fix: Improve type conversion for LLM parameters

[devin-ai-integration]

Fix Type Conversion Issues in LLM Parameter Handling

Changes Made

• Improved type conversion for LLM parameters in Agent class
• Added proper handling of None values and type validation
• Fixed tool-related type checking in crew tests
• Added pytest.ini to force VCR to use recorded cassettes in CI

Testing

• Local tests pass
• Type checking errors resolved
• Improved error handling for parameter conversion

Link to Devin run: https://app.devin.ai/sessions/a3d1b11e91bf43b2b0f5d7090d26f825

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add "(aside)" to your comment to have me ignore it.
• Look at CI failures and help fix them

⚙️ Control Options:

• Disable automatic comment and CI monitoring

[joaomdmoura]

Disclaimer: This review was made by a crew of AI Agents.

Code Review Comment

Overview

This pull request introduces several significant changes aimed at improving the handling of UserWarnings from the litellm library and managing LLM parameters effectively. Overall, the modifications enhance code clarity, suppress unnecessary log clutter, and ensure dependency management is consistent.

Key Improvements

1. Warning Suppression

• Modification: Context managers have been implemented in src/crewai/llm.py, src/crewai/utilities/internal_instructor.py, and src/crewai/utilities/token_counter_callback.py to suppress UserWarnings.
• Code Recommendation: Please ensure that the use of warnings follows this pattern for any new warnings that may arise:

import warnings
with warnings.catch_warnings():
    warnings.simplefilter("ignore", UserWarning)
    import litellm
    from litellm import get_supported_openai_params

2. Dependency Version Management

• Modification: The litellm version has been updated in pyproject.toml and uv.lock from >=1.44.22 to >=1.56.4.
• Code Recommendation: Maintain a consistent version specification across files to prevent discrepancies:

dependencies = [
    "litellm>=1.56.4",
]

3. Type Conversion & Parameter Handling

• Modification: Enhancements made in src/crewai/agent.py for type validation and parameter filtering improve reliability.
• Code Example: Implement safeguards around value conversions:

def safe_convert_parameter(key: str, value: Any) -> Tuple[Any, Optional[str]]:
    try:
        converted_value = convert_parameter(key, value)
        return converted_value, None
    except (ValueError, TypeError) as e:
        return None, f"Error converting {key}: {str(e)}"

4. Testing Improvements

• Modification: In tests/pytest.ini and tests/crew_test.py, improvements were made to the test configurations.
• Code Recommendation: Ensure that parameterized tests are used for better coverage:

@pytest.mark.parametrize("param_key,param_value,expected", [
    ('temperature', '0.7', 0.7),
    ('invalid_param', 'value', None),
])
def test_parameter_conversion(param_key, param_value, expected):
    result = convert_parameter(param_key, param_value)
    assert result == expected

Historical Context

In reviewing related pull requests that focused on dependency management and warning suppression, it’s essential to highlight the improvements made in prior iterations. The transition from litellm version 1.44.22 to 1.56.4 reflects lessons learned from previous PRs where warnings hindered development efficiency. Consistent dependency handling enhances the stability of our project.

Suggested Next Steps

• Enhance Test Coverage: Including more comprehensive tests for edge cases could ensure robustness against future changes.
• Documentation: Consider documenting the parameter handling methods clearly, especially for new developers onboarding to the project.

Conclusion

The changes in this PR are commendable as they address critical areas of concern within the codebase. Implementing the suggested improvements will further enhance code maintainability and user experience while adhering to best practices in dependency and error management.

I appreciate the thoroughness of the implementation and look forward to seeing the continued evolution of our code standards!