[Eduard van Valkenburg (eavanvalkenburg)]

Motivation & Context

PR #9361 detached tool-call payload attributes from function-invocation metrics, but the metric snapshot still copied every base tool-span attribute. As a result, gen_ai.tool.description and the per-invocation gen_ai.tool.call.id remained histogram dimensions.

This follow-up keeps span telemetry unchanged while preventing span-only, potentially sensitive, and high-cardinality metadata from flowing into metrics.

Description & Review Guide

• What are the major changes? Function-duration metric attributes are now built from an explicit allowlist containing operation name, tool name, tool type, and the framework function name. Failure metrics continue to add error.type. Regression tests assert the complete metric attribute set across successful, sensitive-disabled, failed, and asynchronous invocations.
• What is the impact of these changes? Tool descriptions and call IDs remain available on execute_tool spans but no longer become metric dimensions. Existing low-cardinality dimensions and failure classification remain unchanged.
• What do you want reviewers to focus on? Please verify that the explicit allowlist preserves the intended aggregation dimensions while preventing current and future span-only attributes from entering the histogram.

Related Issue

N/A - trivial change. This is an obvious telemetry correctness and cardinality follow-up to #9361. It supersedes #9567 because this branch is rebased on current main, includes the explicit metric dictionary type required by Pyright, and strengthens regression coverage by asserting the complete metric attribute set across all relevant invocation paths.

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: Copilot assisted with implementation, regression tests, validation, and the PR description.

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]

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Copilot review overview

3 open issues

• Indexing span_attributes[...] introduces a hard failure mode (KeyError) if… · New
• This assertion can produce false failures if expression happens to equal one of the… · New
• The annotation dict[str, str] is fairly restrictive for OpenTelemetry attribute values (which… · New

What changed in this PR

This PR tightens function-invocation metric dimensions by switching from copying span attributes to an explicit allowlist, preventing span-only (potentially sensitive / high-cardinality) tool metadata from flowing into histogram metrics.

Changes:

• Build metric attributes from an explicit allowlist (operation, tool name/type, framework function tag) instead of cloning span attributes.
• Update regression tests to assert the complete metric attribute set for sync, async, and failure paths.

File	Description
python/​packages/​core/​agent_framework/​_tools.py	Replaces metric attribute construction to use an explicit allowlist rather than copying span attributes.
python/​packages/​core/​tests/​core/​test_tools.py	Strengthens assertions to validate the exact set of metric attributes recorded across invocation scenarios.

_{Review effort: Lite}

[github-code-quality]

Code coverage overview

Languages: Python

Python / code-coverage/python

main 6ae0f40	remove-tool-metric-a... 4505e79	+/-
92%	92%	0%

ℹ️ Calculations represent line coverage. Results may change as calculations are updated.

---

_{Updated October 22, 2026 11:34 UTC}

[github-actions]

MAF Automated Review — Iteration 1

Result: No findings
Scope: full PR (1 commit(s)): 9140c4f80f87
Model: gpt-5.6-sol

Overview

The PR replaces the invocation histogram's copy of span attributes with an explicit four-key metric allowlist while preserving span telemetry and failure classification. Exact-set regression assertions cover successful, sensitive-disabled, failed, and asynchronous invocations, and the helper contract guarantees the three selected span keys. No residual Critical, High, or Medium defect was established.

Reviewed the supplied pull-request change set across correctness, security/reliability, architecture, and failure behavior.
No publishable findings remained after source verification for this scope.