flexible-mof-gateopen | Agent execution contract
Purpose: fast, auditable conditional equilibrium opening pressure P_eq. No web application, MD, DFT or finite-displacement phonons in the packaged route.
closed/open CIF + clean-cif reviews + gas/T + force field + pressure grid
→ source charges/PACMOF + RDKit/sourced guest geometry
→ static MACE + elastic-network entropy proxy + two RASPA isotherms
→ osmotic integration + PΔV → P_eq and sensitivity.
Install
Install the screening library from PyPI:
python -m pip install flexible-mof-gateopen
Core has no third-party dependencies. Install only the extras a route needs, for
example flexible-mof-gateopen[structure,advanced] or
flexible-mof-gateopen[ml]. Every extra is declared in
pyproject.toml.
The workflow scripts and the Codex skill bundle are not shipped in the wheel. For the full workflow, clone the repository and run the commands in the next section from the checkout:
git clone https://github.com/owenxyzhang-seseTJ/flexible-mof-gateopen.git
cd flexible-mof-gateopen
python -m pip install -r requirements.txt
Execute
Run inside the skill directory with a configured Python environment:
python -m pip install -r requirements.txt
python scripts/verify_bundle.py
python scripts/self_test.py
python scripts/run_fast_screen.py --spec /absolute/job.json --output-dir /absolute/new_run
Fill assets/workflow.template.json using agent_contract.md.
Paths in a job are relative to that JSON; bundled model defaults resolve from
the skill root. Override raspa.binary on a different platform. Never run an
unfilled template or fabricate chemistry evidence. The source repository's
curated skill is under skills/flexible-mof-gateopen.
| Need | Script |
|---|---|
| Full workflow | run_fast_screen.py |
| Host ΔE + entropy/free-energy proxy | quick_host_free_energy.py |
| Clean-CIF gate | check_cleaning.py |
| Charge assignment | run_pacmof.py |
| Guest geometry | prepare_guest_rdkit.py |
| GCMC input/run | prepare_raspa_inputs.py |
| Empty-host dilute uptake audit | run_empty_widom.py |
| Bundle/runtime checks | verify_bundle.py |
| Analytic regressions without material simulation | self_test.py |
Mandatory decisions
- Use bundled clean-cif instructions/auditor. Chemistry decisions remain reviewed; do not blindly remove disorder/solvent. Phases need identical composition and activation per formula unit; Z refers to the CIF cell, not the simulation box.
- PACMOF RF-DDEC requires scikit-learn 1.8.0 for the bundled serialized model. Source charges require neutrality, valid atom mapping and provenance.
- RDKit geometry does not provide adsorption LJ/charges. Custom rigid guests need explicit atom-type mapping, molecule JSON and model/geometry provenance.
- VOCs require sourced Psat(T); never exceed it. GCMC and analysis must use the same actual fugacity. Do not extrapolate above simulated pressure.
- A closed zero-uptake scenario is legitimate but explicit, with residual-uptake sensitivity. No accepted swaps is not exclusion; Widom is dilute evidence, not a rigorous finite-pressure upper bound.
Outputs and status
run/workflow_summary.json contains status/failures/analysis; stages retain
immutable attempts, hashes and native outputs. Exit 0 means a conditional screen
was produced; exit 2 means unresolved execution/scientific gates. A completed run
with no opening roots means no predicted equilibrium opening within the sampled
range, not at every pressure. Dense evaluation nodes do not replace GCMC.
ΔF_screen=ΔE_static+ΔU_vib,proxy−TΔS_proxy combines different models.
The entropy proxy has no calibrated MOF free-energy accuracy. Scenario ranges
are not confidence intervals; do not infer hysteresis, barriers or experimental
accuracy. Report calibration and independent validation separately.
Included / external
Included: required project source; MACE-MOF0-v2 and PACMOF RF-DDEC weights; PACMOF source; clean-cif instructions/auditor; RASPA 3.1.0 and MACE runtime source archives; macOS arm64 RASPA executable/share tree; pinned requirements/runtime record; licenses; model provenance; manifest; JSON template; detailed theory PDF and Markdown. No development trial records, training datasets or application UI.
Python packages and OS shared libraries remain dependencies, not an offline OS distribution. The native binary may need the recorded native libraries; build the supplied source elsewhere. No web service is needed after installing dependencies. Review third-party licenses before redistribution. Only load trusted pickle/Torch files. The Theory PDF explains derivations and limitations; its reproducible XeLaTeX source is included for text/formula inspection and rebuilding.
Metadata
Release files for flexible-mof-gateopen 0.4.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| flexible_mof_gateopen-0.4.0.tar.gz | 40.8 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| flexible_mof_gateopen-0.4.0-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 71.7 kB
Release files / flexible_mof_gateopen-0.4.0.tar.gz
| Download URL | flexible_mof_gateopen-0.4.0.tar.gz |
|---|---|
| Size | 40.8 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
06b307417f60c3fa69bd8fc24b358d2df4db325178e7a67d2964f4164b52e478
|
|
BLAKE2b-256 checksum How to use checksums |
db2c95109ff3a982d7ff37f00bfdb29f8dd62effa395e7e341fed732bbb8a88c
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.14.4
|
Release files / flexible_mof_gateopen-0.4.0-py3-none-any.whl
| Download URL | flexible_mof_gateopen-0.4.0-py3-none-any.whl |
|---|---|
| Size | 30.8 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
8c07a1c85d2f096191aee2d5b70dc6a981ede0ec3c8996f7a64e0b4547da7651
|
|
BLAKE2b-256 checksum How to use checksums |
9d096f72319b9da7c122205be34a86525f89c9351be8057fc187faa807c9b630
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.14.4
|