[tang-ontheway]

Motivation & Context

tool_called_check, tool_calls_present and tool_call_args_match scanned item.conversation — the whole conversation — for function calls. Every other derived value on EvalItem (query, response) and the Foundry provider (item.split_messages(effective_split)) uses the item's split strategy.

As a result, in multi-turn evaluation a tool call made in an earlier turn satisfies the tool check for a later turn. With EvalItem.per_turn_items(...), a turn in which the agent made no tool call still passed tool_called_check("get_weather"), tool_calls_present and tool_call_args_match, because the call from turn 1 sits on the query side of turn 2's split.

That contradicts the per-turn design described in ADR-0023 and in per_turn_items: response_messages "contains the agent's actions up to the next user message", which "lets you evaluate each response independently". ADR-0023 also states: "Changing the split_strategy consistently changes all derived values." For a CI gate, the outcome is a silent false pass on exactly the turns that per-turn evaluation is meant to pinpoint.

Description & Review Guide

• What are the major changes?

  • Adds a private helper _response_messages(item) that returns item.split_messages()[1] — the response side of the item's split, the same messages item.response uses.
  • tool_called_check and _extract_tool_calls (used by tool_calls_present and tool_call_args_match) now read those messages instead of item.conversation.
  • Updates the affected docstrings to describe the response side of the split rather than "the conversation".
  • Adds regression tests covering per-turn items for all three checks, a FULL split that still counts the whole trajectory, and an unchanged single-turn conversation.

• What is the impact of these changes?

  • No public API change. Single-turn items with the default LAST_TURN split behave exactly as before, because the query side there contains only the user message.
  • For multi-turn data, a tool call from an earlier turn no longer satisfies a later turn's tool check.

• What do you want reviewers to focus on?

  • Whether scoping the tool checks to the response side of the split is the intended behaviour, and whether the Foundry provider's tool evaluators (which already use item.split_messages(effective_split)) should stay as they are.

Related Issue

Fixes #10091

AI Assistance

• No material AI assistance was used.
• This is an AI-assisted contribution. I reviewed, understood, and verified all submitted content and accept responsibility for it.

AI assistance details: An AI coding assistant helped audit the evaluation module, reproduce the issue, write the regression tests, and draft this description. The analysis, the reproduction and the test results were verified by me against main at commit dc6617a.

Contribution Checklist

• The code builds clean without any errors or warnings
• All unit tests pass, and I have added new tests where possible
• The PR follows the Contribution Guidelines
• This PR links to an agreed issue with no competing open PR, or the Related Issue section documents a trivial-change or repository-automation exception.
• This is not a breaking change. If it is a breaking change, add the breaking change label (or add "[BREAKING]" to the title prefix, before or after any language prefix) — a workflow keeps the label and title prefix in sync automatically.

[Copilot]

🟡 Changes recommended

The reusable conversation builder in test_local_eval.py violates the required test-helper layout and must be moved to a helper module or converted to an injected fixture in conftest.py.

1 open issue

• Move reusable conversation builder to a test helper or fixture · New

What changed in this PR

This PR fixes #10091 by making local tool checks respect EvalItem’s conversation split, preventing earlier tool calls from incorrectly satisfying later-turn evaluations.

Changes:

• Scope all three local tool checks to response-side messages through a shared private helper.
• Update docstrings to describe split-aware behavior.
• Add regression tests for per-turn isolation, FULL splits, and unchanged single-turn behavior.

File	Description
python/​packages/​core/​tests/​core/​test_local_eval.py	Adds regression coverage for split-aware tool checks.
python/​packages/​core/​agent_framework/​_evaluation.py	Restricts tool checks to response-side messages and updates documentation.

_{Review effort: Balanced}

[microsoft-github-policy-service]

tang-ontheway please read the following Contributor License Agreement(CLA). If you agree with the CLA, please reply with the following information.

@microsoft-github-policy-service agree [company="{your company}"]

Options:

• (default - no company specified) I have sole ownership of intellectual property rights to my Submissions and I am not making Submissions in the course of work for my employer.

@microsoft-github-policy-service agree

• (when company given) I am making Submissions in the course of work for my employer (or my employer has intellectual property rights in my Submissions by contract or applicable law). I have permission from my employer to make Submissions and enter into this Agreement on behalf of my employer. By signing below, the defined term “You” includes me and my employer.

@microsoft-github-policy-service agree company="Microsoft"

Contributor License Agreement

Contribution License Agreement

This Contribution License Agreement (“Agreement”) is agreed to by the party signing below (“You”),
and conveys certain license rights to Microsoft Corporation and its affiliates (“Microsoft”) for Your
contributions to Microsoft open source projects. This Agreement is effective as of the latest signature
date below.

1. Definitions.
   “Code” means the computer software code, whether in human-readable or machine-executable form,
   that is delivered by You to Microsoft under this Agreement.
   “Project” means any of the projects owned or managed by Microsoft and offered under a license
   approved by the Open Source Initiative (www.opensource.org).
   “Submit” is the act of uploading, submitting, transmitting, or distributing code or other content to any
   Project, including but not limited to communication on electronic mailing lists, source code control
   systems, and issue tracking systems that are managed by, or on behalf of, the Project for the purpose of
   discussing and improving that Project, but excluding communication that is conspicuously marked or
   otherwise designated in writing by You as “Not a Submission.”
   “Submission” means the Code and any other copyrightable material Submitted by You, including any
   associated comments and documentation.
2. Your Submission. You must agree to the terms of this Agreement before making a Submission to any
   Project. This Agreement covers any and all Submissions that You, now or in the future (except as
   described in Section 4 below), Submit to any Project.
3. Originality of Work. You represent that each of Your Submissions is entirely Your original work.
   Should You wish to Submit materials that are not Your original work, You may Submit them separately
   to the Project if You (a) retain all copyright and license information that was in the materials as You
   received them, (b) in the description accompanying Your Submission, include the phrase “Submission
   containing materials of a third party:” followed by the names of the third party and any licenses or other
   restrictions of which You are aware, and (c) follow any other instructions in the Project’s written
   guidelines concerning Submissions.
4. Your Employer. References to “employer” in this Agreement include Your employer or anyone else
   for whom You are acting in making Your Submission, e.g. as a contractor, vendor, or agent. If Your
   Submission is made in the course of Your work for an employer or Your employer has intellectual
   property rights in Your Submission by contract or applicable law, You must secure permission from Your
   employer to make the Submission before signing this Agreement. In that case, the term “You” in this
   Agreement will refer to You and the employer collectively. If You change employers in the future and
   desire to Submit additional Submissions for the new employer, then You agree to sign a new Agreement
   and secure permission from the new employer before Submitting those Submissions.
5. Licenses.

• Copyright License. You grant Microsoft, and those who receive the Submission directly or
  indirectly from Microsoft, a perpetual, worldwide, non-exclusive, royalty-free, irrevocable license in the
  Submission to reproduce, prepare derivative works of, publicly display, publicly perform, and distribute
  the Submission and such derivative works, and to sublicense any or all of the foregoing rights to third
  parties.
• Patent License. You grant Microsoft, and those who receive the Submission directly or
  indirectly from Microsoft, a perpetual, worldwide, non-exclusive, royalty-free, irrevocable license under
  Your patent claims that are necessarily infringed by the Submission or the combination of the
  Submission with the Project to which it was Submitted to make, have made, use, offer to sell, sell and
  import or otherwise dispose of the Submission alone or with the Project.
• Other Rights Reserved. Each party reserves all rights not expressly granted in this Agreement.
  No additional licenses or rights whatsoever (including, without limitation, any implied licenses) are
  granted by implication, exhaustion, estoppel or otherwise.

6. Representations and Warranties. You represent that You are legally entitled to grant the above
   licenses. You represent that each of Your Submissions is entirely Your original work (except as You may
   have disclosed under Section 3). You represent that You have secured permission from Your employer to
   make the Submission in cases where Your Submission is made in the course of Your work for Your
   employer or Your employer has intellectual property rights in Your Submission by contract or applicable
   law. If You are signing this Agreement on behalf of Your employer, You represent and warrant that You
   have the necessary authority to bind the listed employer to the obligations contained in this Agreement.
   You are not expected to provide support for Your Submission, unless You choose to do so. UNLESS
   REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING, AND EXCEPT FOR THE WARRANTIES
   EXPRESSLY STATED IN SECTIONS 3, 4, AND 6, THE SUBMISSION PROVIDED UNDER THIS AGREEMENT IS
   PROVIDED WITHOUT WARRANTY OF ANY KIND, INCLUDING, BUT NOT LIMITED TO, ANY WARRANTY OF
   NONINFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
7. Notice to Microsoft. You agree to notify Microsoft in writing of any facts or circumstances of which
   You later become aware that would make Your representations in this Agreement inaccurate in any
   respect.
8. Information about Submissions. You agree that contributions to Projects and information about
   contributions may be maintained indefinitely and disclosed publicly, including Your name and other
   information that You submit with Your Submission.
9. Governing Law/Jurisdiction. This Agreement is governed by the laws of the State of Washington, and
   the parties consent to exclusive jurisdiction and venue in the federal courts sitting in King County,
   Washington, unless no federal subject matter jurisdiction exists, in which case the parties consent to
   exclusive jurisdiction and venue in the Superior Court of King County, Washington. The parties waive all
   defenses of lack of personal jurisdiction and forum non-conveniens.
10. Entire Agreement/Assignment. This Agreement is the entire agreement between the parties, and
    supersedes any and all prior agreements, understandings or communications, written or oral, between
    the parties relating to the subject matter hereof. This Agreement may be assigned by Microsoft.

[Eduard van Valkenburg (eavanvalkenburg)]

tang-ontheway, this response-side scan treats the last role="user" message as the turn boundary, but AgentLoopMiddleware injects user-role nudge/progress messages between iterations in its default aggregated response. A tool call from an earlier loop iteration then falls on the query side and tool_called_check / expected-tool-call checks report it missing. Please distinguish real conversation turns from loop-injected messages (or scope this only to explicit multi-turn items) and add a loop-aggregated regression.

[Qiming Bao (14H034160212)]

Reporter of #10091 here: this fixes my case. I checked out this branch (1997df4) and ran the reproduction from #10091 against it and against main (01de197):

check on the second turn, which made no tool call	main	this PR
tool_called_check("get_weather")	passes	fails
tool_calls_present	passes	fails
tool_call_args_match	passes	fails

The first turn, which made the call, still passes all three. I also applied the tests from my closed #10245 on top of this branch: all 105 tests in test_local_eval.py pass, including the three new ones (a call from an earlier turn, a custom split that decides which calls count, and argument matching within the current turn), which fail on main.

I used Claude (Anthropic) to help run these checks; I reviewed the results myself.

[tang-ontheway]

Thanks Eduard van Valkenburg (@eavanvalkenburg) — good catch, and thanks Qiming Bao (@14H034160212) for running the reproduction.

Loop-injected messages vs. real turns. AgentLoopMiddleware injects user-role messages (the nudge, the progress summary) between iterations of a single run, and the default aggregated response contains them. The fix marks them instead of guessing:

• AgentLoopMiddleware now tags every injected message with LOOP_INJECTED_MESSAGE_PROPERTY in Message.additional_properties ("agent_framework_loop_injected"), including the streaming updates that surface them (_harness/_loop.py).
• The built-in splits (LAST_TURN, FULL) and EvalItem.per_turn_items skip loop-injected messages when they look for a turn boundary, so a nudge is never mistaken for a user turn (_evaluation.py).

With that, the response side of a loop-aggregated conversation starts after the real user query, so a tool call made in the first iteration is still found by tool_called_check / tool_calls_present / tool_call_args_match, while a per-turn item still only sees its own turn's calls.

Regression coverage. New tests in tests/core/test_local_eval.py cover the loop-aggregated shape (tool call in an earlier iteration is still matched, the nudge does not start a turn, and a real user turn after a nudge still splits), plus a loop-side test that the injected messages carry the marker (tests/core/test_harness_loop.py).

Test layout. The reusable conversation builders moved out of the test module into tests/core/local_eval_helpers.py and are imported explicitly, per the coding standard.

Qiming Bao (@14H034160212) — on the table you posted: the second turn now fails the tool checks because that turn made no tool call, which is the point of #10091 (the check is scoped to the turn under evaluation instead of the whole conversation). The first turn still passes all three, and the tests from #10245 pass unchanged on this head — the only behaviour change since 1997df4 is that loop-injected nudges no longer act as turn boundaries.