$ infer --source regional-sales.csv
Select a mark for its exact value.
The chart choice is a fixed rule, not a model. A column is typed as a date, a number, or a category by inspecting every value in it; a category paired with a number becomes a bar — or a pie, when there are five or fewer categories and the number is a whole-number total whose shares mean something. A date paired with a number becomes a line, two numbers become a scatter, one number on its own becomes a histogram, and two categories with a number become a heatmap. Nothing here was trained on anything, and the same rule decides the chart whether the data is one of the four samples or something you paste yourself.
The suggested views are a script, not a conversation. Each one preselects the columns to look at — it does not understand the question, and there is no chat behind it. What is real is that the chart form is re-derived from scratch every time the columns change.
The four built-in tables are illustrative sample data, not a live feed or a real company’s numbers. Anything you paste is parsed in this page and never leaves it — there is no upload, and nothing is kept after reload.