feat(gemini): add thought signature preservation for thinking models

[pgrayy]

Description

Related Issues

Documentation PR

Type of Change

Bug fix
New feature
Breaking change
Documentation update
Other (please describe):

Testing

How have you tested the change? Verify that the changes do not break functionality or introduce warnings in consuming repositories: agents-docs, agents-tools, agents-cli

• I ran hatch run prepare

Checklist

• I have read the CONTRIBUTING document
• I have added any necessary tests that prove my fix is effective or my feature works
• I have updated the documentation accordingly
• I have added an appropriate example to the documentation to outline the feature, or no new docs are needed
• My changes generate no new warnings
• Any dependent changes have been merged and published

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[codecov]

Codecov Report

❌ Patch coverage is 72.72727% with 6 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/strands/event_loop/streaming.py	0.00%	2 Missing and 2 partials ⚠️
src/strands/models/gemini.py	86.66%	1 Missing and 1 partial ⚠️

📢 Thoughts on this report? Let us know!

[mkmeral]

/strands review

[github-actions]

Issue: Missing test coverage for reasoningSignature handling.

Codecov reports 0% patch coverage for the streaming.py changes. The existing parameterized tests for handle_content_block_start and handle_content_block_stop should be extended to cover the reasoningSignature field.

Suggestion: Add test cases to the existing parameterized tests in test_streaming.py:

# In test_handle_content_block_start parameters:
(
    {"start": {"toolUse": {"toolUseId": "test", "name": "test", "reasoningSignature": "YWJj"}}},
    {"toolUseId": "test", "name": "test", "input": "", "reasoningSignature": "YWJj"},
),

# In test_handle_content_block_stop parameters (with reasoningSignature):
(
    {
        "content": [],
        "current_tool_use": {"toolUseId": "123", "name": "test", "input": '{}', "reasoningSignature": "YWJj"},
        "text": "",
        "reasoningText": "",
        "citationsContent": [],
        "redactedContent": b"",
    },
    {
        "content": [{"toolUse": {"toolUseId": "123", "name": "test", "input": {}, "reasoningSignature": "YWJj"}}],
        ...
    },
),

[github-actions]

Issue: PR description is incomplete.

Please update the PR description to include:

• A detailed description of what this change does (thought signature preservation for thinking models)
• Select the appropriate type of change (appears to be "New feature")
• Confirm testing checklist items (particularly hatch run prepare)
• Link any related issues if applicable

This helps reviewers understand the context and purpose of the changes.

[github-actions]

Review Summary

Assessment: Request Changes

Overview

This PR adds thought signature preservation for Gemini thinking models by:

1. Properly handling reasoningSignature in tool use content blocks
2. Using base64 encoding/decoding for the thought signatures
3. Updating type definitions to include the new field

Key Themes

Test Coverage Gap (Important)
The streaming.py changes have 0% test coverage according to Codecov. The existing parameterized tests in test_streaming.py should be extended to cover the reasoningSignature field in both handle_content_block_start and handle_content_block_stop.

PR Documentation
The PR description template should be filled out to help reviewers understand the purpose and testing status of these changes.

What Looks Good

• ✅ Correct base64 encoding/decoding for thought signatures
• ✅ Comprehensive tests for the Gemini model request formatting
• ✅ Type definitions properly updated with NotRequired annotation
• ✅ Follows existing code patterns and conventions

Requested Changes

1. Add test cases for reasoningSignature handling in test_streaming.py
2. Consider adding a response-side test for tool use with thought signature
3. Update the PR description with context about the changes

Overall the implementation looks solid - just needs the test coverage addressed before merging.