.NET: Propagate EnableSensitiveData to auto-wired inner OpenTelemetryChatClient

[Copilot]

Motivation and Context

• Fixes .NET: Use OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT to control sensitive data capture in telemetry #5873

When EnableSensitiveData = true is set on OpenTelemetryAgent and auto-wiring is active, the inner OpenTelemetryChatClient created by GetRunOptionsWithChatClientWiring was always instantiated with the default EnableSensitiveData = false. This caused the inner chat span to silently omit gen_ai.input.messages / gen_ai.output.messages even when the caller explicitly opted in to sensitive data capture — violating the OpenTelemetry GenAI semantic conventions contract described in the issue.

Description

Root cause: WrapIfNeeded was a zero-parameter static local that called UseOpenTelemetry(sourceName: sourceName) with no configure callback, so EnableSensitiveData was never forwarded to the auto-wired client.

Fix (OpenTelemetryAgent.cs):

// Before
static IChatClient WrapIfNeeded(IChatClient cc, string sourceName) =>
    cc.GetService(typeof(OpenTelemetryChatClient)) is not null
        ? cc
        : cc.AsBuilder().UseOpenTelemetry(sourceName: sourceName).Build();

// After
bool enableSensitiveData = this.EnableSensitiveData;
static IChatClient WrapIfNeeded(IChatClient cc, string sourceName, bool enableSensitiveData) =>
    cc.GetService(typeof(OpenTelemetryChatClient)) is not null
        ? cc
        : cc.AsBuilder()
            .UseOpenTelemetry(sourceName: sourceName, configure: o => o.EnableSensitiveData = enableSensitiveData)
            .Build();

The value is read once at GetRunOptionsWithChatClientWiring call-time and captured in the factory closure, consistent with how sourceName is already handled. Both the ChatClientAgentRunOptions clone path and the plain AgentRunOptions → new-options path are updated.

Test (OpenTelemetryAgentTests.cs): Added AutoWireChatClient_EnableSensitiveData_PropagatedToInnerChatClient_Async — a 4-variant theory (enableSensitiveData × streaming) written red-first to replicate the bug, then green after the fix. Asserts that the inner chat span includes/excludes gen_ai.input.messages and gen_ai.output.messages in lock-step with the outer agent's EnableSensitiveData flag.

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the Contribution Guidelines
• All unit tests pass, and I have added new tests where possible
• Is this a breaking change? If yes, add "[BREAKING]" prefix to the title of the PR.

[Copilot]

Pull request overview

This PR fixes a telemetry configuration propagation bug in the .NET OpenTelemetryAgent auto-wiring path: when the agent is configured with EnableSensitiveData = true, the implicitly-created inner OpenTelemetryChatClient is now configured the same way, ensuring GenAI message-content tags are emitted consistently.

Changes:

• Propagate OpenTelemetryAgent.EnableSensitiveData into the auto-wired inner OpenTelemetryChatClient created by GetRunOptionsWithChatClientWiring.
• Add a regression theory test covering EnableSensitiveData × (streaming vs non-streaming) and asserting gen_ai.input.messages / gen_ai.output.messages presence in the inner chat span.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
dotnet/src/Microsoft.Agents.AI/OpenTelemetryAgent.cs	Forwards EnableSensitiveData to the auto-wired inner OpenTelemetryChatClient via the UseOpenTelemetry(..., configure: ...) callback.
dotnet/tests/Microsoft.Agents.AI.UnitTests/OpenTelemetryAgentTests.cs	Adds a regression theory verifying sensitive-data capture tags are present/absent on the inner chat span as expected.