molcrafts-molrs

Python bindings for the molrs molecular modeling toolkit.

Install with pip install molcrafts-molrs and import molrs. Full docs: https://docs.molcrafts.org/molrs/.

Install

pip install molcrafts-molrs

Requires Python 3.12+.

Quick start

import molrs

# SMILES → atomistic graph (class API under molrs.io)
mol = molrs.io.smiles.read_smiles("CCO").to_atomistic()

# 3D coordinates
from molrs.conformer import Conformer

mol, report = Conformer().generate(mol)

# Force field: typify → pairs → potentials
from molrs.ff.typifier import Mmff94Typifier
from molrs.ff.potential import intramolecular_pairs
from molrs.ff.compile import PotentialCompiler

typifier = Mmff94Typifier()
typed = typifier.typify(mol)
frame = typed.to_frame()
ff = typifier.forcefield()  # a copy of exactly the types typify assigned
frame["pairs"] = intramolecular_pairs(frame, ff)
pots = PotentialCompiler(ff).compile(frame)
energy, forces = pots.evaluate(frame)
assert forces.shape == (frame["atoms"].n_rows, 3)

Package layout

The top level is the subsystems, plus the core data classes a user handles directly (molrs.Frame is molrs.core.Frame: Frame, Block, Trajectory, Box, MolGraph, Atomistic, CoarseGrain, the entity classes, Element, Topology). Everything else has one path, named after its Rust owner.

An analysis is configured when it is built and given its data when it computes: Steinhardt(l=[6]) then .compute(...). The constructor takes configuration only, every optional setting is None unless given (all defaults live in the core; Msd.spec() lists them), and compute takes the data as keyword arguments named after the core input's fields. Msd, Vacf and Rdf are streams, as in the core: acc = Msd(max_lag).accumulator(), acc.push(positions, dt=...) per frame, then acc.finish(). A result is a read-only object that the next compute accepts as is — LinearFit(0.2, 0.8).compute(correlation=msd_result, channel=0).

Kernels take dt in the time unit of your trajectory, and time-valued results come back in that unit. MSD needs unwrapped coordinates (Unwrap produces them). VACF is the unbiased (C(\tau)) used for Green–Kubo D and VDOS.

Development

maturin develop --release
pytest -q