fix(test): update reasoning_effort test to expect dict format

[jquinter]

Summary

Fixes broken test that was expecting reasoning_effort to be a string, but the code now returns a dict format.

Problem

Test test_openai_model_with_thinking_converts_to_reasoning_effort was failing with:

AssertionError: reasoning_effort should be 'minimal' for budget_tokens=1024, 
got {'effort': 'minimal', 'summary': 'detailed'}

Test fails on main branch ❌ - This is a pre-existing broken test, not a regression.

Root Cause

Code in litellm/llms/anthropic/experimental_pass_through/adapters/handler.py (lines 72-74) transforms reasoning_effort from a string to a dict:

completion_kwargs["reasoning_effort"] = {
    "effort": reasoning_effort,  # "minimal"
    "summary": "detailed",
}

The test was never updated to expect this dict format.

Solution

Updated test expectations from:

assert call_kwargs["reasoning_effort"] == "minimal"

To:

expected_reasoning_effort = {"effort": "minimal", "summary": "detailed"}
assert call_kwargs["reasoning_effort"] == expected_reasoning_effort

Testing

pytest tests/.../test_anthropic_experimental_pass_through_messages_handler.py::test_openai_model_with_thinking_converts_to_reasoning_effort -v
======================== 1 passed in 0.14s ========================

Related

This test failure was reported on PR #21217, but NOT caused by PR #21217 (which only modifies test_anthropic_structured_output.py). This is a pre-existing bug that also fails on the main branch.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
litellm	Ready	Preview, Comment	Feb 15, 2026 10:09pm

[greptile-apps]

Greptile Summary

Fixes a broken test assertion in test_openai_model_with_thinking_converts_to_reasoning_effort where the expected value for reasoning_effort was a plain string ("minimal"), but the actual code pipeline transforms it into a dict ({"effort": "minimal", "summary": "detailed"}).

• The test exercises the full pipeline: translate_anthropic_to_openai first sets reasoning_effort as a string, then _route_openai_thinking_to_responses_api_if_needed transforms it into a dict with effort and summary fields for OpenAI models routed through the Responses API.
• The test assertion is updated to match the actual dict output. The companion unit test (test_non_claude_model_converts_thinking_to_reasoning_effort) correctly continues to expect a string since it tests translate_thinking_for_model in isolation.
• The PR includes evidence of the fix: passing test output is provided in the description.

Confidence Score: 5/5

• This PR is safe to merge — it only fixes a test assertion to match existing production behavior.
• The change is a minimal, correct test fix. It aligns the test expectation with the actual code behavior in the handler pipeline. No production code is modified, and the fix has been verified with a passing test run.
• No files require special attention.

Important Files Changed

Filename	Overview
tests/test_litellm/llms/anthropic/experimental_pass_through/messages/test_anthropic_experimental_pass_through_messages_handler.py	Updates test assertion for reasoning_effort from expecting a string to expecting the dict format ({"effort": "minimal", "summary": "detailed"}) that the full pipeline actually produces. Correct fix for a pre-existing broken test.

Flowchart

flowchart TD
    A["anthropic_messages_handler\n(model: openai/gpt-5.2, thinking: {type: enabled, budget_tokens: 1024})"] --> B["_prepare_completion_kwargs"]
    B --> C["translate_anthropic_to_openai\n(sets reasoning_effort = 'minimal')"]
    C --> D["_route_openai_thinking_to_responses_api_if_needed\n(transforms to dict for OpenAI models)"]
    D --> E["reasoning_effort = {effort: 'minimal', summary: 'detailed'}"]
    E --> F["litellm.completion(**completion_kwargs)"]

Loading

_{Last reviewed commit: 0812323}

[greptile-apps]

_{1 file reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}