reacnet-md-tools 0.1.5

Helpers around reactive MD post-processing workflows (ReacNetGenerator).

reacnet-md-tools

Small, dependency-light helpers around reactive MD post-processing workflows, built to make ReacNetGenerator runs more repeatable.

Design goals:

• minimal dependencies for the core pipeline
• safe subprocess execution (no shell=True)
• clear, agent-friendly CLIs
• packaged analysis helpers that can be called directly from Agent Skills

Package layout

Reusable code now lives inside the installable package:

• src/reacnet_md_tools/
  • core CLIs: pipeline.py, xs2x.py, infer_atomnames.py
  • analysis/query CLI: query_cli.py
  • reusable analysis modules: rng_tools/
  • lightweight browser backend: webapp/

For automation and skills, prefer the packaged entry points below.

Install / run

Run without installing (recommended)

uvx --from reacnet-md-tools rng-pipeline --help
uvx --from reacnet-md-tools rng-query --help

Install into an environment

python -m pip install -U reacnet-md-tools

Optional analysis extras for richer query/web workflows:

python -m pip install -U 'reacnet-md-tools[analysis]'

Commands

rng-xs2x

Convert a LAMMPS dump with scaled coordinates xs/ys/zs into absolute coordinates x/y/z.

rng-xs2x -i in.lammpstrj -o out.lammpstrj

Current packaged converter supports orthorhombic boxes in the standalone CLI.

rng-infer-atomnames

Infer reacnetgenerator -a atomname order from a LAMMPS data file by parsing the Masses section and matching masses to common elements.

rng-infer-atomnames path/to/system.data
# -> e.g. "C H O Cl"

rng-pipeline

End-to-end helper:

1. If the input is a LAMMPS dump and uses xs/ys/zs, convert it to x/y/z into out/<basename>/dump_with_xyz.lammpstrj.
2. If --atomnames is not provided, try to infer them from a nearby LAMMPS data file (or ask you to pick one with --pick-data).
3. Run ReacNetGenerator, writing everything into out/<basename>/.

Example:

rng-pipeline \
  --input /path/to/traj.lammpstrj \
  --type dump \
  --outroot out \
  --pick-data \
  --stepinterval 10 \
  --maxspecies 50

ReacNetGenerator runner selection

• default --runner auto:
  • use local reacnetgenerator if found on PATH
  • otherwise run via: uvx --from reacnetgenerator reacnetgenerator ...

rng-query

Unified CLI for post-analysis on .reactionabcd / .species outputs.

Typical uses:

rng-query species --reac path/to/run.reactionabcd --formula C6H6
rng-query next --reac path/to/run.reactionabcd --smiles '[H][H]'
rng-query rxn-formula --reac path/to/run.reactionabcd --reactants C6H6 --products C6H5+H

rng-webapp

Launch the lightweight local browser backend for interactive exploration of reaction datasets.

rng-webapp --reac path/to/run.reactionabcd --host 127.0.0.1 --port 8000

For Agent Skills / automation

Prefer these stable commands:

• rng-pipeline for generating outputs from trajectories
• rng-query for species / reaction / pathway / CSV summary queries
• rng-webapp only when the user explicitly wants an interactive local browser UI

Preferred import surface for Python code:

from reacnet_md_tools.rng_tools.network import ReactionNetwork, parse_reactionabcd
from reacnet_md_tools.rng_tools.reaction import parse_species_file

Development

Build (wheel + sdist):

python -m pip install -U build
python -m build

Run tests:

python -m pip install -U '.[dev]'
pytest

Publishing

This package is intended to be published from GitHub Actions via PyPI Trusted Publishing.

Typical release flow:

1. Push commits to GitHub
2. Create a version tag such as v0.1.5
3. Let the GitHub Actions publish workflow build and upload to PyPI
4. Install/run with uvx --from reacnet-md-tools ...

License

MIT (see LICENSE).