create_csv_agent#

langchain_experimental.agents.agent_toolkits.csv.base.create_csv_agent(llm: LanguageModelLike, path: str | IOBase | List[str | IOBase], pandas_kwargs: dict | None = None, **kwargs: Any) → AgentExecutor[source]#

Create pandas dataframe agent by loading csv to a dataframe.

Parameters:

llm (LanguageModelLike) – Language model to use for the agent.
path (Union[str, IOBase, List[Union[str, IOBase]]]) – A string path, file-like object or a list of string paths/file-like objects that can be read in as pandas DataFrames with pd.read_csv().
pandas_kwargs (Optional[dict]) – Named arguments to pass to pd.read_csv().
kwargs (Any) – Additional kwargs to pass to langchain_experimental.agents.agent_toolkits.pandas.base.create_pandas_dataframe_agent().

Returns:

An AgentExecutor with the specified agent_type agent and access to a PythonAstREPLTool with the loaded DataFrame(s) and any user-provided extra_tools.

Return type:

AgentExecutor

Example

from langchain_openai import ChatOpenAI
from langchain_experimental.agents import create_csv_agent

llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0)
agent_executor = create_pandas_dataframe_agent(
    llm,
    "titanic.csv",
    agent_type="openai-tools",
    verbose=True
)