CSV_agent

[akabass]

@dosu-bot
i want to create a conversational chatbot that is aware of the exisiting chat and it can anlayze and talk with me about the csv i have tried to use the create_csv_agent() function but the problem is that agent cannot be conversation and it doesn't have a memory where i can ask any question its' answer is in the chat it doesn't remember or doesn't store the conversation and take it as an input or something, could someone please provide me how could i achieve this chatbot

[dosubot[bot]]

🤖

Hello,

Based on your request, it seems you want to create a conversational chatbot that can analyze and discuss CSV data, and also has the ability to remember and reference previous conversations.

To achieve this, you can modify the create_csv_agent() function to include a memory component. This can be done by creating a new class that inherits from AgentExecutor and adding a memory attribute to it. Here's an example of how you could modify the function:

class MemoryAgentExecutor(AgentExecutor):
    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)
        self.memory = []

    def execute(self, text: str) -> str:
        response = super().execute(text)
        self.memory.append((text, response))
        return response

    def recall(self, text: str) -> Optional[str]:
        for question, answer in reversed(self.memory):
            if text in question:
                return answer
        return None


def create_csv_agent(
    llm: BaseLanguageModel,
    path: Union[str, IOBase, List[Union[str, IOBase]]],
    pandas_kwargs: Optional[dict] = None,
    **kwargs: Any,
) -> MemoryAgentExecutor:
    """Create csv agent by loading to a dataframe and using pandas agent."""
    try:
        import pandas as pd
    except ImportError:
        raise ImportError(
            "pandas package not found, please install with `pip install pandas`"
        )

    _kwargs = pandas_kwargs or {}
    if isinstance(path, (str, IOBase)):
        df = pd.read_csv(path, **_kwargs)
    elif isinstance(path, list):
        df = []
        for item in path:
            if not isinstance(item, (str, IOBase)):
                raise ValueError(f"Expected str or file-like object, got {type(path)}")
            df.append(pd.read_csv(item, **_kwargs))
    else:
        raise ValueError(f"Expected str, list, or file-like object, got {type(path)}")
    return MemoryAgentExecutor(llm, df, **kwargs)

In this modification, a new class MemoryAgentExecutor is created that inherits from AgentExecutor. This class has a memory attribute which is a list that stores pairs of questions and responses. The execute method is overridden to store the question and response in the memory before returning the response. A new method recall is added that searches the memory for a given text and returns the corresponding response if found. The create_csv_agent function is then modified to return an instance of MemoryAgentExecutor instead of AgentExecutor.

This modification should allow your chatbot to remember and reference previous conversations.

I hope this helps! If you have any other questions, feel free to ask.

Best,
Dosu

Sources

• Issue: Dataframe with conversation agent for chat models
• Issue: Not answering questions out of context using RetrievalQA Chain and ConversationalChatAgent
• ConversationalRetrievalChain did not look into chat histroy while making an answer
• libs/langchain/langchain/agents/agent_toolkits/csv/base.py

---

This response is meant to be useful and save you time. It is not meant to be a precise solution, but rather a starting point for your own research.

Help me be more useful! Please leave a 👍 if this is helpful and 👎 if it is irrelevant.

If you want to continue the conversation, start your reply with @dosu-bot.