[alexcastgonz (pcbeingused333)]

Motivation and Context

With a thinking Gemini model through the Google AI connector, the usage in the response metadata under-reports the completion tokens. Gemini reports the thinking tokens in usage_metadata.thoughts_token_count, separately from candidates_token_count, and bills them as output tokens, but GoogleAIChatCompletion and GoogleAITextCompletion took completion_tokens from candidates_token_count alone. The telemetry that reads this usage, and any caller that tracks cost from it, missed every thinking token.

Description

A new get_completion_token_count helper in google_ai/services/utils.py returns candidates_token_count + thoughts_token_count (either may be missing, and it is None only when both are). Both Google AI completion services use it in _get_metadata_from_response, which streaming also goes through. The Vertex AI connector is not changed.

Tests in test_google_ai_utils.py cover the helper (both counts, one missing, zero, none, no usage metadata) and check that the chat and text services report 50 completion tokens for 10 candidate and 40 thinking tokens. The Google AI unit tests pass (93 passed), and ruff check and format are clean on the touched paths. A live Gemini call was not made.

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the SK Contribution Guidelines and the pre-submission formatting script raises no violations
• All unit tests pass, and I have added new tests where possible
• I didn't break anyone 😄

This change was made with Claude Code.

🤖 Generated with Claude Code

[Copilot]

🟢 Approval recommended

No active review issues are reported for the update to include Gemini thinking tokens in completion-token accounting.

No open issues 🎉

What changed in this PR

Corrects completion-token accounting in the Python Google AI connector by including Gemini thinking tokens, so response usage and downstream telemetry reflect billed output tokens. The Vertex AI connector is unchanged.

Changes:

• Adds a shared helper that sums candidate and thinking tokens while handling missing counts.
• Uses the helper in both chat and text completion response metadata.
• Adds unit tests for token-count edge cases and both service integrations.

File	Description
python/​tests/​unit/​connectors/​ai/​google/​google_ai/​services/​test_google_ai_utils.py	Tests the helper and chat/text usage accounting.
python/​semantic_kernel/​connectors/​ai/​google/​google_ai/​services/​utils.py	Adds the completion-token counting helper.
python/​semantic_kernel/​connectors/​ai/​google/​google_ai/​services/​google_ai_text_completion.py	Includes thinking tokens in text completion usage.
python/​semantic_kernel/​connectors/​ai/​google/​google_ai/​services/​google_ai_chat_completion.py	Includes thinking tokens in chat completion usage.

_{Review effort: Balanced}