User Guide

Learn GDBS, then check its work.

The two introductory guides are open to everyone and below on this page. The full guide, 41 topics across every module, is inside the app; it opens after you sign in or register.

Open the full guide Register Videos

Getting Started with GDBS Pro

GDBS Pro is a research-grade physics simulation platform that runs in your browser. No cluster queue, no Python install, no conda environment.

What's here

Domain modules cover the textbook ranges of:

Each module ships in two tiers:

How to read a result

Every solver reports:

What makes this research-grade

  1. GeoNum drift tracking. Engines that route their arithmetic through GeoNum carry an uncertainty band on each value, surfaced in the UI. Engines that do not are labelled [UNTRACKED] rather than shown as a pass.
  2. Validation matrix. Solvers in the matrix are benchmarked against a published, peer-reviewed value. The per-module Validation Suite runs the full matrix and marks each row green (within tolerance of the cited result) or red (out of tolerance), with the reference and DOI shown inline. Only published references are cited - the matrix is checkable against the source paper, not against us.
  3. Honest runtimes. Most interactive solvers return in seconds; heavier engines state their expected runtime up front (the Euler3D GPU example below takes about 40 seconds).
  4. Reproducibility. Results export with their inputs and drift readout, so a run can be repeated and checked.

Plugins

Beyond the built-in modules, GDBS loads drop-in plugins. Each tier (Standard, Pro, HPC) has a Plugins link in its module nav that opens a shared plugin catalog.

One HPC example to try: Euler3D GPU Taylor-Green runs a WebGPU compressible-Euler solver and checks its GPU output against the CPU path on the Taylor-Green vortex, up to 128^3 in-browser (a full default run is about 40 seconds and reports a GeoNum Exact verdict). As with every engine, this is browser-side pre-cluster prototyping with honest runtime notes, not an HPC replacement.

A share link opens the app directly into one target - a single view, engine, module, or plugin - in read-only kiosk mode. Anyone with the link (no login) sees that one target with its inputs locked to a frozen snapshot: only Run works, and any other action routes to Register. Access is live-gated, so if the link is revoked the page locks.

Visitors land in kiosk mode by opening a URL of the form ...?share={token}. Minting share links is an account action, but the kiosk view is what a trial-curious visitor sees first, so a shared result is a no-login way to show exactly one run.

Beyond running engines

Two workspace features sit alongside the physics modules:

Next steps

Your first run in 5 minutes

This walkthrough takes you from sign-in to a checked, converged plasma equilibrium.

1. Sign in

If you don't have an account, click Register on the landing page. The Equilibrium and Validation tabs used below are Pro features; the pricing section offers a 14-day free trial of Pro.

2. Open the Plasma module

Left nav -> Plasma. Alongside the standard configuration tabs you'll see the Pro tabs, including Equilibrium, Ballooning, Mercier, Multi-mode, Sensitivity, Convergence, EFIT Import and Validation.

3. Run a Soloviev equilibrium

Click Equilibrium. The form is pre-filled with ITER-class inputs:

Click Solve Soloviev. In under a second you'll see:

4. Verify against the literature

Click Validation. You'll see a table of canonical references: ITER, DIII-D, JET, NSTX, W7-X, LHD, TAE Norman, Soloviev. Click run on the ITER row. The Computed and Error columns fill in - error should be under the listed tolerance (typically 15%). The tokamak and theory rows compute a β_N; the W7-X, LHD and TAE Norman rows deliberately do not, because β_N is not the figure of merit for a stellarator or an FRC - see the Validation Suite doc.

5. Explore the s-α diagram

Click Ballooning. Defaults scan s ∈ [0, 4] and α ∈ [0, 4] on a 17² grid. Click Run s-α scan. You get the textbook ballooning stability map (Wesson §6.6.5), with the marginal-stability boundary traced in yellow.

6. Find what matters

Click Sensitivity. Click Compute sensitivities. You get a ranked bar chart of which parameter (κ, δ, A, B₀, q_edge) has the largest elasticity for β_N. For ITER-class geometry, κ usually wins.

7. Prove your result is converged

Click Convergence. Click Run convergence study. We sweep n_radial ∈ {32, 64, 128, 256, 512}, fit the convergence order via Richardson extrapolation, and tell you the smallest grid that gives < 0.1% error.

That's an equilibrium, a literature check, a stability map, a sensitivity ranking and a convergence study, in five tabs, each result showing the reference it was checked against.