Standardize reasoning field names ascross akd #284
Description
Is your feature request related to a problem?
Currently, there are various schema that uses several different name for the field that commonly represent reasoning traces such as reasoning: str, reasoning_traces: list[str], reasoning_steps: list[str]
Describe the solution you'd like
I propose we just unify this into one field name convention. I'd prefer reasoning_traces: list[str] which are list of concise reasoning to come to an answer for the agent.
Describe alternatives you've considered
Alterantively, maybe make this directly embedded into akd._base.InputSchema. Or even some metaclass at pydnatic base-model level where in compile time if we say enable_reasoning it will automatically add the reasoning_traces as a new field. Complicated, yes.
Additional context
It probably blows up output tokens if the input content is too large (as we have seen in some of our deepeval experiments in the past.