Independent operator level cache + doc update

[StanChan03]

Independent operator level caching + doc updates

[sidjha1]

LGTM, left some minor comments

[sidjha1]

Note: It is recommended to enable operator level caching -> Note: It is recommended to enable caching?

[sidjha1]

I think the serialize function can be cleaner. Consider something like what ChatGPT spits out

def serialize(value: Any) -> Any:
    """
    Serialize a value into a JSON-serializable format.
    Supports basic types, pandas DataFrames, and objects with a `dict` or `__dict__` method.
    """
    if value is None or isinstance(value, (str, int, float, bool)):
        # Basic types
        return value
    elif isinstance(value, pd.DataFrame):
        # Serialize DataFrame to JSON string in 'split' format
        return value.to_json(orient="split")
    elif isinstance(value, (list, tuple)):
        # Recursively serialize lists and tuples
        return [serialize(item) for item in value]
    elif isinstance(value, dict):
        # Recursively serialize dictionary keys and values
        return {key: serialize(val) for key, val in value.items()}
    elif hasattr(value, "dict"):
        # Serialize objects with a `dict` method (e.g., Pydantic models)
        return value.dict()
    elif hasattr(value, "__dict__"):
        # Serialize objects with a `__dict__` attribute
        return {
            key: serialize(val)  # Recursively serialize attributes
            for key, val in vars(value).items()
            if not key.startswith("_")  # Optionally ignore private attributes
        }
    else:
        # For unsupported types, convert to string (last resort)
        lotus.logger.warning(f"Unsupported type {type(value)} for serialization. Converting to string.")
        return str(value)

[sidjha1]

Is if hash: needed?

[StanChan03]

yea not needed