Summary

This PR fixes a logic error in find_combinations where valid evaluation results were silently discarded when their value was falsy (e.g. 0.0). As a result, combinations with legitimate zero-valued scores were incorrectly excluded from the output.

Bug / Issue

evaluation_func is typed to return a float, and evaluate_combination returns (candidate_group, value) when evaluation succeeds.
However, find_combinations previously filtered results using a truthiness check (if result:), which caused valid falsy values such as 0.0 to be dropped unintentionally.

This behavior is inconsistent with the function’s type contract and makes it impossible for users to represent neutral or zero-utility combinations.

Implementation

• Replaced the truthiness check with an explicit is not None check to distinguish between “no evaluation result” and valid falsy values

• Renamed a local variable for clarity to reflect that the group is a tuple rather than a set

The change is behavior-preserving for all non-falsy values and only affects cases that were previously handled incorrectly.

Testing

• Added a regression test that verifies combinations with zero-valued evaluation results are preserved

• All existing meta_agents tests pass