Vizzy uses ChatGPT to visualize any kind of data.
Play with the demo, or check out these sample projects:
- Scripts from TV's "The Office"
- Same-sex Marriage Over Time
- U.S. State Obesity Rates
- Cause of Death News Coverage
- U.S. National Parks
A typical workflow goes like this:
Once you upload a new JSON, CSV, XML, or other data file, Vizzy will tell you what it can infer about the format, structure, and meaning of the data.
Example: "This is a CSV file containing facts about different car models, with the fields
year,make,model,horsepowerandweight
Sometimes there's not enough information in the file for Vizzy to know what's going on--maybe that list of numbers and dates represents a stock's price, or rainfall measurements in Montana, or something else. You can always edit the summary to give Vizzy more context.
A good summary will improve the quality of any visualizations.
Vizzy will give you a lit of suggestion, or you can make your own request.
Example: "Draw a scatterplot of horsepower versus weight"
Example: "Show how horsepower has grown over time"
It's best to start simple; you can add in features like legends, tooltips, color schemes, etc with follow-on prompts.
More often than not, Vizzy gets it right!
Sometimes Vizzy may reply with an error message, or a broken graph. See Failure Patterns for common issues and resolutions, or to report a new issue.
Now that you've got a working visualization, you can start tweaking it to suit your needs. You can add tooltips, change the title, create a new color scheme, or add an animation.
Example: "Color the data points by manufacturer"
Example: "Add a tooltip showing the make and model"
You can prompt Vizzy to make these changes, or you can edit the code yourself. The latter can be faster and cheaper for trivial changes, like changing a color. It can also be helpful for changes Vizzy doesn't quite understand.
You can always revert back to a previous version if you're not happy with the change.
See Contributing.md for instructions on running Vizzy locally.