# Gemini Pro leaking internal Python unit tests / SchemaNode code instead of responding

**URL:** <https://discuss.ai.google.dev/t/gemini-pro-leaking-internal-python-unit-tests-schemanode-code-instead-of-responding/128320>\
**Category:** Gemini API\
**Tags:** python, models, bug, gemini\
**Created:** [March 3, 2026, 4:00pm UTC](https://discuss.ai.google.dev/t/gemini-pro-leaking-internal-python-unit-tests-schemanode-code-instead-of-responding/128320 "2026-03-03T16:00:05Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![RubyEnrique](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/rubyenrique/32/32083_2.png) [@RubyEnrique](https://discuss.ai.google.dev/u/RubyEnrique)\
**Post date:** [March 3, 2026, 4:00pm UTC](https://discuss.ai.google.dev/t/gemini-pro-leaking-internal-python-unit-tests-schemanode-code-instead-of-responding/128320/1 "2026-03-03T16:00:05Z")

</div>

Hello,

Starting around an hour ago (22:40 UTC+8), Gemini pro 3.1 has been bugging out and have been outputting raw python unit tests and internal schema definitions (specifically referencing SchemaNode, extract\_content\_parts, and unittest), instead of providing normal text responses.

I’m wondering if anyone else had the same problem, or maybe a heads up on what happened. An example output of the model is `self.assertEqual(extracted, [('text', 'In real estate and financial advisory...')])`, and the problems remained consistent even if I started a new chat or switched accounts. For reference, I am using Gemini Web as a Gemini pro user using Gemini 3.1.