[pylaterreur]

Description

Since google-genai 2.34.0 (2026-10-29), the loongsuite-instrumentation-google-genai-latest jobs fail on Python 3.11 to 3.15. On 3.10 they pass, because 2.34.0 needs Python 3.11 and 3.10 still gets 2.33.0. test_instrumentation_does_not_break_core_functionality (sync and async) gets <MagicMock name='mock.model_copy().id'> instead of test-id. main and every pull request that runs the full matrix fail the same way.

google-genai 2.34.0 adds automatic continuation to interactions.create (googleapis/python-genai#3226). The public create merges the hops it sends, even a single one, into a copy made with model_copy(update=...), both for the interaction and for its usage. The tests return MagicMock interactions from the generated create, and MagicMock().model_copy() is a new mock without the configured fields.

Only the tests are affected. The instrumentation wraps the generated create and reads each hop's response, as before. With real objects over a mocked HTTP API:

• a single-hop interaction gives the same object and the same span on 2.33.0 and 2.34.0;
• an interaction continued over two hops gives one span per request.

This PR mocks the interaction and its usage as pydantic-like models whose model_copy keeps their fields.

Type of change

• Bug fix (non-breaking change which fixes an issue)

How Has This Been Tested?

tox -c tox-loongsuite.ini -e py3{10,11,12,13,14,15}-test-loongsuite-instrumentation-google-genai-{latest,oldest} (tox-uv, GitHub CPython builds), online and offline (bwrap --unshare-net):

• latest, Python 3.11 to 3.15: 226 passed, with google-genai 2.34.0 and with 2.33.0. Without this change, 2.34.0 fails the same 2 tests as CI.
• latest, Python 3.10 (google-genai 2.33.0): 226 passed.
• oldest (google-genai 1.32.0), Python 3.10 to 3.15: 182 passed.
• tox -e precommit passes.

Does This PR Require a Core Repo Change?

• No.

Checklist:

• Followed the style guidelines of this project
• Changelogs have been updated
• Unit tests have been added
• Documentation has been updated

🤖 Generated with Claude Code

https://claude.ai/code/session_01GwUvhmzgfkiZ6vo4uGyesN

🤖 Generated with Claude Code

https://claude.ai/code/session_01GwUvhmzgfkiZ6vo4uGyesN

[Copilot]

🟢 Approval recommended

No active review issues were identified, with the changes focused on updating test fixtures and documenting compatibility with google-genai 2.34.0.

No open issues 🎉

What changed in this PR

Updates Google GenAI Interactions test fixtures for compatibility with google-genai 2.34.0, whose interactions.create implementation copies interaction and usage models via model_copy.

Changes:

• Added field-preserving, pydantic-like MagicMock fixtures.
• Updated interaction and usage mocks to preserve fields across copies.
• Documented the compatibility fix in the changelog.

File	Description
instrumentation-loongsuite/​loongsuite-instrumentation-google-genai/​tests/​interactions/​util.py	Adds model-copy-aware mocks for interaction responses and usage data.
instrumentation-loongsuite/​loongsuite-instrumentation-google-genai/​CHANGELOG.md	Records the google-genai 2.34 compatibility fix.

_{Review effort: Lite}

[pylaterreur]

Same as on #295: this PR now also contains #295's docstring fix (same content), so its CI can go green on its own. Merging either PR makes the other's duplicate commit empty.

🤖 Generated with Claude Code

https://claude.ai/code/session_01GwUvhmzgfkiZ6vo4uGyesN

[pylaterreur]

@ralf0131 same as on #295: all checks pass now, and the approval was dismissed by my push of the other PR's commit (same content). Could you approve again?

🤖 Generated with Claude Code

https://claude.ai/code/session_01GwUvhmzgfkiZ6vo4uGyesN