Realm
Oa Plot module interface with data layers and analysis views

Ui and Plot

Model diagnostics and retained interfaces on one GPU canvas.

One canvas from model state to presentation.

OA Plot dark model diagnostics gallery

Retained figures

Ordered artists become one reusable figure for live presentation or deterministic output.

OA Plot light model diagnostics gallery

Theme parity

Dark and light themes preserve the same composition, measurements, and artist contract.

OA Plot training and evaluation diagnostics

Model diagnostics

Training curves, heatmaps, histograms, and image grids stay inside the OA data path.

OA Plot projected scalar-field wireframes

Projected fields

Scalar fields and wireframes lower through the same compact Vulkan compositor.

Compose once. Present anywhere.

Ui is the retained command layer over Renderer. Viewer presents images and video interactively; Plot builds composable figures for training, evaluation, diagnostics, and image grids. The same scene can render live, become a semantic Image, or save headlessly through one engine-owned Vulkan compositor.

01

Compose

Record semantic artists into a small retained Figure without exposing rendering machinery.
02

Render

Lower text, grids, heatmaps, images, and vector geometry through the GPU compositor.
03

Present

Reuse the same Figure in Viewer, return a semantic Image, or write a deterministic PNG.
04

Regenerate

Change an artist, rerun the source tutorial, and verify the resulting presentation artifact.

One figure. Two language surfaces.

These generated artifacts preserve qualified donor evidence while the Rust Ui and Plot surface is ported. They are historical C++/Python results, not a claim that the Rust modules ship today.

1#include <oa/oa.h>
2
3oa::plot::Figure figure({
4 .title = "Training",
5 .rows = 1,
6 .cols = 1,
7 .theme = oa::plot::Theme::Dark,
8});
9
10auto& axes = figure.ax(0, 0);
11axes.title("Cross entropy");
12axes.xLabel("optimizer step");
13axes.yLabel("loss");
14
15axes.plot(steps, loss);

Dark, light, diagnostic, and projected views.

Every presentation is copied from the OA tutorial artifact directory with a checked SHA-256 digest. Use the carousel to inspect all six source-generated figures.

OA Plot C++ and Python parity intro
Six generated OA Plot examples

A tiny renderer foundation, already useful for real model work.

Figures, viewers, images, text, and diagnostics share one retained presentation contract. The surface stays small enough to reason about while supporting live and headless model workflows.

01

Retained figures

Compose one Figure from ordered artists, then replay it through a live Viewer or a terminal image sink.
02

Model diagnostics

Line, scatter, bar, histogram, heatmap, labels, legends, and explicit limits cover OA training and evaluation.
03

GPU compositor

Plots lower through the same Vulkan Ui compositor as images, text, grids, overlays, and Viewer controls.
04

Dark and light

Realm dark is the default; Realm light is an explicit peer with the same layout and artist contract.
05

Rust port in progress

The retained Figure and artist contracts are being translated from the qualified C++ donor without publishing a fake Rust surface.
06

Live or headless

Show the retained Figure in Viewer, save a PNG, or render a semantic Image for another OA pipeline.