ref(google-genai): Remove set_data_normalized for the gen_ai.response.finish_reasons attribute

[alexander-alderman-webb]

Description

Remove set_data_normalized() for attributes that do not require normalization.

Since extract_finish_reasons(response) returns a list of strings, we can directly serialize to JSON.

This attribute does not adhere to our conventions as the conventions indicate type it be a string: https://github.com/getsentry/sentry-conventions/blob/0e79c16961a747152afc2c8d86311351d2c05554/model/attributes/gen_ai/gen_ai__response__finish_reasons.json#L4

Issues

Reminders

• Please add tests to validate your changes, and lint your code using tox -e linters.
• Add GH Issue ID & Linear ID (if applicable)
• PR title should use conventional commit style (feat:, fix:, ref:, meta:)
• For external contributors: CONTRIBUTING.md, Sentry SDK development docs, Discord community

[github-actions]

Semver Impact of This PR

🟢 Patch (bug fixes)

📋 Changelog Preview

This is how your changes will appear in the changelog.
Entries from this PR are highlighted with a left border (blockquote style).

---

Bug Fixes 🐛

Openai

• Avoid consuming iterables passed to the Completions API by alexander-alderman-webb in #5489
• Avoid consuming iterables passed to the Embeddings API by alexander-alderman-webb in #5491

Other

• (anthropic) Fix token accounting by shellmayr in #5490
• (google-genai) Remove agent spans for simple requests by alexander-alderman-webb in #5443

Documentation 📚

• New integration guide by alexander-alderman-webb in #5476

Internal Changes 🔧

• (agents) Add sentry skills to be used by warden in CI reviews by ericapisani in #5485
• (ai) Add configuration for dotagents by ericapisani in #5480
• (github) Add warden configuration by ericapisani in #5484

• (google-genai) Remove set_data_normalized for the gen_ai.response.finish_reasons attribute by alexander-alderman-webb in #5506

• (openai-agents) Expect new tool fields by alexander-alderman-webb in #5471
• (repo) Add .serena to .gitignore by ericapisani in #5464
• 🤖 Update test matrix with new releases (02/19) by github-actions in #5483
• 🤖 Update test matrix with new releases (02/18) by github-actions in #5475

---

_{🤖 This preview updates automatically when you update the PR.}

[sentry]

Bug: Using json.dumps instead of set_data_normalized incorrectly serializes single-element finish_reasons lists, creating an inconsistency with the streaming path.
_{Severity: MEDIUM}

Suggested Fix

Revert to using set_data_normalized(span, SPANDATA.GEN_AI_RESPONSE_FINISH_REASONS, finish_reasons). This function correctly handles both single-element lists by unpacking them and multi-element lists by JSON-serializing them, maintaining consistency with the streaming code path.

Prompt for AI Agent

Review the code at the location below. A potential bug has been identified by an AI
agent.
Verify if this is a real issue. If it is, propose a fix; if not, explain why it's not
valid.

Location: sentry_sdk/integrations/google_genai/utils.py#L884-L886

Potential issue: The switch from `set_data_normalized` to `json.dumps` for setting the
`finish_reasons` data on a span introduces a data format inconsistency. The
`set_data_normalized` function unpacks single-element lists, meaning a list like
`["STOP"]` becomes the plain string `"STOP"`. The new implementation using `json.dumps`
will serialize it as the string `'["STOP"]'`. This creates a discrepancy with the
streaming implementation, which still uses `set_data_normalized` and produces a plain
string for a single finish reason. This change will break consumers of the span data
that expect a plain string for a single reason in non-streaming scenarios.

_{Did we get this right? 👍 / 👎 to inform future reviews.}

[alexander-alderman-webb]

The aim is to have a consistent type for an attribute, so if lists are possible they should always be lists.

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.}

[cursor]

Inconsistent finish_reasons format between streaming and non-streaming

High Severity

The change creates inconsistent serialization for gen_ai.response.finish_reasons between streaming and non-streaming paths. Non-streaming now always stores as JSON array (e.g., '["STOP"]'), while streaming still uses set_data_normalized which unpacks single-element lists to plain strings (e.g., "STOP"). OpenTelemetry specifies this attribute as string[] type, requiring consistent array format. Consumers parsing this attribute will encounter different formats depending on whether streaming was used, breaking data consistency.

 

[alexander-alderman-webb]

The aim is to have a consistent type for an attribute, so if lists are possible they should always be lists.