[roanny]

Summary

Projects finish_reason from LlmResponse to the attributes JSON column of LLM_RESPONSE events written to BigQuery by BigQueryAgentAnalyticsPlugin, and exposes it in the generated per-event-type SQL view.

Without this projection, classifying model failure modes (MAX_TOKENS, SAFETY, MALFORMED_FUNCTION_CALL, RECITATION, etc.) via SQL against the analytics table is impossible — the signal lives only in unstructured Cloud Logging output that is not joinable to the structured event stream.

Fixes #5644.

Changes

• EventData gets a finish_reason: Optional[str] field.
• after_model_callback reads llm_response.finish_reason.name and passes it through to EventData.
• _enrich_attributes writes attrs["finish_reason"] when present (omitted entirely when None, so the JSON shape stays backwards-compatible for existing consumers).
• _EVENT_VIEW_DEFS["LLM_RESPONSE"] projects JSON_VALUE(attributes, '$.finish_reason') AS finish_reason so it shows up as a typed column on the per-event-type view.
• Schema-field description for the attributes column mentions the new key with example values.

Backwards compatibility

The new key is only added to the JSON when LlmResponse.finish_reason is not None, so events emitted before this change and events for streaming partials (where finish_reason is unset) remain unchanged. SQL consumers reading the existing documented fields are unaffected. New consumers can opt in via JSON_VALUE(attributes, '$.finish_reason') or via the finish_reason column exposed by the generated view.

Test plan

New unit tests (all passing locally; 229 passed, 5 skipped for the full plugin suite):

• test_after_model_callback_projects_finish_reason — asserts MAX_TOKENS round-trips end-to-end into the BQ row's attributes.
• test_after_model_callback_omits_finish_reason_when_absent — asserts the key is omitted when LlmResponse.finish_reason is None.
• TestEnrichAttributes::test_finish_reason_included_when_set — direct _enrich_attributes coverage.
• TestEnrichAttributes::test_finish_reason_omitted_when_none — direct _enrich_attributes coverage of the None branch.
• Existing 229 tests in the plugin suite still pass (no regressions).

To run locally:

pytest tests/unittests/plugins/test_bigquery_agent_analytics_plugin.py -k finish_reason -v

cc @sanketpatil06 @darshil3011

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[roanny]

@googlebot I signed it!

[xuanyang15]

@shobsi Could you please take a look, it is for #5644?

[rohityan]

Hi @roanny , Thank you for your contribution! We appreciate you taking the time to submit this pull request. Your PR has been received by the team and is currently under review. We will provide feedback as soon as we have an update to share.

[GWeale]

shipped in 04b8b72, v2.7.0: EventData.finish_reason, the attributes write, reading the enum name off LlmResponse in after_model_callback, and the JSON_VALUE column on the LLM_RESPONSE view. that's the approach from this pr, and the commit should have credited it. head additionally suppresses the key on streaming partials and tolerates a plain string value. #5644 is closed.

thanks for the pr and for the issue that drove it.