Usage

Installation

To use pandasai, first install it:

# Using poetry (recommended)
poetry add pandasai

# Using pip
pip install pandasai

Before installation, we recommend you create a virtual environment using your preferred choice of environment manager e.g Poetry, Pipenv, Conda, Virtualenv, Venv etc.

Optional dependencies

In order to keep the installation size small, pandasai does not include all the dependencies that it supports by default. You can install the extra dependencies by running the following command:

pip install pandasai[extra-dependency-name]

You can replace extra-dependency-name with any of the following:

  • google-ai: this extra dependency is required if you want to use Google PaLM as a language model.
  • google-sheet: this extra dependency is required if you want to use Google Sheets as a data source.
  • excel: this extra dependency is required if you want to use Excel files as a data source.
  • modin: this extra dependency is required if you want to use Modin dataframes as a data source.
  • polars: this extra dependency is required if you want to use Polars dataframes as a data source.
  • langchain: this extra dependency is required if you want to support the LangChain LLMs.
  • numpy: this extra dependency is required if you want to support numpy.
  • ggplot: this extra dependency is required if you want to support ggplot for plotting.
  • seaborn: this extra dependency is required if you want to support seaborn for plotting.
  • plotly: this extra dependency is required if you want to support plotly for plotting.
  • statsmodels: this extra dependency is required if you want to support statsmodels.
  • scikit-learn: this extra dependency is required if you want to support scikit-learn.
  • streamlit: this extra dependency is required if you want to support streamlit.

SmartDataframe

The SmartDataframe class is the main class of pandasai. It is used to interact with a single dataframe. Below is simple example to get started with pandasai.

import pandas as pd
from pandasai import SmartDataframe
from pandasai.llm import OpenAI

# Sample DataFrame
sales_by_country = pd.DataFrame({
    "country": ["United States", "United Kingdom", "France", "Germany", "Italy", "Spain", "Canada", "Australia", "Japan", "China"],
    "sales": [5000, 3200, 2900, 4100, 2300, 2100, 2500, 2600, 4500, 7000]
})

# Instantiate a LLM
llm = OpenAI(api_token="YOUR_API_TOKEN")

df = SmartDataframe(sales_by_country, config={"llm": llm})
df.chat('Which are the top 5 countries by sales?')
# Output: China, United States, Japan, Germany, Australia

If you want to learn more about the SmartDataframe class, check out this video:

Intro to SmartDataframe

How to generate an OpenAI API Token

In order to use the OpenAI language model, users are required to generate a token. Follow these simple steps to generate a token with openai:

  1. Go to https://openai.com/api/ and signup with your email address or connect your Google Account.
  2. Go to View API Keys on left side of your Personal Account Settings.
  3. Select Create new Secret key.

The API access to OPENAI is a paid service. You have to set up billing. Make sure you read the Pricing information before experimenting.

Passing name and description for a dataframe

Sometimes, in order to help the LLM to work better, you might want to pass a name and a description of the dataframe. You can do this as follows:

df = SmartDataframe(df, name="My DataFrame", description="Brief description of what the dataframe contains")

SmartDatalake

PandasAI also supports queries with multiple dataframes. To perform such queries, you can use a SmartDatalake instead of a SmartDataframe.

Similarly to a SmartDataframe, you can instantiate a SmartDatalake as follows:

import pandas as pd
from pandasai import SmartDatalake
from pandasai.llm import OpenAI

employees_data = {
    'EmployeeID': [1, 2, 3, 4, 5],
    'Name': ['John', 'Emma', 'Liam', 'Olivia', 'William'],
    'Department': ['HR', 'Sales', 'IT', 'Marketing', 'Finance']
}

salaries_data = {
    'EmployeeID': [1, 2, 3, 4, 5],
    'Salary': [5000, 6000, 4500, 7000, 5500]
}

employees_df = pd.DataFrame(employees_data)
salaries_df = pd.DataFrame(salaries_data)


llm = OpenAI()
dl = SmartDatalake([employees_df, salaries_df], config={"llm": llm})
dl.chat("Who gets paid the most?")
# Output: Olivia gets paid the most

PandasAI will automatically figure out which dataframe or dataframes are relevant to the query and will use only those dataframes to answer the query.

Intro to the SmartDatalake

Agent

While a SmartDataframe or a SmartDatalake can be used to answer a single query and are meant to be used in a single session and for exploratory data analysis, an agent can be used for multi-turn conversations.

To instantiate an agent, you can use the following code:

from pandasai import Agent
import pandas as pd
from pandasai.llm import OpenAI


# Sample DataFrames
sales_by_country = pd.DataFrame({
    "country": ["United States", "United Kingdom", "France", "Germany", "Italy", "Spain", "Canada", "Australia", "Japan", "China"],
    "sales": [5000, 3200, 2900, 4100, 2300, 2100, 2500, 2600, 4500, 7000],
    "deals_opened": [142, 80, 70, 90, 60, 50, 40, 30, 110, 120]
    "deals_closed": [120, 70, 60, 80, 50, 40, 30, 20, 100, 110]
})

llm = OpenAI(api_token="YOUR_API_TOKEN")
agent = Agent([sales_by_country], config={"llm": llm})
agent.chat('Which are the top 5 countries by sales?')
# Output: China, United States, Japan, Germany, Australia

Contrary to a SmartDataframe or a SmartDatalake, an agent will keep track of the state of the conversation and will be able to answer multi-turn conversations. For example:

agent.chat('And which one has the most deals?')
# Output: United States has the most deals

Clarification questions

An agent will also be able to ask clarification questions if it does not have enough information to answer the query. For example:

agent.clarification_question('What is the GDP of the United States?')

this will return up to 3 clarification questions that the agent can ask to the user to get more information to answer the query.

Explanation

An agent will also be able to explain the answer given to the user. For example:

response = agent.chat('What is the GDP of the United States?')
explanation = agent.explain()

print("The answer is", response)
print("The explanation is", explanation)

Rephrase Question

Rephrase question to get accurate and comprehensive response from the model. For example:

rephrased_query = agent.rephrase_query('What is the GDP of the United States?')

print("The rephrased query is", rephrased_query)

Config

To customize PandasAI's SmartDataframe, you can either pass a config object with specific settings upon instantiation or modify the pandasai.json file in your project's root. The latter serves as the default configuration but can be overridden by directly specifying settings in the config object at creation. This approach ensures flexibility and precision in how PandasAI handles your data.

Settings:

  • llm: the LLM to use. You can pass an instance of an LLM or the name of an LLM. You can use one of the LLMs supported. You can find more information about LLMs here.
  • llm_options: the options to use for the LLM (for example the api token, etc). You can find more information about the settings here.
  • save_logs: whether to save the logs of the LLM. Defaults to True. You will find the logs in the pandasai.log file in the root of your project.
  • verbose: whether to print the logs in the console as PandasAI is executed. Defaults to False.
  • enforce_privacy: whether to enforce privacy. Defaults to False. If set to True, PandasAI will not send any data to the LLM, but only the metadata. By default, PandasAI will send 5 samples that are anonymized to improve the accuracy of the results.
  • save_charts: whether to save the charts generated by PandasAI. Defaults to False. You will find the charts in the root of your project or in the path specified by save_charts_path.
  • save_charts_path: the path where to save the charts. Defaults to exports/charts/. You can use this setting to override the default path.
  • open_charts: whether to open the chart during parsing of the response from the LLM. Defaults to True. You can completely disable displaying of charts by setting this option to False.
  • enable_cache: whether to enable caching. Defaults to True. If set to True, PandasAI will cache the results of the LLM to improve the response time. If set to False, PandasAI will always call the LLM.
  • use_error_correction_framework: whether to use the error correction framework. Defaults to True. If set to True, PandasAI will try to correct the errors in the code generated by the LLM with further calls to the LLM. If set to False, PandasAI will not try to correct the errors in the code generated by the LLM.
  • max_retries: the maximum number of retries to use when using the error correction framework. Defaults to 3. You can use this setting to override the default number of retries.
  • custom_whitelisted_dependencies: the custom whitelisted dependencies to use. Defaults to {}. You can use this setting to override the default custom whitelisted dependencies. You can find more information about custom whitelisted dependencies here.

Demo in Google Colab

Try out PandasAI in your browser:

Open in Colab

Other Examples

You can find all the other examples here.