alisaad / plotgpt

An AI assistant for making visualizations

Introducing PlotGPT, a lightweight AI assistant to help you visualize your data.

Built on top of langchain and openai

Currently it will (most likely) only support plotting in matplotlib and seaborn

Quickstart

To get started, all you need to do is give PlotGPT a pandas dataframe to inspect.

import seaborn as sns
df = sns.load_dataset('iris')

ai = PlotGPT()
ai.inspect(df)

And then prompt it to start plotting.

ai.ask("plot sepal width vs sepal length")

PlotGPT will remember previous prompts (up to a certain point), so you can build off of previous plots, mimicking iterative explorate data analysis.

ai.ask("now colored it by specied")

By default, PlotGPT will return the plotting code. You can turn it off by setting the show_code flag.

ai = PlotGPT(show_code=False)

However, it is this author's recommendation to leave this setting on be able to sanity check the plots.

Questions

Should this completely replace my data analysis?

Probably not. Since this is an LLM, it's definitely not perfect. I'm more envisioning this as empowering initial discovery, where we use the AI to create a series of initial exploratory visualizations, and then have the human take over once you want to go deeper / more nuanced.