fix(anthropic): handle arbitrary dictionary tool outputs

[QuentinFuxa]

Please ensure you have read the contribution guide before creating a pull request.

Link to Issue or Description of Change

1. Link to an existing issue (if applicable):

• Closes: # (Not applicable)
• Related: # (Not applicable)

2. Or, if no issue exists, describe the change:

Problem:
The Claude model implementation in google-adk fails to pass tool results to the model if the tool's output dictionary doesn't explicitly contain a "result" key.

In the current implementation of part_to_message_block for function_response, the code initializes content = "" and only populates it if "result" is present in the response dictionary. As a result, tools returning standard result structures (like {"status": "success", "data": [...]}) appear to have empty outputs to the Claude model, making tool use unreliable or non-functional.

Solution:
Modify the function_response handling in google/adk/models/anthropic_llm.py to be more robust:

1. If the response contains only a "result" key, extract its value (preserving existing behavior).
2. In all other cases, stringify the entire response dictionary so that Claude can perceive the full tool output (status, data, error messages, etc.).
3. Ensure the result is converted to a string to comply with ToolResultBlockParam requirements.

Testing Plan

Unit Tests:

• I have added or updated unit tests for my change.
• All unit tests pass locally.

Manual End-to-End (E2E) Tests:

1. Create an agent using google.adk.models.anthropic_llm.Claude.
2. Define a tool that returns a dictionary without a "result" key (e.g., return {"status": "success", "data": [1, 2, 3]}).
3. Ask the agent a question requiring that tool.
4. Before fix: Claude receives "" as tool output and claims it doesn't see any data.
5. After fix: Claude receives "{'status': 'success', 'data': [1, 2, 3]}" and correctly processes the information.

Checklist

• I have read the CONTRIBUTING.md document.
• I have performed a self-review of my own code.
• I have commented my code, particularly in hard-to-understand areas.
• I have added tests that prove my fix is effective or that my feature works.
• New and existing unit tests pass locally with my changes.
• I have manually tested my changes end-to-end.
• Any dependent changes have been merged and published in downstream modules.

Additional context

This fix is critical for interoperability between Gemini and Claude models within the ADK, as it allows tools designed for Gemini's flexible output handling to work seamlessly with Claude without requiring tool-specific modifications.

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[gemini-code-assist]

Summary of Changes

Hello @QuentinFuxa, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request addresses a critical issue in the Anthropic LLM integration within the ADK, specifically concerning how tool outputs are interpreted by the Claude model. By making the tool response parsing more flexible and robust, it ensures that Claude can correctly process diverse tool outputs, thereby enhancing the reliability and interoperability of agents utilizing these models.

Highlights

• Improved Anthropic Tool Output Handling: The Claude model in google-adk now correctly processes tool outputs that do not explicitly contain a "result" key. Previously, such outputs were treated as empty, hindering tool functionality.
• Robust function_response Processing: The part_to_message_block function has been updated to stringify the entire tool response dictionary if it doesn't fit the {"result": ...} or {"content": [...]} formats, ensuring Claude receives complete information.
• Enhanced Interoperability: This fix allows tools designed for Gemini's flexible output handling to work seamlessly with Claude, removing the need for tool-specific modifications and improving the ADK's overall compatibility.

🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[adk-bot]

Response from ADK Triaging Agent

Hello @QuentinFuxa, thank you for creating this PR!

This PR is a bug fix, could you please associate the github issue with this PR? If there is no existing issue, could you please create one?

In addition, could you please add unit tests for your change?

This information will help reviewers to review your PR more efficiently. Thanks!

[gemini-code-assist]

Code Review

This pull request effectively addresses the issue of handling arbitrary dictionary tool outputs for the Claude model. The logic to stringify the entire response dictionary as a fallback is a solid approach. I've added one minor suggestion to improve code robustness by ensuring type checks are consistent. Additionally, I recommend adding a new unit test case to tests/unittests/models/test_anthropic_llm.py that specifically covers this new fallback behavior with an arbitrary dictionary (e.g., {'status': 'success', 'data': [...]}). This will help prevent future regressions and verify the fix.