[Go] Google GenAI: reasoning/thinking content not extracted from model responses

[yesudeep]

Summary

The Go Google GenAI plugin does not support extracting reasoning/thinking content from model responses. The conformance test expects reasoning content but finds none:

reasoning content not found

Expected behavior

When a model returns reasoning/thinking tokens (e.g. Gemini models with thinking enabled), the Go plugin should extract and expose them in the response, matching the Python plugin's behavior.

Reproduction

cd py && bin/conform check-model google-genai --runtime go --runner native

Look for "Reasoning conformance" failure on googleai/gemini-3-pro-preview.

Affected models

• googleai/gemini-3-pro-preview

Notes

This is a feature gap — the Go plugin may not yet support the reasoning/thinking content extraction that was added to the Python plugin.

[hugoaguirre]

@yesudeep I'll investiagate on this. Reasoning tokens/output should be part of the standard response. That information is only provided when explicitly setting the thinking configuration on the generate request.