[Ilia Sokolov (ilia-sokolov)]

Motivation and Context

The streaming chat-history reducer sample reports aggregate token usage but gives no indication of the latency users experience before the first response content becomes available.

This change addresses #9580 by making that latency visible for the reducer-enabled streaming path. The label explicitly states that chat-history reduction is included, avoiding the misleading implication that the value represents provider-only latency.

Description

• start a per-turn stopwatch immediately before requesting the streaming response
• stop and report it once, on the first non-empty textual update
• preserve the existing response assembly and token accounting
• report a truthful fallback when a stream returns no text, such as a metadata-only or tool-only stream

The change is confined to MultipleProviders_ChatHistoryReducer.cs, introduces no dependencies or public APIs, and intentionally avoids timing thresholds because provider latency is nondeterministic.

Validation

The exact modified sample source compiled successfully with 0 warnings and 0 errors against the released Semantic Kernel 1.78 packages.

The repository-wide Concepts build and formatter traverse a large connector project graph and did not complete in the available validation window. The modified Semantic Kernel API expressions are unchanged from main; the added implementation uses only System.Diagnostics.Stopwatch and local control flow.

Contribution Checklist

• The complete repository build finishes without any errors or warnings
• The repository-wide pre-submission formatting script raises no violations
• All repository unit tests pass
• The focused sample change compiles cleanly and preserves existing behavior
• I didn't break anyone 😄

[github-actions]

Automated Code Review

Reviewers: 5 | Confidence: 92% | Result: All clear

Reviewed: Correctness, Security Reliability, Test Coverage, Failure Modes, Design Approach

---

Automated review by ilia-sokolov's agents

[Copilot]

Pull request overview

Adds a latency metric to the .NET streaming chat-history reducer sample so users can see how long it takes before the first assistant text becomes available when chat-history reduction is enabled.

Changes:

• Start a per-turn Stopwatch before initiating streaming and print elapsed time on the first non-empty text update.
• Preserve existing response assembly and token usage aggregation while streaming.
• Add a fallback message for streams that produce no text content.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Ilia Sokolov (ilia-sokolov)]

Hi @microsoft/octo-semantickernel-pr-dotnet, could someone review this focused .NET sample change when you have a chance? It addresses #9580 by reporting reducer-inclusive time to first non-empty text, and reports stream-completion time separately for textless responses. The earlier Copilot feedback has been addressed. Please let me know whether this is still the desired scope or if any adjustments or additional validation would help. Thanks!