fix(anthropic): allow images in user messages while filtering model turns

[Happy-ryan]

Link to Issue or Description of Change

Related issue: #3250

Problem:
The content_to_message_param function in anthropic_llm.py was filtering out image parts from all message roles, including user messages. This was a bug because Claude API supports images in user turns but not in model/assistant turns. As a result, users couldn't send images to Claude models even though the API supports it.

Solution:
Fixed the bug by adding a role check (content.role != "user") before filtering image parts. Now:

• Images in user messages are preserved and sent to Claude API ✅
• Images in model/assistant messages are filtered out (as intended) ✅
• A warning is logged when images are filtered to help with debugging

This ensures Claude models in ADK can properly handle image attachments in user messages while preventing errors from unsupported images in model turns.

Testing Plan

Unit Tests:

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

Test Results:

$ pytest tests/unittests/models/test_anthropic_llm.py::test_content_to_message_param_with_images -v
==================================================== test session starts =====================================================
platform darwin -- Python 3.11.14, pytest-8.4.2, pluggy-1.6.0 -- /Users/joonpyo.hong/Documents/GitHub/adk-python/.venv/bin/python3
cachedir: .pytest_cache
rootdir: /Users/joonpyo.hong/Documents/GitHub/adk-python
configfile: pyproject.toml
plugins: mock-3.15.1, asyncio-1.2.0, anyio-4.11.0, xdist-3.8.0, langsmith-0.4.38
asyncio: mode=Mode.AUTO, debug=False, asyncio_default_fixture_loop_scope=function, asyncio_default_test_loop_scope=function
collected 3 items                                                                                                            

tests/unittests/models/test_anthropic_llm.py::test_content_to_message_param_with_images[user_role_with_text_and_image] PASSED [ 33%]
tests/unittests/models/test_anthropic_llm.py::test_content_to_message_param_with_images[model_role_with_text_and_image] PASSED [ 66%]
tests/unittests/models/test_anthropic_llm.py::test_content_to_message_param_with_images[assistant_role_with_text_and_image] PASSED [100%]

Manual End-to-End (E2E) Tests:

Tested with Claude agent processing messages containing images:

• User messages with images: Successfully processed ✅
• Mixed content (text + images): Text preserved, images handled correctly based on role ✅

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 PR fixes issue #3250 point 2: "No support for embedded attachments in message part".

Bug Fix Details:

• Before: Image parts were filtered from ALL roles → users couldn't send images to Claude
• After: Image parts only filtered from model/assistant roles → users can now send images properly

This enables proper Claude API image support in ADK while maintaining safety by preventing unsupported images in model responses.

[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 @Happy-ryan, 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 bug in the Anthropic LLM integration that prevented the proper handling of image attachments in user messages. By introducing a role-specific check, the system now correctly allows images to be sent by users to Claude models, while maintaining the necessary filtering for model-generated content. This enhancement significantly improves the multimodal capabilities of the ADK for Claude models, ensuring alignment with the Claude API's specifications.

Highlights

• Bug Fix: Image Part Filtering: Resolved an issue where image parts were incorrectly filtered from all message roles within the content_to_message_param function, preventing users from sending images to Claude models.
• Role-Based Image Handling: Implemented a role check (content.role != "user") to ensure that images in user messages are preserved and sent to the Claude API, while images in model/assistant messages continue to be filtered as per API limitations.
• Improved Debugging: Added a warning log when image parts are filtered, providing better visibility and assistance for debugging.
• New Unit Tests: Introduced new parameterized unit tests to thoroughly validate the correct behavior of image handling in content_to_message_param across different message roles and confirm warning log invocation.

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. ↩

[gemini-code-assist]

Code Review

This pull request correctly fixes a bug where image parts were being filtered from user messages sent to Claude models. The logic change is concise and effective. The addition of comprehensive, parameterized unit tests is excellent, as it validates the fix across different message roles (user, model, assistant) and ensures the expected logging behavior. I have one suggestion to make the new warning log message more specific, which will improve its utility for debugging. Overall, this is a well-executed fix.

[gemini-code-assist]

This logic correctly filters images for non-user roles. However, the accompanying warning message on line 146 is now slightly misleading as it only mentions "model turns", but this block also executes for the "assistant" role. To improve debugging clarity, consider making the log message more general or dynamic. Note that this would require updating the assertion in the new unit test.

Suggested change

	if content.role != "user" and _is_image_part(part):
	logger.warning("Image data is not supported in Claude for model turns.")
	continue
	if content.role != "user" and _is_image_part(part):
	logger.warning("Image data is not supported in Claude for role '%s'. Filtering image part.", content.role)
	continue