PandasAI — Chat with Your Data Using AI

What it is

PandasAI is a Python library that adds conversational capabilities to data analysis. You ask questions about your data in natural language, and PandasAI generates the Python code or SQL query to answer them. It works with pandas DataFrames, SQL databases, CSV files, and Parquet files. It can also generate visualizations automatically.

PandasAI targets data analysts, scientists, and business users who want to explore datasets without writing code manually. It bridges the gap between natural language questions and programmatic data analysis.

How it saves time or tokens

PandasAI eliminates the translation step between 'what do I want to know' and 'how do I write the code.' Instead of composing groupby, merge, and pivot operations manually, you describe what you need. The library generates and executes the code, returning results or charts. For exploratory analysis where you run dozens of ad-hoc queries, this reduces each query from minutes to seconds. Estimated token usage is around 500 tokens per query.

How to use

1. Install PandasAI:

pip install pandasai

1. Create a SmartDataframe and ask questions:

import pandas as pd
from pandasai import SmartDataframe

df = pd.DataFrame({
    'country': ['USA', 'China', 'Japan', 'Germany', 'India'],
    'gdp': [21400, 14700, 5100, 3800, 2900],
    'population': [331, 1400, 126, 83, 1380]
})

sdf = SmartDataframe(df)
result = sdf.chat('Which country has the highest GDP per capita?')
print(result)

1. Generate visualizations:

sdf.chat('Plot a bar chart of GDP by country')
# Generates and displays a matplotlib chart

Example

Connecting to a SQL database:

from pandasai import SmartDataframe
from pandasai.connectors import PostgreSQLConnector

connector = PostgreSQLConnector(
    config={
        'host': 'localhost',
        'port': 5432,
        'database': 'analytics',
        'username': 'user',
        'password': 'pass',
        'table': 'sales'
    }
)

sdf = SmartDataframe(connector)
sdf.chat('What were the total sales by region last quarter?')

Related on TokRepo

• AI coding tools — AI-powered development and analysis tools
• Research tools — data exploration and research frameworks

Common pitfalls

• PandasAI executes generated code automatically. Review the generated code before running on sensitive datasets, especially when connected to production databases.
• Complex multi-step queries sometimes produce incorrect code. Break complex questions into simpler sub-questions for more reliable results.
• LLM API costs accumulate during heavy exploratory sessions. Consider using local models via Ollama for high-volume analysis.