Python: fix: use getattr for non-OpenAI provider response compatibility

[Oxygen56]

Summary

• Fix crash when non-OpenAI providers return responses without system_fingerprint (fixes Python: [Bug]: OpenAIChatCompletionClient fails with 'str' object has no attribute 'system_fingerprint' when provider returns non-OpenAI conformant response #6234)
• Fix crash when non-OpenAI providers return responses without output attribute (fixes Python: [Bug]: OpenAIChatClient fails with 'str' object has no attribute 'output' when provider returns non-OpenAI conformant/structured response #6235)
• Both use getattr(x, attr, None) pattern for defensive access

Fixes #6234
Fixes #6235

Test Plan

• Existing OpenAI tests pass
• Import and basic smoke test passes
• getattr pattern is consistent with existing defensive checks in codebase

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

This PR hardens OpenAI response parsing by making attribute access resilient to optional/missing fields in different response shapes.

Changes:

• Use getattr(..., default) when reading system_fingerprint from chat and streaming chat responses.
• Use getattr(..., []) when iterating response.output to avoid attribute errors when the field is absent.

Reviewed changes

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

File	Description
python/packages/openai/agent_framework_openai/_chat_completion_client.py	Makes metadata extraction tolerant to missing system_fingerprint.
python/packages/openai/agent_framework_openai/_chat_client.py	Prevents failures when response.output is not present by defaulting to an empty list.

[github-actions]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
packages/openai/agent_framework_openai
_chat_client.py	1085	148	86%	276, 289, 631–635, 643–646, 652–656, 706–713, 715–717, 724–726, 772, 780, 803, 921, 1020, 1079, 1081, 1083, 1085, 1151, 1165, 1245, 1255, 1260, 1303, 1414–1415, 1430, 1639, 1644, 1648–1650, 1654–1655, 1738, 1748, 1775, 1781, 1791, 1797, 1802, 1808, 1813–1814, 1833, 1836–1839, 1853, 1855, 1863–1864, 1876, 1918, 2009, 2031–2032, 2047–2048, 2066–2067, 2110, 2276, 2314–2315, 2333, 2413–2421, 2451, 2561, 2596, 2611, 2631–2641, 2654, 2665–2669, 2683, 2697–2708, 2717, 2749–2752, 2762–2763, 2774–2776, 2790–2792, 2802–2803, 2809, 2824
_chat_completion_client.py	363	25	93%	431, 527–528, 532, 764–771, 773–776, 864, 866, 883, 904, 932, 945, 969, 989, 1304
TOTAL	37824	4420	88%

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
7526	34 💤	0 ❌	0 🔥	2m 7s ⏱️

[eavanvalkenburg]

@Oxygen56 there are checks failing, please check, and ideally run them locally first.

[Oxygen56]

Thanks for the review @eavanvalkenburg! Fixed the pyright error — the getattr(response, "output", []) was losing type info for the match statement below. Changed to assign to a typed outputs: list variable first. CI should pass now.

[eavanvalkenburg]

@Oxygen56 both checks and tests now failing, please have a look

[Oxygen56]

@eavanvalkenburg sorry about that! Fixed the mypy error — the outputs variable was colliding with an existing outputs in a case block further down. Renamed to response_outputs: list[Any]. CI should pass now.

[eavanvalkenburg]

@Oxygen56 pyright failed... thanks for keeping at it!