvaluateAsync (run As Run, evaluator As IAgentEvaluator, Optional includeOverall As Boolean = true, Optional includePerAgent As Boolean = true, Optional evalName As String = "Workflow Eval", Optional splitter As IConversationSplitter = Nothing, Optional expectedOutput As String = Nothing, Optional cancellationToken As CancellationToken = Nothing) As Task(Of AgentEvaluationResults)Parameters
- run
- Run
The completed workflow run.
- evaluator
- IAgentEvaluator
The evaluator to score results.
- includeOverall
- Boolean
Whether to include an overall evaluation.
- includePerAgent
- Boolean
Whether to include per-agent breakdowns.
- evalName
- String
Display name for this evaluation run.
- splitter
- IConversationSplitter
Optional conversation splitter to apply to all items. Use LastTurn, Full, or a custom IConversationSplitter implementation.
- expectedOutput
- String
Optional ground-truth/expected output for the workflow's overall final answer.
When provided, it is stamped onto the overall ExpectedOutput
so reference-based evaluators (for example, similarity) can compare the
workflow's response against a golden answer. Ground truth is only applied
to the overall item; per-agent items are intentionally left without an
expected output, since ground truth is defined against the final response.
When using a reference-based evaluator that requires ground truth, set
includePerAgent to false to avoid
invoking the evaluator on per-agent items that have no expected output.
- cancellationToken
- CancellationToken
Cancellation token.
Returns
Evaluation results with optional per-agent sub-results.