Use Cases

Use Cases#

Teaching and Tutorials#

Problem: Distribute pre-configured simulations to students that they can run and explore immediately.

Solution: Create a Zoomy project with multiple sessions (one per exercise), save as ZIP, host alongside the GUI. Share a single URL:

https://mbd-rwth.github.io/Zoomy/gui/?project=tutorials/course-exercises.zip

Students open the link, see all sessions in the sidebar, select one, and click Run. No installation required – everything runs in the browser via Pyodide.

For permanent archival (e.g., linked from a publication), upload the ZIP to Zenodo:

https://mbd-rwth.github.io/Zoomy/gui/?project=zenodo:12345/exercises.zip

Reproducible Research#

Problem: A publication requires that simulation results are reproducible from a single link.

Solution: Save the exact model + mesh + solver configuration as a Zoomy project ZIP. Upload to Zenodo (gets a DOI). Include the link in the paper.

The Zenodo record preserves the configuration permanently. The GUI loads it and reproduces the simulation in-browser.

Model Exploration and Debugging#

Problem: Understand how a model behaves before committing to a large simulation run.

Solution: Open the model’s code editor in the GUI, use display() calls to inspect intermediate state:

model = SME(level=2)
display(model.describe())                    # show model equations
display(mermaid="graph LR; h-->hu-->hv")     # visualize structure
display(latex=model.flux_expression())       # render flux

Enable auto mode. As you add display() calls, the script re-runs and output cells update live.

Backend Comparison#

Problem: Compare the same simulation across different solver backends (NumPy vs JAX vs AMReX).

Solution: Create one session per backend:

Session	Model	Mesh	Solver
NumPy baseline	SWE L0	1D (200 cells)	NumPy
JAX comparison	SWE L0	1D (200 cells)	JAX
AMReX high-res	SWE L0	2D (100x100)	AMReX

Connect the required backends, run each session, compare results on the Dashboard.

Batch Processing (CLI)#

Problem: Run many simulations non-interactively, e.g., for convergence studies or CI pipelines.

Solution: Use the CLI in a shell script:

for N in 50 100 200 400; do
  zoomy start
  zoomy select model swe
  zoomy select mesh create-1d
  zoomy select solver numpy
  zoomy run --local
done

Or load a pre-configured project and run:

zoomy load convergence-study.zip
zoomy session switch "Fine mesh"
zoomy run --local