Gravicode.Visual.Net.Core 1.0.0

Visual.Net

The Python visualization stack, rewritten for .NET 10 — three libraries with an idiomatic .NET API, plus a desktop IDE whose assistant writes visualization apps for you.

Bahasa Indonesia: README.id.md

Quick start

git clone https://github.com/DotNetVibeCoderz/Vibe_Charting.git
cd Visual.Net
dotnet build
dotnet test

dotnet run --project samples/Matplotlib.Console

using Matplotlib.Net;
using Matplotlib.Net.Rendering;
using Matplotlib.Net.Styling;

var figure = new Figure(9, 5, style: PlotStyle.SeabornWhiteGrid());
var ax = figure.AddSubplot();

var x = Enumerable.Range(0, 400).Select(i => i * 4 * Math.PI / 399).ToArray();

ax.Plot(x, x.Select(Math.Sin).ToArray(), "sin(x)");
ax.Plot(x, x.Select(Math.Cos).ToArray(), "cos(x)", style: LineStyle.Dashed);

ax.SetTitle("Trigonometric functions");
ax.SetXLabel("x (radians)");
ax.Legend();

figure.Save("chart.png");   // or .svg, .pdf, .jpg, .webp

What is in the box

src/            five class libraries
samples/        console apps in C#, F# and VB.Net, an Avalonia gallery, a Blazor site
tools/          VisualGen — a desktop IDE whose assistant builds visualization apps
notebooks/      Polyglot notebooks
docs/           English, with Bahasa Indonesia in docs/id/
tests/          167 tests, including 12 image baselines
datasets/       Iris and Titanic
benchmarks/     the matplotlib comparison, both harnesses and the join script

VisualGen

A desktop IDE with an assistant, Jack — the Code Bender, that writes the files and runs the build itself rather than printing code for you to copy.

dotnet run --project tools/VisualGen

An editor with syntax highlighting and a file explorer, 12 project templates that all compile, and a chat panel backed by Semantic Kernel with kernel functions for project and file operations, build and run, web search, scraping, arithmetic, the clock, and a curated Visual.Net API reference.

Works with OpenAI, Azure OpenAI, Anthropic, Google and Ollama. Claude is served by a hand-written chat completion service, because Semantic Kernel has no official Anthropic connector.

Every template is verified in CI:

dotnet run --project tools/VisualGen -- --verify-templates
# All 12 templates build.

Design

Matplotlib.Net is the foundation. Seaborn.Net layers on it exactly as seaborn layers on matplotlib, so every seaborn call returns an ordinary Figure you can then adjust with the lower-level API. Plotly.Net is independent: it emits JSON, not pixels.

GraviNum ── GraviFrame ── Visual.Net.Core ──┬── Matplotlib.Net ── Seaborn.Net
                                            └── Plotly.Net

Two rendering backends, one drawing interface. IRenderer is small on purpose — paths, text, images, clipping. SkiaRenderer produces PNG, JPEG, WebP and PDF; SvgRenderer is hand-written so that text stays as <text> elements rather than being flattened to paths. Both draw the same geometry, so a figure looks identical in every format.

The numerics are borrowed, not rewritten. NdArray and DataFrame come from Gravicode.Science — GraviNum is the NumPy layer, GraviFrame the pandas layer. Reimplementing them would have been a worse copy of code that already exists and is tested.

Contours are triangulated, not marching-squared. Each grid cell is split into two triangles and the field is treated as linear inside each one. That costs twice the primitives but removes the saddle ambiguity entirely, so both the lines and the filled bands always close.

3D is orthographic with a painter's algorithm. Every 3D artist reduces to a stream of facets, and the scene depth-sorts the whole stream before drawing — which is what lets a surface, a wireframe and a scatter interleave correctly in one figure.

How it compares to matplotlib, plotly and seaborn

The honest summary: Visual.Net covers the chart types you reach for daily and the output formats you ship, and does not attempt matplotlib's long tail — mathtext, interactive GUI backends, or map projections. The perceptual colormaps now ship matplotlib's own tables, so they match colour for colour. Where something remains approximate — the synthetic Make* datasets, GIF's 256-colour palette — it is documented rather than glossed over.

Gallery

Seaborn.Net:

Plotly.Net writes HTML, so the samples in docs/charts/ are live — open any of them in a browser.

Samples

dotnet run --project samples/Seaborn.Console
dotnet run --project samples/Visual.Net.Gallery
dotnet run --project samples/Visual.Net.Web        # then open the printed URL

Notebooks

#r "nuget: Gravicode.Visual.Net.Interactive"

using Seaborn.Net;
using Visual.Net.Core.Data;
using Visual.Net.Interactive;

VisualNetFormatters.Register();
Sns.SetTheme(SeabornStyle.WhiteGrid);

var data = Datasets.MakePenguins();
var (figure, ax) = Sns.Subplots(8, 4.5);
Sns.KdePlot(data, "body_mass_g", hue: "species", ax: ax, fill: true);

figure    // renders inline

Ready-made notebooks are in notebooks/.

Documentation

• Getting started
• Matplotlib.Net · Plotly.Net · Seaborn.Net · Visual.Net.Core
• Notebooks · VisualGen
• Comparison with the Python libraries · Benchmarks
• Bahasa Indonesia: docs/id/

Roadmap and status

PLAN.md is the roadmap; Progress.md tracks what is done.

All ten stages are done, with one deliberate exception: the packages build and the release workflow is written, but nothing has been published to nuget.org — that needs an account and a key, and a published version can be delisted but never withdrawn.

The last two stages were performance (benchmarked against matplotlib, honestly, in docs/Benchmarks.md) and distribution (CI, image-comparison baselines, and a dotnet new template pack).

Licence

MIT.

Dibuat oleh Gravicode Studios, dipimpin oleh Kang Fadhil.