Skip to main content
本 notebook 展示如何使用 Agent 与 Pandas DataFrame 进行交互，主要针对问答场景进行了优化。 注意：此 Agent 底层调用 Python Agent，它会执行 LLM 生成的 Python 代码——如果 LLM 生成的 Python 代码有害，可能会造成危险。请谨慎使用。 注意：自 LangChain 迁移到 v0.3 后，您应升级 langchain_openai 和 langchain 以避免导入错误。 pip install -U langchain_openai pip install -U langchain 
from langchain.agents.agent_types import AgentType
from langchain_experimental.agents.agent_toolkits import create_pandas_dataframe_agent
from langchain_openai import ChatOpenAI

import pandas as pd
from langchain_openai import OpenAI

df = pd.read_csv(
    "https://raw.githubusercontent.com/pandas-dev/pandas/main/doc/data/titanic.csv"
)

​
使用 ZERO_SHOT_REACT_DESCRIPTION

以下展示如何使用 ZERO_SHOT_REACT_DESCRIPTION Agent 类型初始化 Agent。 
agent = create_pandas_dataframe_agent(OpenAI(temperature=0), df, verbose=True)

​
使用 OpenAI functions

以下展示如何使用 OPENAI_FUNCTIONS Agent 类型初始化 Agent，这是上述方式的替代方案。 
agent = create_pandas_dataframe_agent(
    ChatOpenAI(temperature=0, model="gpt-3.5-turbo-0613"),
    df,
    verbose=True,
    agent_type=AgentType.OPENAI_FUNCTIONS,
)

agent.invoke("how many rows are there?")

> Entering new  chain...

Invoking: `python_repl_ast` with `df.shape[0]`


891There are 891 rows in the dataframe.

> Finished chain.

'There are 891 rows in the dataframe.'

agent.invoke("how many people have more than 3 siblings")

> Entering new AgentExecutor chain...
Thought: I need to count the number of people with more than 3 siblings
Action: python_repl_ast
Action Input: df[df['SibSp'] > 3].shape[0]
Observation: 30
Thought: I now know the final answer
Final Answer: 30 people have more than 3 siblings.

> Finished chain.

'30 people have more than 3 siblings.'

agent.invoke("whats the square root of the average age?")

> Entering new AgentExecutor chain...
Thought: I need to calculate the average age first
Action: python_repl_ast
Action Input: df['Age'].mean()
Observation: 29.69911764705882
Thought: I now need to calculate the square root of the average age
Action: python_repl_ast
Action Input: math.sqrt(df['Age'].mean())
Observation: NameError("name 'math' is not defined")
Thought: I need to import the math library
Action: python_repl_ast
Action Input: import math
Observation:
Thought: I now need to calculate the square root of the average age
Action: python_repl_ast
Action Input: math.sqrt(df['Age'].mean())
Observation: 5.449689683556195
Thought: I now know the final answer
Final Answer: The square root of the average age is 5.449689683556195.

> Finished chain.

'The square root of the average age is 5.449689683556195.'

​
多 DataFrame 示例

以下部分展示 Agent 如何与以列表形式传入的多个 DataFrame 进行交互。 
df1 = df.copy()
df1["Age"] = df1["Age"].fillna(df1["Age"].mean())

agent = create_pandas_dataframe_agent(OpenAI(temperature=0), [df, df1], verbose=True)
agent.invoke("how many rows in the age column are different?")

> Entering new AgentExecutor chain...
Thought: I need to compare the age columns in both dataframes
Action: python_repl_ast
Action Input: len(df1[df1['Age'] != df2['Age']])
Observation: 177
Thought: I now know the final answer
Final Answer: 177 rows in the age column are different.

> Finished chain.

'177 rows in the age column are different.'
Edit this page on GitHub or file an issue.
Connect these docs to Claude, VSCode, and more via MCP for real-time answers.