fix: serialize non-standard FunctionResponse dicts in AnthropicLlm

[giulio-leone]

Summary

Fixes #4779

part_to_message_block() only handled function_response dicts containing "content" or "result" keys. Any other dict structure (e.g. SkillToolset returning {"skill_name": ..., "instructions": ...}) fell through to an empty string, causing Claude to never see the tool output.

Changes

• anthropic_llm.py: Added json.dumps() fallback after the existing "content" / "result" checks, so non-standard response dicts are serialized to JSON text instead of being silently dropped.
• test_anthropic_llm.py: Added regression test verifying that a SkillToolset-style response dict is round-tripped correctly through part_to_message_block().

Testing

• All 18 test_anthropic_llm.py tests pass (including the new one)
• Full unit test suite: 4623 passed, 0 regressions

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

[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.

[rohityan]

Hi @giulio-leone , Thank you for your contribution! It appears you haven't yet signed the Contributor License Agreement (CLA). Please visit https://cla.developers.google.com/ to complete the signing process. Once the CLA is signed, we'll be able to proceed with the review of your PR. Thank you!

[giulio-leone]

Rebased onto latest main to resolve merge conflicts.

The upstream branch introduced improved json.dumps() serialization for standard result values (dict/list). I kept that improvement and preserved the fallback path for non-standard response dicts (e.g. SkillToolset returns) that was the purpose of this PR.

Both the upstream PDF document tests and this PR's non-standard response test are included.

[giulio-leone]

@rohityan Thanks for the heads-up! I'll get the Google CLA signed. Will follow up once it's done.

[giulio-leone]

I have signed the Google CLA. Could you please re-run the CLA check? Thank you!

[giulio-leone]

Hi @rohityan — the CLA is now signed and passing ✅ (it was a Co-authored-by trailer issue that has been resolved). Ready for review!

[giulio-leone]

Refreshed this branch onto current origin/main and re-ran validation from rebased head bc878087.

Validation (two consecutive clean passes, no code changes between passes):

• uv run --project /Users/giulioleone/Sviluppo/pr/work/adk-python-4807 --extra dev --extra test pyink --check src/google/adk/models/anthropic_llm.py tests/unittests/models/test_anthropic_llm.py
• uv run --project /Users/giulioleone/Sviluppo/pr/work/adk-python-4807 --extra dev --extra test isort --check-only src/google/adk/models/anthropic_llm.py tests/unittests/models/test_anthropic_llm.py
• uv run --project /Users/giulioleone/Sviluppo/pr/work/adk-python-4807 --extra dev --extra test pytest tests/unittests/models/test_anthropic_llm.py -q
• result: 40 passed on both passes
• repo-aligned mypy diff check against current origin/main: branch errors 1342, main errors 1342, new errors introduced 0

Direct behavioral proof for the non-standard FunctionResponse payload this PR targets:

• input payload:
  • {"skill_name": "search_docs", "instructions": "Use the search API to find documents.", "frontmatter": {"version": "1.0"}}
• rebased branch serializes that payload into the Anthropic tool_result.content JSON string and parses back correctly
• current origin/main now does the same

So after refreshing the branch, the runtime fix itself is now already present upstream as well; the remaining contributor-side value of this PR is the explicit regression coverage that locks the non-standard response path in place and the fact that the branch is no longer stale/conflicting.

[giulio-leone]

✅ Autonomous Validation — Pass 2/2 Clean

Branch: fix/anthropic-function-response-fallback
HEAD: bc878087
Rebased onto: current origin/main (0 commits behind)

Test Results (double-pass, no code changes between passes)

Pass	Tests	Result
1	40/40	✅
2	40/40	✅

Runner: python -m pytest <test_file> -v (Python venv)

Validated autonomously — double-pass strict protocol