Skip to main content
Every Experiment can carry four kinds of output. Config and Summary are simple; Series and Figures are where people get tripped up, because they look similar in the app but behave differently.

​
Config and Summary — small key-value facts

Both are small, named scalar values. They show up as columns and filters in the Experiments list, and as fields on the experiment detail page. The only difference is direction: Config is what you fed in, Summary is what came out.

​
Series — the plottable measurement data

A Series is a named set of numeric x/y traces. It comes from one of two places:
  1. Automatic classifier detection — Braven recognizes plain numeric arrays in your data and offers them as plottable fields.
  2. braven.plot_series(name, y, x=...) — an explicit call in your pipeline script or SDK code (see Write a pipeline script).
Series are the source of truth for interactive plots: fully explorable, one trace at a time, overlayable with any other Series in Analysis.

​
Figures — uploaded matplotlib plots

When your script calls braven.upload(fig, name) with a live matplotlib Figure, Braven does two things:
  1. Saves the rendered image exactly as fig.savefig() would — this is what you see in the experiment’s Images tab.
  2. Best-effort extracts the line/scatter data from the figure and stores it as an interactive companion — this is what shows up as a Figure tile in Analysis.
A Figure is addressed as one object, not per line. All of its subplots and traces travel together — you can’t pick out a single line from inside a Figure and compare it against something else. If you need that level of control, promote that specific trace to a first-class Series with an explicit plot_series() call.

​
Why the distinction exists

Early on, every line inside every matplotlib subplot was tracked as its own Series. In practice this flooded every picker with hundreds of entries per experiment (one real experiment produced 783), and nobody actually compared individual fit lines across figures — people compare whole figures, panel by panel. So Figures are intentionally opaque, and Series stays reserved for data you explicitly asked to be plottable.

​
Overlay rules

​
Interactive plot

An “interactive plot” isn’t a stored thing — it’s what Analysis renders at view time from one or more Series, or one or more Figures. Nothing about opening or overlaying a plot changes what’s stored.

See it in action

Continue to the Analysis guide to see how Series and Figures actually get combined on screen.