feat(google_genai): add google_server_tools to GoogleGenAIChatGenerator

[Aftabbs]

Summary

• Adds google_server_tools: list[dict[str, Any]] | None = None parameter to GoogleGenAIChatGenerator.__init__
• Enables users to pass Google's built-in server-side tools (e.g. google_search, code_execution) to the Gemini API
• Server tools are merged with Haystack Tool/Toolset at request time and passed in GenerateContentConfig.tools
• Serializes/deserializes cleanly via to_dict() / from_dict()

Motivation

Google's Gemini API supports built-in tools that are not Haystack Tool objects and cannot be expressed as function declarations. For example:

from google.genai import types
# Native Google search tool
types.Tool(google_search=types.GoogleSearch())

# Code execution tool
types.Tool(code_execution=types.ToolCodeExecution())

Currently there is no way to pass these through GoogleGenAIChatGenerator because the tools parameter only accepts ToolsType (Haystack Tools/Toolsets). This PR adds google_server_tools to fill that gap.

Usage:

generator = GoogleGenAIChatGenerator(
    model="gemini-2.5-flash",
    google_server_tools=[{"google_search": {}}],
)

Or combined with Haystack tools:

generator = GoogleGenAIChatGenerator(
    model="gemini-2.5-flash",
    tools=[my_haystack_tool],
    google_server_tools=[{"google_search": {}}, {"code_execution": {}}],
)

Relation to existing work

Follows the same pattern as anthropic_server_tools introduced in #3386 for AnthropicChatGenerator.

Test plan

• test_init_with_google_server_tools — verifies _google_server_tools is stored on init
• test_to_dict_with_google_server_tools — verifies serialization round-trip
• test_from_dict_with_google_server_tools — verifies deserialization round-trip
• test_to_dict_default_google_server_tools_is_none — verifies default is None
• test_run_with_google_server_tools — verifies native tool is passed to GenerateContentConfig
• test_run_merges_haystack_and_google_server_tools — verifies Haystack + native tools are merged

[github-actions]

Heads-up for maintainers

This PR is from a fork and touches integrations whose integration tests require API keys.
Those tests are skipped in CI because fork PRs don't have access to repo secrets for security reasons.

Affected integrations:

• google_genai

Please run the integration tests locally (hatch run test:integration inside each folder) before approving.

[github-actions]

Coverage report (google_genai)

Click to see where and how coverage changed

File	Statements	Missing	Coverage	Coverage (new stmts)	Lines missing
integrations/google_genai/src/haystack_integrations/components/generators/google_genai/chat
chat_generator.py					597
Project Total

This report was generated by python-coverage-comment-action

[anakin87]

Thanks for the contribution.

I found some opportunities for improvement.

Quick question: have you tried this feature locally yourself?

[anakin87]

Could you explain how this syntax works?

We should support all Google built-in tools if possible, also allowing to pass potential parameters they accept.

So let's verify that this syntax is functional for this goal or choose another one; also provide users with an example in docstring and a link to Google tools.

[anakin87]

Let's incorporate these new tests into existing tests or at least move them close to other similar ones

[anakin87]

Let's add a live integration test in the appropriate section.

[Aftabbs]

Yes — I tested the parameter wiring and serialization locally (unit tests pass). For the live API path, I've added test_live_run_with_google_server_tools to TestGoogleGenAIChatGeneratorInference which exercises the full end-to-end flow when GOOGLE_API_KEY is present. Since integration tests require an API key (which fork PRs don't have in CI), the live test is guarded and will only run with a real key. The unit tests cover the serialization, deserialization, and that the google_server_tools list reaches the request params correctly.