Python: Fix Gemini client support for Gemini API and Vertex AI

[eavanvalkenburg]

Motivation and Context

This update fixes the Python Gemini connector so it works with both google-genai connection modes: the Gemini Developer API and Vertex AI. It also addresses follow-up issues that showed up in the Gemini samples when switching models and when using Gemini built-in tools.

Description

• add SDK-aligned Google settings alongside the connector aliases so GeminiChatClient can resolve either Gemini Developer API or Vertex AI configuration
• make the SDK-style GOOGLE_* values lead when both namespaces are present, including support for GOOGLE_MODEL
• preserve Gemini thought_signature data across tool-call round trips so tool-enabled conversations continue to work with newer reasoning-capable models
• resolve the effective service URL from the instantiated google-genai client instead of hardcoding endpoints
• fix core tool serialization for Gemini built-in tool specs so the samples no longer emit the Can't parse tool. warning
• refresh Gemini sample docs/comments so the sample set is complete and consistent with the repository guidance

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the Contribution Guidelines
• All unit tests pass, and I have added new tests where possible
• Is this a breaking change? If yes, add "[BREAKING]" prefix to the title of the PR.

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

This PR updates the Python Gemini connector to support both google-genai connection modes (Gemini Developer API and Vertex AI), aligns environment variable handling with the SDK-style GOOGLE_* settings, and fixes tool/thinking metadata handling so the samples and tool loops behave correctly with newer Gemini models.

Changes:

• Add GoogleGeminiSettings and update GeminiChatClient/RawGeminiChatClient to prefer GOOGLE_* settings, support Vertex AI configuration, and resolve the service URL from the instantiated genai.Client.
• Preserve thought_signature across tool-call round trips and improve tool serialization to handle Pydantic tool models.
• Refresh tests and sample docs to reflect the new configuration options and behaviors.

Reviewed changes

Copilot reviewed 14 out of 14 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
python/packages/gemini/agent_framework_gemini/_chat_client.py	Adds Google/Vertex settings, service URL resolution, Vertex AI options, and thought signature preservation in tool calls.
python/packages/gemini/agent_framework_gemini/init.py	Exports GoogleGeminiSettings from the package.
python/packages/gemini/tests/test_gemini_client.py	Adds env precedence and Vertex AI tests; updates integration-test skipping logic; adds thought_signature regression test.
python/packages/core/agent_framework/_tools.py	Extends tool serialization to support Pydantic BaseModel specs via model_dump().
python/packages/core/tests/core/test_tools.py	Adds a regression test covering Pydantic tool spec serialization.
python/packages/gemini/samples/*.py	Updates sample prerequisites/comments and adds inline “Sample output” blocks.
python/packages/gemini/samples/README.md	Documents both Gemini Developer API and Vertex AI environment variable options and how to run samples.
python/packages/gemini/README.md	Documents both auth modes and corresponding env vars.
python/packages/gemini/AGENTS.md	Updates package description and lists GoogleGeminiSettings.
python/packages/gemini/samples/init.py	Adds copyright header.

[moonbox3]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
packages/core/agent_framework
_tools.py	950	87	90%	190–191, 364, 366, 379, 404–406, 414, 432, 446, 453, 460, 483, 485, 492, 500, 539, 583, 587, 619–621, 623, 629, 674–676, 678, 701, 727, 731, 769–771, 775, 797, 912–918, 954, 966, 968, 970, 973–976, 997, 1001, 1005, 1019–1021, 1362, 1384, 1471–1477, 1606, 1610, 1656, 1717–1718, 1833, 1853, 1855, 1911, 1974, 2146–2147, 2167, 2223–2224, 2362–2363, 2430, 2435, 2442
packages/gemini/agent_framework_gemini
_chat_client.py	331	1	99%	375
TOTAL	27724	3201	88%

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
5602	20 💤	0 ❌	0 🔥	1m 33s ⏱️