Python: fix: AzureAIClient always shows unsupported warning (#4681)

[LEDazzio01]

Summary

Fixes #4681

AzureAIClient always emits a spurious warning:

"AzureAIClient does not support runtime tools or structured_output overrides after agent creation. Use AzureOpenAIResponsesClient instead."

…even when the tool set hasn't actually changed.

Root Cause

In _remove_agent_level_run_options(), when warn_runtime_tools_and_structure_changed is False (the use_latest_version path that fetches an existing agent rather than creating one), the code sets:

tools_changed = runtime_tools is not None

Since the framework always populates runtime_tools, this is always True, causing the warning on every single request — regardless of whether tools actually differ.

Fix

Move the warn_runtime_tools_and_structure_changed check to be the outer guard condition. When the flag is False (no agent was created by this client), we skip the entire warning block. When True, we compare the runtime tools against the creation-time set and only warn if they actually differ.

-        if runtime_tools is not None or runtime_structured_output is not None:
-            tools_changed = runtime_tools is not None
-            structured_output_changed = runtime_structured_output is not None
-
-            if self.warn_runtime_tools_and_structure_changed:
+        if self.warn_runtime_tools_and_structure_changed and (
+            runtime_tools is not None or runtime_structured_output is not None
+        ):
+            tools_changed = False
+            structured_output_changed = False
+
+            if runtime_tools is not None:

Tests

Added test_azure_ai_client_runtime_warning.py with tests covering:

• ✅ No warning when warn_runtime_tools_and_structure_changed is False (regression test for Python: [Bug]: always show warning: "AzureAIClient does not support runtime tools or structured_output overrides after agent creation." #4681)
• ✅ Warning emitted when tools actually differ after agent creation
• ✅ No warning when tools are identical after agent creation
• ✅ Warning when structured output changes
• ✅ Agent-level keys always stripped from run_options

[Copilot]

Pull request overview

Fixes a spurious AzureAIClient warning that was emitted on every request (notably in the use_latest_version=True path) by only performing the runtime-tools/structured-output mismatch warning logic when the client actually created the agent and can compare against creation-time settings.

Changes:

• Gate the runtime warning behind warn_runtime_tools_and_structure_changed and compare tool/signature deltas only when tracking is enabled.
• Add unit tests covering warning/no-warning scenarios and ensuring agent-level keys are stripped from run_options.
• Add a contribution log markdown file (appears unrelated to the PR purpose).

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
python/packages/azure-ai/agent_framework_azure_ai/_client.py	Fixes warning logic so it only triggers when tracking agent creation-time config and actual mismatches occur.
python/packages/azure-ai/tests/test_azure_ai_client_runtime_warning.py	Adds regression/unit tests for the warning behavior and option stripping.
contribution-logs/2026-03-14.md	Adds non-functional contribution status content (unrelated to the described fix).

---

You can also share your feedback on Copilot code review. Take the survey.

[LEDazzio01]

Thanks for the review! Fixed the unused imports (Mapping, MagicMock) in commit fe8a84c. The contribution-logs/2026-03-14.md is a stale artifact from the fork's main branch — will sync the fork to remove it from the diff.