pantometry

Physics for simulated worlds — a kernel that knows nothing about any particular physics, and fourteen domains built on it that do: light, heat, motion, sound, electricity, electromagnetic fields, elastic deformation, incompressible flow, flow through a packed bed, matter one atom at a time, a quantum wavefunction in a well, the motions a measured protein has, a drug distributing through a body, and a drug molecule read atom by atom and typed for a force field. Two layers above them place a simulation in the world and draw it, and neither knows a domain either.

Dimensions live in the type system, so Length + Time does not compile. Conservation is audited rather than assumed, and a Violation names what went missing and where. Every result is reproducible bit for bit — across platforms, across optimisation levels, in WebAssembly, and under sixteen threads as readily as one.

Every claim is checked against a closed form or an independent computation, never against another implementation that might be wrong in the same direction. Where no closed form exists, EVIDENCE.md says so.

cargo run --release --example lens_spots -- lens.svg. The three SVGs here are an example's output and every example checks itself against a closed form. The four screenshots are the figures CI cannot take — they need a GPU, and the editor's needs a display — so each carries a machine-readable record of what it is a picture of, and the thing that writes that record fails when the picture goes stale.

cargo run --release --example optical_bench bench.svg, or bench.html for the same scene rotatable in a browser, or bench.gltf for somebody else's renderer. Every surface in it is drawn as the surface the rays are traced against, checked point by point against the same intersection the trace uses — and the two flats are drawn to the footprint of the beam, because a prescription gives a fold mirror no size at all.

The same run through the viewer, which shades it: pantometry view bench.json. The glass is a solid here rather than an outline — 6 404 vertices and 12 288 triangles, every edge of each element shared by exactly two faces and every vertex still a point a ray lands on. A line behind it is hidden by it, which is a thing measured rather than looked at.

A different domain through the same viewer, and the one figure here that moves: cargo run --release --example ligand_binding closing.json, then pantometry view closing.json, where space plays it. This is the window — all forty-eight frames of it, carrying what the run measured as it goes and what the keys do, because a picture of an application that says neither is a picture of a shape. The enzyme is walking its own softest mode — the model is shown only the open structure, and that one direction scores 0.799 against the motion the enzyme is observed to perform, where a direction that knew nothing would score 0.039 in 642 dimensions. What the animation adds to that number is the thing no still can: the lids close over the molecule, and then leave again, because a harmonic direction is a straight line through a curved path and only goes so far — 7.13 Å from the closed form down to 4.29 and back.

The backbone is a tube of 7 680 triangles swept along 640 spline samples that pass through every one of the 214 alpha carbons. Its radius is not a taste — the chain turns through six places where the radius of curvature is about 0.9 Å, and a tube wider than that passes through its own axis and renders as a bead, so the run measures the ratio and refuses above one. It is also checked closed, right-side-out, and clear of the places the chain folds back past itself. The molecule is C20 N10 O22 P5 at Bondi radii, and that formula is the check: AP5A is C20H29N10O22P5, a structure at this resolution has no hydrogens, and counted before the reader's alternate-location rule the same records come to O28 P6 — sixty-four atoms, seven of them one atom modelled twice.

And the editor, on a protein of its own: pantometry edit 31-a-protein-shaking-at-body-temperature.json --run. The scene is the text — a PDB file, a cutoff, a stiffness and a temperature — and everything else is what running it produced. The seven readings on the right are the model answering for itself: six rigid modes, which is how many a body with nothing holding it has, and a mode separation of 1.4e14 saying the softest motion is well clear of the rest.

Install

cargo add pantometry     # one dependency, all twenty published crates
pip install pantometry   # or from Python — see bindings/python

pantometry = "0.22"

Run

In a clone of this repository:

cargo run --release --example melting            # a crystal melting, read off its own structure
cargo run --release --example beam_hot_spot      # a laser on a mirror, and the hot spot a lumped
                                                 # model misses
cargo test --workspace                           # the suite, all against closed forms

Most examples take an output path and draw the result — an SVG, or a page you open:

cargo run --release --example lens_spots -- lens.svg

Everything a person runs is one binary, in app/ — the CLI, the viewer, the editor and the GPU accelerator:

cd app
cargo run --release -- run    scene.json out.html   # run it and write an asset, by extension
cargo run --release -- verify scene.json --deep     # margins, determinism, sweeps, geometry
cargo run --release -- edit   scene.json            # the editor: check as you type, run, verify

pantometry edit with no file opens on a start screen: a new project, a scene to open, or one you had open before.

pantometry edit 29-a-designed-bracket-becomes-cells.json --run. An L-bracket designed as an STL, filled into 4 100 cells at 2 mm and cooled through a bolt pad and to still air. The scene is the text on the left and the picture is what running it produced — the readings, the colour bar and the frame transport are the run's, and the note in orange is the mesher saying what it did.

A laser on a mirror. The thin line is the lumped mean, which every curve has exactly — and the thick ones are what a lumped model cannot tell you: the peak is well above it.

Where everything else is

read	for
AGENTS.md	the whole API on one page. cargo run --example agents_quickstart is a runnable version, including a deliberate 10% energy leak so you can see what the audit says
ARCHITECTURE.md	the map: three layers, the twenty crates, what is built, what is missing, and what is deliberately not here
EVIDENCE.md	how the claims are checked, and what checking them found — including two real defects the conservation audit reported as clean
EXAMPLES.md	every example and what it demonstrates
app/README.md	the binary: CLI, viewer, editor, accelerator
CONTRIBUTING.md	the conventions, and everything CI runs in the order it runs it
CHANGELOG.md	what was found, as well as what was added
CLAUDE.md	working on pantometry rather than with it

pantometry-world was the first consumer — the editor has been the second since it needed World::advance made public, and the Python bindings are a third — and its first job was not to be a good application but to use the SDK the way a stranger would. app/pantometry-world/FRICTION.md is what it came back with: forty-nine findings, forty-three fixed and six argued down in writing — forty-eight from building it and one from an example, which says so. Not one of the library's own tests could have found any of them: a test is written by somebody who already knows the shape.

Citation

If this software contributes to work you publish, please cite it. CITATION.cff is beside this file and GitHub renders it as a Cite this repository button; the same content as BibTeX:

@software{park_pantometry,
  author  = {Park, Younghyeon},
  title   = {pantometry: physics for simulated worlds, checked against closed forms},
  version = {0.22.0},
  year    = {2026},
  doi     = {10.5281/zenodo.22979609},
  url     = {https://doi.org/10.5281/zenodo.22024817},
  note    = {ORCID: 0000-0002-4733-5049. The `doi` is 0.22.0's; the `url` is the concept DOI,
             which always resolves to the newest version. Both move with a release --
             this block named 0.16.0 through the whole of 0.17.0, and its `version`
             was left at 0.19.0 through the 0.20.0 bump because no list covered it},
  license = {MIT OR Apache-2.0}
}

Co-authorship is not requested, and here is why that is the honest answer

It would be easy to write "please add me as an author" here, and it would be wrong on three counts.

It could not be required. MIT OR Apache-2.0 is already granted irrevocably, and a licence with an authorship condition attached would no longer be an open-source licence under the OSI definition — it would impose a restriction the two licences people already rely on do not have.

It is contrary to how authorship works. ICMJE and COPE both rest authorship on an intellectual contribution to the specific work being published. Supplying a tool is not that, however much work the tool was. Authorship asked for on those grounds is what the literature calls gift or honorary authorship, and an editor who saw it as a condition of use would strike it.

And it would cost the thing it was reaching for. A tool with a tax on it loses users, and users are where citations come from.

What is invited instead

If you build something with this, I would like to hear about it — and if there is a question the library cannot answer yet, open an issue. Where that turns into real involvement in your work — a domain built for your problem, an analysis argued through together — a joint contribution is earned in the ordinary way, and I am glad to do it. That is an invitation and not a condition.

An acknowledgement, if the work does not warrant more, is always welcome and never expected.

Licence

Dual-licensed under either MIT or Apache-2.0 at your option — the Rust ecosystem's convention, and not indecision. Each covers a gap the other has. Apache-2.0 carries an explicit patent grant and MIT does not, which is the first thing a corporate legal review asks about. MIT is compatible with GPLv2 and Apache-2.0 is not, which matters here because the scientific-computing world has plenty of GPLv2 code. Offering both lets a consumer take whichever they need, so OR is strictly less restrictive than either alone.

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 licence, shall be dual-licensed as above, without any additional terms or conditions.

Every dependency is permissive too, and that is checked rather than remembered: deny.toml holds an allow-list and CI fails on anything outside it. Twelve external crates at the time of writing, of which three reach a published artifact — glam, serde and serde_core, all under the same MIT OR Apache-2.0. serde_json and its three transitive crates are a dev-dependency of four of the published crates, so they reach no artifact either. The rest are compile-time or test-only.

Two crates were added in 0.9.0 and the count did not move. pantometry-view has no external dependency at all — pantometry-scene, and two of this workspace's own crates to develop against: SVG and HTML are text, so a renderer is a format! and a file write, and the alternative would have put a plotting stack into the tree of every consumer who wanted a picture.