EvaluationInputs
NexusLabs.Needlr.AgentFramework.Evaluation¶
EvaluationInputs Struct¶
Inputs shaped for Microsoft.Extensions.AI.Evaluation evaluators, derived from a
captured agent run. Consumers pass Messages and ModelResponse
to IEvaluator.EvaluateAsync (or to a ScenarioRun obtained via
ReportingConfiguration.CreateScenarioRunAsync).
public readonly record struct EvaluationInputs : System.IEquatable<NexusLabs.Needlr.AgentFramework.Evaluation.EvaluationInputs>
Implements System.IEquatable<EvaluationInputs>
Constructors¶
EvaluationInputs(IReadOnlyList<ChatMessage>, ChatResponse) Constructor¶
Inputs shaped for Microsoft.Extensions.AI.Evaluation evaluators, derived from a
captured agent run. Consumers pass Messages and ModelResponse
to IEvaluator.EvaluateAsync (or to a ScenarioRun obtained via
ReportingConfiguration.CreateScenarioRunAsync).
public EvaluationInputs(System.Collections.Generic.IReadOnlyList<Microsoft.Extensions.AI.ChatMessage> Messages, Microsoft.Extensions.AI.ChatResponse ModelResponse);
Parameters¶
Messages System.Collections.Generic.IReadOnlyList<Microsoft.Extensions.AI.ChatMessage>
The conversation messages that were sent to the agent, exactly as captured at the agent-run boundary. Safe to replay without re-invoking the model.
ModelResponse Microsoft.Extensions.AI.ChatResponse
The agent's aggregated response materialized as a Microsoft.Extensions.AI.ChatResponse. When the captured run produced no output, this is an empty assistant response so evaluators still receive a non-null value.
Properties¶
EvaluationInputs.Messages Property¶
The conversation messages that were sent to the agent, exactly as captured at the agent-run boundary. Safe to replay without re-invoking the model.
public System.Collections.Generic.IReadOnlyList<Microsoft.Extensions.AI.ChatMessage> Messages { get; init; }
Property Value¶
System.Collections.Generic.IReadOnlyList<Microsoft.Extensions.AI.ChatMessage>
EvaluationInputs.ModelResponse Property¶
The agent's aggregated response materialized as a Microsoft.Extensions.AI.ChatResponse. When the captured run produced no output, this is an empty assistant response so evaluators still receive a non-null value.