License

prophys is proprietary software by RhineQC GmbH. This page explains what that means in practice: what runs without a license, what needs one, and how the two are enforced.

Free tier vs. licensed use

prophys is one installable package with two parts inside it. The Python frontend is the DSL you write models in, fully open to read, documented, and unrestricted. The compiled native core, prophys._core, performs the actual numerics and enforces a free-tier object limit:

Tier	Limit	What counts
Free	Up to 250 objects per structural kernel call	Points in a PointList, vertices in a Polygon, segments in a LineList, halfspaces in a Polyhedron, cells in a raster Field, …
Licensed	No limit	Same, unbounded

A model that stays within the free tier, the kind of model this documentation’s examples and gallery use, runs with no license file at all. This is deliberate: evaluation, teaching, prototyping, and small studies should never require going through a licensing step.

What does not count towards the limit

The 250-object limit applies specifically to structural model size, the geometry and data structures you declare. It does not apply to:

• Monte-Carlo marginalization sample counts (n_samples in ProbabilityModel.compile)

• The number of observations passed to calibrate() / finetune()

• Dense evaluation grids used for plotting (e.g. a risk-map Grid)

• Training iterations, gradient steps, or optimizer calls

In other words, you can calibrate a small model against as much data, at as high a Monte-Carlo precision, as you like, entirely for free. The license question is only ever “how big is the model itself, how many points, vertices, segments, or raster cells does it declare”, matching the intuitive reading of “250 objects, e.g. points or contents in a PointList.”

import prophys as prp

# Free tier: 200 points in a PointList, 4 polygon vertices, well under 250.
houses = prp.PointList(house_coords_200x2, site)
zone = prp.Polygon(four_vertices, site)

# Also free: marginalizing over 10,000 Monte-Carlo samples of a tiny model.
compiled = model.compile(n_samples=10_000)

# Requires a license: 5,000 points in a PointList.
many_points = prp.PointList(coords_5000x2, site)

Configuring a license

export PROPHYS_LICENSE=/path/to/license.key

PROPHYS_LICENSE may point to a license file or contain the license text directly. A license is an Ed25519-signed document (licensee, issued, expiry). The verification key is compiled into prophys._core, so a license cannot be forged or the check bypassed from Python.

from prophys import _core as prophys_core

valid, licensee, detail = prophys_core.license_status()
print(valid, licensee, detail)

print(prophys_core.free_tier_limit())  # 250

Exceeding the free tier without a license raises a clear error naming the kernel and the object count involved. Nothing is silently truncated, degraded, or approximated:

INVALID_ARGUMENT: prophys_points_closest_hard_fwd: model size (300 objects)
exceeds the prophys free tier (250 objects). Set PROPHYS_LICENSE to a
valid license to run models of this size.