[BUG] Incorrect memory_size for SmartDataFrame

[jamie0725]

System Info

pandasai 2.3.0

🐛 Describe the bug

Hi! I've discovered a bug in pandasai v2 where there's a discrepancy between the documented behavior and actual implementation.

According to the documentation, SmartDataFrame should have a memory_size of 1 by default (designed for single-round conversations). However, the current implementation sets memory_size to 10 by default in the Agent class constructor.

class SmartDataframe:
    def __init__(
        self,
        df: Union[DataFrameType, BaseConnector],
        name: str = None,
        description: str = None,
        custom_head: pd.DataFrame = None,
        config: Config = None,
    ):
        # ... other code ...
        self._agent = Agent([df], config=config)  # No memory_size override

class Agent(BaseAgent):
    def __init__(
        self,
        # ... other params ...
        memory_size: Optional[int] = 10,  # Should be 1 for SmartDataFrame
        # ... other params ...
    ):

This inconsistency affects users who expect SmartDataFrame to handle single-round conversations as documented, but instead get multi-round conversation behavior due to the higher memory size.

Questions:

Is there a current workaround to explicitly set memory_size=1 when using SmartDataFrame?
Is there a fix planned for an upcoming release?

[1aryan102002]

self._agent = Agent([df], config=config, memory_size=1)

if we change agents default memory_size to 1 this would break existing Agent usage where multiturn is expected

the best i think is that 10 seems unlikely since the whole point of SmartDataframe is single-round.