mlmm-toolkit 0.3.1__tar.gz → 0.3.4__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (945) hide show
  1. mlmm_toolkit-0.3.4/CHANGELOG.md +1182 -0
  2. mlmm_toolkit-0.3.4/CITATION.cff +32 -0
  3. mlmm_toolkit-0.3.4/MANIFEST.in +14 -0
  4. mlmm_toolkit-0.3.4/PKG-INFO +281 -0
  5. mlmm_toolkit-0.3.4/README.md +204 -0
  6. mlmm_toolkit-0.3.4/THIRD_PARTY_NOTICES.txt +70 -0
  7. mlmm_toolkit-0.3.4/docs/add-elem-info.md +85 -0
  8. mlmm_toolkit-0.3.4/docs/all.md +361 -0
  9. mlmm_toolkit-0.3.4/docs/architecture.md +399 -0
  10. mlmm_toolkit-0.3.4/docs/backends.md +204 -0
  11. mlmm_toolkit-0.3.4/docs/bond-summary.md +67 -0
  12. mlmm_toolkit-0.3.4/docs/cif.md +93 -0
  13. mlmm_toolkit-0.3.4/docs/cli-conventions.md +248 -0
  14. mlmm_toolkit-0.3.4/docs/concepts.md +388 -0
  15. mlmm_toolkit-0.3.4/docs/define-layer.md +75 -0
  16. mlmm_toolkit-0.3.4/docs/device-hpc.md +222 -0
  17. mlmm_toolkit-0.3.4/docs/dft.md +143 -0
  18. mlmm_toolkit-0.3.4/docs/energy-diagram.md +62 -0
  19. mlmm_toolkit-0.3.4/docs/extract.md +240 -0
  20. mlmm_toolkit-0.3.4/docs/fix-altloc.md +124 -0
  21. mlmm_toolkit-0.3.4/docs/freq.md +200 -0
  22. mlmm_toolkit-0.3.4/docs/getting-started.md +244 -0
  23. mlmm_toolkit-0.3.4/docs/glossary.md +146 -0
  24. mlmm_toolkit-0.3.4/docs/index.md +164 -0
  25. mlmm_toolkit-0.3.4/docs/irc.md +193 -0
  26. mlmm_toolkit-0.3.4/docs/ja/add-elem-info.md +77 -0
  27. mlmm_toolkit-0.3.4/docs/ja/all.md +373 -0
  28. mlmm_toolkit-0.3.4/docs/ja/architecture.md +386 -0
  29. mlmm_toolkit-0.3.4/docs/ja/backends.md +194 -0
  30. mlmm_toolkit-0.3.4/docs/ja/cif.md +66 -0
  31. mlmm_toolkit-0.3.4/docs/ja/cli-conventions.md +382 -0
  32. mlmm_toolkit-0.3.4/docs/ja/concepts.md +369 -0
  33. mlmm_toolkit-0.3.4/docs/ja/define-layer.md +80 -0
  34. mlmm_toolkit-0.3.4/docs/ja/device-hpc.md +230 -0
  35. mlmm_toolkit-0.3.4/docs/ja/dft.md +145 -0
  36. mlmm_toolkit-0.3.4/docs/ja/energy-diagram.md +62 -0
  37. mlmm_toolkit-0.3.4/docs/ja/extract.md +246 -0
  38. mlmm_toolkit-0.3.4/docs/ja/fix-altloc.md +134 -0
  39. mlmm_toolkit-0.3.4/docs/ja/freq.md +202 -0
  40. mlmm_toolkit-0.3.4/docs/ja/getting-started.md +558 -0
  41. mlmm_toolkit-0.3.4/docs/ja/glossary.md +150 -0
  42. mlmm_toolkit-0.3.4/docs/ja/index.md +303 -0
  43. mlmm_toolkit-0.3.4/docs/ja/irc.md +234 -0
  44. mlmm_toolkit-0.3.4/docs/ja/json-output.md +350 -0
  45. mlmm_toolkit-0.3.4/docs/ja/mcp_server.md +140 -0
  46. mlmm_toolkit-0.3.4/docs/ja/mlmm-calc.md +126 -0
  47. mlmm_toolkit-0.3.4/docs/ja/mm-parm.md +94 -0
  48. mlmm_toolkit-0.3.4/docs/ja/oniom-export.md +86 -0
  49. mlmm_toolkit-0.3.4/docs/ja/oniom-gaussian.md +70 -0
  50. mlmm_toolkit-0.3.4/docs/ja/oniom-import.md +79 -0
  51. mlmm_toolkit-0.3.4/docs/ja/oniom-orca.md +72 -0
  52. mlmm_toolkit-0.3.4/docs/ja/opt.md +304 -0
  53. mlmm_toolkit-0.3.4/docs/ja/output-layout.md +96 -0
  54. mlmm_toolkit-0.3.4/docs/ja/path-opt.md +127 -0
  55. mlmm_toolkit-0.3.4/docs/ja/path-search.md +152 -0
  56. mlmm_toolkit-0.3.4/docs/ja/python-api.md +157 -0
  57. mlmm_toolkit-0.3.4/docs/ja/quickstart-all.md +36 -0
  58. mlmm_toolkit-0.3.4/docs/ja/quickstart-scan-spec.md +66 -0
  59. mlmm_toolkit-0.3.4/docs/ja/quickstart-tsopt-freq.md +49 -0
  60. mlmm_toolkit-0.3.4/docs/ja/recipes-common-errors.md +91 -0
  61. mlmm_toolkit-0.3.4/docs/ja/reproducibility.md +48 -0
  62. mlmm_toolkit-0.3.4/docs/ja/scan.md +269 -0
  63. mlmm_toolkit-0.3.4/docs/ja/scan2d.md +212 -0
  64. mlmm_toolkit-0.3.4/docs/ja/scan3d.md +199 -0
  65. mlmm_toolkit-0.3.4/docs/ja/sp.md +94 -0
  66. mlmm_toolkit-0.3.4/docs/ja/trj2fig.md +87 -0
  67. mlmm_toolkit-0.3.4/docs/ja/troubleshooting.md +575 -0
  68. mlmm_toolkit-0.3.4/docs/ja/tsopt.md +377 -0
  69. mlmm_toolkit-0.3.4/docs/ja/yaml-reference.md +723 -0
  70. mlmm_toolkit-0.3.4/docs/json-output.md +412 -0
  71. mlmm_toolkit-0.3.4/docs/mcp_server.md +162 -0
  72. mlmm_toolkit-0.3.4/docs/mlmm-calc.md +115 -0
  73. mlmm_toolkit-0.3.4/docs/mm-parm.md +103 -0
  74. mlmm_toolkit-0.3.4/docs/oniom-export.md +88 -0
  75. mlmm_toolkit-0.3.4/docs/oniom-gaussian.md +73 -0
  76. mlmm_toolkit-0.3.4/docs/oniom-import.md +78 -0
  77. mlmm_toolkit-0.3.4/docs/oniom-orca.md +76 -0
  78. mlmm_toolkit-0.3.4/docs/opt.md +168 -0
  79. mlmm_toolkit-0.3.4/docs/output-layout.md +92 -0
  80. mlmm_toolkit-0.3.4/docs/path-opt.md +128 -0
  81. mlmm_toolkit-0.3.4/docs/path-search.md +139 -0
  82. mlmm_toolkit-0.3.4/docs/python-api.md +159 -0
  83. mlmm_toolkit-0.3.4/docs/quickstart-all.md +40 -0
  84. mlmm_toolkit-0.3.4/docs/quickstart-scan-spec.md +68 -0
  85. mlmm_toolkit-0.3.4/docs/quickstart-tsopt-freq.md +50 -0
  86. mlmm_toolkit-0.3.4/docs/recipes-common-errors.md +98 -0
  87. mlmm_toolkit-0.3.4/docs/reference/commands/add_elem_info.md +25 -0
  88. mlmm_toolkit-0.3.4/docs/reference/commands/all.md +383 -0
  89. mlmm_toolkit-0.3.4/docs/reference/commands/define_layer.md +30 -0
  90. mlmm_toolkit-0.3.4/docs/reference/commands/dft.md +109 -0
  91. mlmm_toolkit-0.3.4/docs/reference/commands/energy_diagram.md +29 -0
  92. mlmm_toolkit-0.3.4/docs/reference/commands/extract.md +68 -0
  93. mlmm_toolkit-0.3.4/docs/reference/commands/fix_altloc.md +31 -0
  94. mlmm_toolkit-0.3.4/docs/reference/commands/freq.md +174 -0
  95. mlmm_toolkit-0.3.4/docs/reference/commands/irc.md +176 -0
  96. mlmm_toolkit-0.3.4/docs/reference/commands/mm_parm.md +48 -0
  97. mlmm_toolkit-0.3.4/docs/reference/commands/oniom_export.md +56 -0
  98. mlmm_toolkit-0.3.4/docs/reference/commands/oniom_import.md +33 -0
  99. mlmm_toolkit-0.3.4/docs/reference/commands/opt.md +204 -0
  100. mlmm_toolkit-0.3.4/docs/reference/commands/path_opt.md +177 -0
  101. mlmm_toolkit-0.3.4/docs/reference/commands/path_search.md +183 -0
  102. mlmm_toolkit-0.3.4/docs/reference/commands/scan.md +166 -0
  103. mlmm_toolkit-0.3.4/docs/reference/commands/scan2d.md +159 -0
  104. mlmm_toolkit-0.3.4/docs/reference/commands/scan3d.md +163 -0
  105. mlmm_toolkit-0.3.4/docs/reference/commands/sp.md +133 -0
  106. mlmm_toolkit-0.3.4/docs/reference/commands/trj2fig.md +43 -0
  107. mlmm_toolkit-0.3.4/docs/reference/commands/tsopt.md +226 -0
  108. mlmm_toolkit-0.3.4/docs/reference/yaml.md +95 -0
  109. mlmm_toolkit-0.3.4/docs/reproducibility.md +79 -0
  110. mlmm_toolkit-0.3.4/docs/scan.md +264 -0
  111. mlmm_toolkit-0.3.4/docs/scan2d.md +215 -0
  112. mlmm_toolkit-0.3.4/docs/scan3d.md +216 -0
  113. mlmm_toolkit-0.3.4/docs/sp.md +92 -0
  114. mlmm_toolkit-0.3.4/docs/trj2fig.md +90 -0
  115. mlmm_toolkit-0.3.4/docs/troubleshooting.md +189 -0
  116. mlmm_toolkit-0.3.4/docs/tsopt.md +323 -0
  117. mlmm_toolkit-0.3.4/docs/yaml-reference.md +760 -0
  118. mlmm_toolkit-0.3.4/examples/beza/1.R.pdb +9221 -0
  119. mlmm_toolkit-0.3.4/examples/beza/2.IM.pdb +9221 -0
  120. mlmm_toolkit-0.3.4/examples/beza/3.P.pdb +9217 -0
  121. mlmm_toolkit-0.3.4/examples/beza/README.md +30 -0
  122. mlmm_toolkit-0.3.4/examples/beza/run.sh +19 -0
  123. mlmm_toolkit-0.3.4/examples/comt/1.R.pdb +3427 -0
  124. mlmm_toolkit-0.3.4/examples/comt/3.P.pdb +3427 -0
  125. mlmm_toolkit-0.3.4/examples/comt/README.md +18 -0
  126. mlmm_toolkit-0.3.4/examples/comt/run.sh +13 -0
  127. mlmm_toolkit-0.3.4/examples/methyltransferase/run_all.sh +15 -0
  128. mlmm_toolkit-0.3.4/examples/methyltransferase/run_stepwise.sh +63 -0
  129. mlmm_toolkit-0.3.4/examples/mlmm_colab.ipynb +12725 -0
  130. mlmm_toolkit-0.3.4/examples/toy_system/ml_region_p.pdb +48 -0
  131. mlmm_toolkit-0.3.4/examples/toy_system/opt_ase.py +22 -0
  132. mlmm_toolkit-0.3.4/examples/toy_system/p_toy.pdb +125 -0
  133. mlmm_toolkit-0.3.4/examples/toy_system/run.sh +107 -0
  134. mlmm_toolkit-0.3.4/examples/toy_system/test_core.py +25 -0
  135. mlmm_toolkit-0.3.4/hessian_ff/active_space.py +94 -0
  136. mlmm_toolkit-0.3.4/hessian_ff/analytical_hessian.py +613 -0
  137. mlmm_toolkit-0.3.4/hessian_ff/forcefield.py +361 -0
  138. mlmm_toolkit-0.3.4/hessian_ff/loaders.py +674 -0
  139. mlmm_toolkit-0.3.4/hessian_ff/native/Makefile +9 -0
  140. mlmm_toolkit-0.3.4/hessian_ff/native/bonded_ext.cpp +713 -0
  141. mlmm_toolkit-0.3.4/hessian_ff/native/loader.py +742 -0
  142. mlmm_toolkit-0.3.4/hessian_ff/native/nonbonded_ext.cpp +1145 -0
  143. mlmm_toolkit-0.3.4/hessian_ff/terms/angle.py +75 -0
  144. mlmm_toolkit-0.3.4/hessian_ff/terms/dihedral.py +132 -0
  145. mlmm_toolkit-0.3.4/hessian_ff/terms/nonbonded.py +276 -0
  146. mlmm_toolkit-0.3.4/hessian_ff/tests/test_active_space.py +69 -0
  147. mlmm_toolkit-0.3.4/hessian_ff/tests/test_energy_force.py +151 -0
  148. mlmm_toolkit-0.3.4/hessian_ff/tests/test_hessian.py +136 -0
  149. mlmm_toolkit-0.3.4/hessian_ff/tests/test_input_boundaries.py +177 -0
  150. mlmm_toolkit-0.3.4/hessian_ff/tests/test_native_loader.py +421 -0
  151. mlmm_toolkit-0.3.4/hessian_ff/tests/test_runtime_cache.py +104 -0
  152. mlmm_toolkit-0.3.4/hessian_ff/tests/test_validation.py +91 -0
  153. mlmm_toolkit-0.3.4/hessian_ff/workflows.py +910 -0
  154. mlmm_toolkit-0.3.4/mlmm/__init__.py +38 -0
  155. mlmm_toolkit-0.3.4/mlmm/_version.py +23 -0
  156. mlmm_toolkit-0.3.4/mlmm/backends/__init__.py +252 -0
  157. mlmm_toolkit-0.3.4/mlmm/backends/_determinism.py +124 -0
  158. mlmm_toolkit-0.3.4/mlmm/backends/custom.py +158 -0
  159. mlmm_toolkit-0.3.4/mlmm/backends/methods.py +65 -0
  160. mlmm_toolkit-0.3.4/mlmm/backends/mlmm_calc.py +3711 -0
  161. mlmm_toolkit-0.3.4/mlmm/backends/xtb_embedcharge_correction.py +482 -0
  162. mlmm_toolkit-0.3.4/mlmm/cli/app.py +1010 -0
  163. mlmm_toolkit-0.3.4/mlmm/cli/common_options.py +389 -0
  164. mlmm_toolkit-0.3.4/mlmm/cli/decorators.py +218 -0
  165. mlmm_toolkit-0.3.4/mlmm/cli/default_group.py +614 -0
  166. mlmm_toolkit-0.3.4/mlmm/cli/help_pages.py +93 -0
  167. mlmm_toolkit-0.3.4/mlmm/cli/preflight.py +37 -0
  168. mlmm_toolkit-0.3.4/mlmm/core/__init__.py +19 -0
  169. mlmm_toolkit-0.3.4/mlmm/core/calc_eval.py +53 -0
  170. mlmm_toolkit-0.3.4/mlmm/core/defaults.py +530 -0
  171. mlmm_toolkit-0.3.4/mlmm/core/logging.py +41 -0
  172. mlmm_toolkit-0.3.4/mlmm/core/output.py +76 -0
  173. mlmm_toolkit-0.3.4/mlmm/core/pes_composition.py +248 -0
  174. mlmm_toolkit-0.3.4/mlmm/core/residue_data.py +167 -0
  175. mlmm_toolkit-0.3.4/mlmm/core/result_commit.py +284 -0
  176. mlmm_toolkit-0.3.4/mlmm/core/utils.py +3849 -0
  177. mlmm_toolkit-0.3.4/mlmm/domain/add_elem_info.py +426 -0
  178. mlmm_toolkit-0.3.4/mlmm/domain/bond_changes.py +264 -0
  179. mlmm_toolkit-0.3.4/mlmm/domain/bond_summary.py +204 -0
  180. mlmm_toolkit-0.3.4/mlmm/io/__init__.py +15 -0
  181. mlmm_toolkit-0.3.4/mlmm/io/altloc.py +78 -0
  182. mlmm_toolkit-0.3.4/mlmm/io/energy_diagram.py +323 -0
  183. mlmm_toolkit-0.3.4/mlmm/io/gaussian_input.py +133 -0
  184. mlmm_toolkit-0.3.4/mlmm/io/hessian_cache.py +695 -0
  185. mlmm_toolkit-0.3.4/mlmm/io/hessian_calc.py +150 -0
  186. mlmm_toolkit-0.3.4/mlmm/io/hessian_file.py +310 -0
  187. mlmm_toolkit-0.3.4/mlmm/io/oniom_identity.py +77 -0
  188. mlmm_toolkit-0.3.4/mlmm/io/path_mode_cache.py +381 -0
  189. mlmm_toolkit-0.3.4/mlmm/io/pdb_fix.py +484 -0
  190. mlmm_toolkit-0.3.4/mlmm/io/pdb_indexing.py +346 -0
  191. mlmm_toolkit-0.3.4/mlmm/io/plotly_image.py +144 -0
  192. mlmm_toolkit-0.3.4/mlmm/io/structure_formats.py +1209 -0
  193. mlmm_toolkit-0.3.4/mlmm/io/summary.py +1577 -0
  194. mlmm_toolkit-0.3.4/mlmm/io/trj2fig.py +537 -0
  195. mlmm_toolkit-0.3.4/mlmm/io/xyz_trajectory.py +290 -0
  196. mlmm_toolkit-0.3.4/mlmm/mcp/__init__.py +33 -0
  197. mlmm_toolkit-0.3.4/mlmm/mcp/_runner.py +439 -0
  198. mlmm_toolkit-0.3.4/mlmm/mcp/_tools.py +1048 -0
  199. mlmm_toolkit-0.3.4/mlmm/workflows/_all_helpers.py +622 -0
  200. mlmm_toolkit-0.3.4/mlmm/workflows/_microiteration.py +581 -0
  201. mlmm_toolkit-0.3.4/mlmm/workflows/_opt_freq_common.py +90 -0
  202. mlmm_toolkit-0.3.4/mlmm/workflows/_outcomes.py +559 -0
  203. mlmm_toolkit-0.3.4/mlmm/workflows/_run_session.py +593 -0
  204. mlmm_toolkit-0.3.4/mlmm/workflows/align_freeze.py +620 -0
  205. mlmm_toolkit-0.3.4/mlmm/workflows/all.py +8156 -0
  206. mlmm_toolkit-0.3.4/mlmm/workflows/charge_prep.py +275 -0
  207. mlmm_toolkit-0.3.4/mlmm/workflows/define_layer.py +676 -0
  208. mlmm_toolkit-0.3.4/mlmm/workflows/dft.py +1442 -0
  209. mlmm_toolkit-0.3.4/mlmm/workflows/extract.py +2216 -0
  210. mlmm_toolkit-0.3.4/mlmm/workflows/freq.py +1897 -0
  211. mlmm_toolkit-0.3.4/mlmm/workflows/irc.py +1346 -0
  212. mlmm_toolkit-0.3.4/mlmm/workflows/mm_parm.py +1241 -0
  213. mlmm_toolkit-0.3.4/mlmm/workflows/oniom_export.py +2105 -0
  214. mlmm_toolkit-0.3.4/mlmm/workflows/oniom_import.py +586 -0
  215. mlmm_toolkit-0.3.4/mlmm/workflows/opt.py +2474 -0
  216. mlmm_toolkit-0.3.4/mlmm/workflows/path_opt.py +2023 -0
  217. mlmm_toolkit-0.3.4/mlmm/workflows/path_search.py +2897 -0
  218. mlmm_toolkit-0.3.4/mlmm/workflows/restraints.py +303 -0
  219. mlmm_toolkit-0.3.4/mlmm/workflows/scan.py +1368 -0
  220. mlmm_toolkit-0.3.4/mlmm/workflows/scan2d.py +1587 -0
  221. mlmm_toolkit-0.3.4/mlmm/workflows/scan3d.py +1660 -0
  222. mlmm_toolkit-0.3.4/mlmm/workflows/scan_common.py +301 -0
  223. mlmm_toolkit-0.3.4/mlmm/workflows/sp.py +619 -0
  224. mlmm_toolkit-0.3.4/mlmm/workflows/tsopt.py +5577 -0
  225. mlmm_toolkit-0.3.4/mlmm_toolkit.egg-info/PKG-INFO +281 -0
  226. mlmm_toolkit-0.3.4/mlmm_toolkit.egg-info/SOURCES.txt +577 -0
  227. mlmm_toolkit-0.3.4/mlmm_toolkit.egg-info/entry_points.txt +3 -0
  228. mlmm_toolkit-0.3.4/mlmm_toolkit.egg-info/requires.txt +60 -0
  229. mlmm_toolkit-0.3.4/pyproject.toml +133 -0
  230. mlmm_toolkit-0.3.4/pysisyphus/Geometry.py +1934 -0
  231. mlmm_toolkit-0.3.4/pysisyphus/_array.py +114 -0
  232. mlmm_toolkit-0.3.4/pysisyphus/calculators/Calculator.py +844 -0
  233. mlmm_toolkit-0.3.4/pysisyphus/calculators/Dimer.py +836 -0
  234. mlmm_toolkit-0.3.4/pysisyphus/config.py +131 -0
  235. mlmm_toolkit-0.3.4/pysisyphus/cos/ChainOfStates.py +750 -0
  236. mlmm_toolkit-0.3.4/pysisyphus/cos/GrowingChainOfStates.py +74 -0
  237. mlmm_toolkit-0.3.4/pysisyphus/cos/GrowingString.py +541 -0
  238. mlmm_toolkit-0.3.4/pysisyphus/cos/__init__.py +19 -0
  239. mlmm_toolkit-0.3.4/pysisyphus/helpers.py +630 -0
  240. mlmm_toolkit-0.3.4/pysisyphus/helpers_pure.py +666 -0
  241. mlmm_toolkit-0.3.4/pysisyphus/intcoords/BondedFragment.py +48 -0
  242. mlmm_toolkit-0.3.4/pysisyphus/intcoords/CartesianCoords.py +140 -0
  243. mlmm_toolkit-0.3.4/pysisyphus/intcoords/DLC.py +195 -0
  244. mlmm_toolkit-0.3.4/pysisyphus/intcoords/DistanceFunction.py +41 -0
  245. mlmm_toolkit-0.3.4/pysisyphus/intcoords/DummyImproper.py +101 -0
  246. mlmm_toolkit-0.3.4/pysisyphus/intcoords/DummyTorsion.py +110 -0
  247. mlmm_toolkit-0.3.4/pysisyphus/intcoords/LinearDisplacement.py +152 -0
  248. mlmm_toolkit-0.3.4/pysisyphus/intcoords/PrimTypes.py +302 -0
  249. mlmm_toolkit-0.3.4/pysisyphus/intcoords/RedundantCoords.py +662 -0
  250. mlmm_toolkit-0.3.4/pysisyphus/intcoords/Torsion.py +100 -0
  251. mlmm_toolkit-0.3.4/pysisyphus/intcoords/augment_bonds.py +133 -0
  252. mlmm_toolkit-0.3.4/pysisyphus/intcoords/setup.py +973 -0
  253. mlmm_toolkit-0.3.4/pysisyphus/intcoords/setup_fast.py +182 -0
  254. mlmm_toolkit-0.3.4/pysisyphus/intcoords/update.py +267 -0
  255. mlmm_toolkit-0.3.4/pysisyphus/io/hessian.py +34 -0
  256. mlmm_toolkit-0.3.4/pysisyphus/io/pdb.py +313 -0
  257. mlmm_toolkit-0.3.4/pysisyphus/io/qcschema.py +40 -0
  258. mlmm_toolkit-0.3.4/pysisyphus/irc/DWI.py +121 -0
  259. mlmm_toolkit-0.3.4/pysisyphus/irc/EulerPC.py +407 -0
  260. mlmm_toolkit-0.3.4/pysisyphus/irc/IRC.py +1281 -0
  261. mlmm_toolkit-0.3.4/pysisyphus/irc/Instanton.py +304 -0
  262. mlmm_toolkit-0.3.4/pysisyphus/linalg.py +411 -0
  263. mlmm_toolkit-0.3.4/pysisyphus/modefollow/lanczos.py +109 -0
  264. mlmm_toolkit-0.3.4/pysisyphus/normal_modes.py +450 -0
  265. mlmm_toolkit-0.3.4/pysisyphus/optimizers/HessianOptimizer.py +1650 -0
  266. mlmm_toolkit-0.3.4/pysisyphus/optimizers/LBFGS.py +343 -0
  267. mlmm_toolkit-0.3.4/pysisyphus/optimizers/Optimizer.py +1353 -0
  268. mlmm_toolkit-0.3.4/pysisyphus/optimizers/RFOptimizer.py +266 -0
  269. mlmm_toolkit-0.3.4/pysisyphus/optimizers/StringOptimizer.py +179 -0
  270. mlmm_toolkit-0.3.4/pysisyphus/optimizers/checkpoint.py +494 -0
  271. mlmm_toolkit-0.3.4/pysisyphus/optimizers/closures.py +303 -0
  272. mlmm_toolkit-0.3.4/pysisyphus/optimizers/gdiis.py +282 -0
  273. mlmm_toolkit-0.3.4/pysisyphus/optimizers/hessian_updates.py +540 -0
  274. mlmm_toolkit-0.3.4/pysisyphus/optimizers/poly_fit.py +296 -0
  275. mlmm_toolkit-0.3.4/pysisyphus/optimizers/restrict_step.py +145 -0
  276. mlmm_toolkit-0.3.4/pysisyphus/stocastic/Pipeline.py +303 -0
  277. mlmm_toolkit-0.3.4/pysisyphus/tr_projection.py +292 -0
  278. mlmm_toolkit-0.3.4/pysisyphus/tsoptimizers/RSIRFOptimizer.py +85 -0
  279. mlmm_toolkit-0.3.4/pysisyphus/tsoptimizers/RSPRFOptimizer.py +211 -0
  280. mlmm_toolkit-0.3.4/pysisyphus/tsoptimizers/TRIM.py +112 -0
  281. mlmm_toolkit-0.3.4/pysisyphus/tsoptimizers/TSHessianOptimizer.py +1826 -0
  282. mlmm_toolkit-0.3.4/pysisyphus/xyzloader.py +176 -0
  283. mlmm_toolkit-0.3.4/skills/README.md +50 -0
  284. mlmm_toolkit-0.3.4/skills/colab-local-gpu-runtime/SKILL.md +45 -0
  285. mlmm_toolkit-0.3.4/skills/colab-local-gpu-runtime/agents/openai.yaml +4 -0
  286. mlmm_toolkit-0.3.4/skills/colab-local-gpu-runtime/references/colab-connection.md +27 -0
  287. mlmm_toolkit-0.3.4/skills/colab-local-gpu-runtime/scripts/setup.ps1 +82 -0
  288. mlmm_toolkit-0.3.4/skills/colab-local-gpu-runtime/scripts/start.ps1 +58 -0
  289. mlmm_toolkit-0.3.4/skills/colab-local-gpu-runtime/scripts/status.ps1 +20 -0
  290. mlmm_toolkit-0.3.4/skills/colab-local-gpu-runtime/scripts/stop.ps1 +10 -0
  291. mlmm_toolkit-0.3.4/skills/mlmm-architecture/SKILL.md +79 -0
  292. mlmm_toolkit-0.3.4/skills/mlmm-cli/SKILL.md +161 -0
  293. mlmm_toolkit-0.3.4/skills/mlmm-cli/add-elem-info.md +64 -0
  294. mlmm_toolkit-0.3.4/skills/mlmm-cli/all-endpoint-mep.md +126 -0
  295. mlmm_toolkit-0.3.4/skills/mlmm-cli/all-scan-list.md +151 -0
  296. mlmm_toolkit-0.3.4/skills/mlmm-cli/all-ts-only.md +200 -0
  297. mlmm_toolkit-0.3.4/skills/mlmm-cli/all.md +199 -0
  298. mlmm_toolkit-0.3.4/skills/mlmm-cli/bond-summary.md +77 -0
  299. mlmm_toolkit-0.3.4/skills/mlmm-cli/define-layer.md +107 -0
  300. mlmm_toolkit-0.3.4/skills/mlmm-cli/dft.md +147 -0
  301. mlmm_toolkit-0.3.4/skills/mlmm-cli/extract.md +118 -0
  302. mlmm_toolkit-0.3.4/skills/mlmm-cli/fix-altloc.md +65 -0
  303. mlmm_toolkit-0.3.4/skills/mlmm-cli/freq.md +168 -0
  304. mlmm_toolkit-0.3.4/skills/mlmm-cli/irc.md +177 -0
  305. mlmm_toolkit-0.3.4/skills/mlmm-cli/mm-parm.md +144 -0
  306. mlmm_toolkit-0.3.4/skills/mlmm-cli/oniom-export.md +110 -0
  307. mlmm_toolkit-0.3.4/skills/mlmm-cli/oniom-import.md +91 -0
  308. mlmm_toolkit-0.3.4/skills/mlmm-cli/opt.md +124 -0
  309. mlmm_toolkit-0.3.4/skills/mlmm-cli/path-opt.md +103 -0
  310. mlmm_toolkit-0.3.4/skills/mlmm-cli/path-search.md +132 -0
  311. mlmm_toolkit-0.3.4/skills/mlmm-cli/scan.md +105 -0
  312. mlmm_toolkit-0.3.4/skills/mlmm-cli/scan2d.md +89 -0
  313. mlmm_toolkit-0.3.4/skills/mlmm-cli/scan3d.md +74 -0
  314. mlmm_toolkit-0.3.4/skills/mlmm-cli/sp.md +103 -0
  315. mlmm_toolkit-0.3.4/skills/mlmm-cli/trj2fig.md +78 -0
  316. mlmm_toolkit-0.3.4/skills/mlmm-cli/tsopt.md +183 -0
  317. mlmm_toolkit-0.3.4/skills/mlmm-env-detect/SKILL.md +178 -0
  318. mlmm_toolkit-0.3.4/skills/mlmm-hpc/SKILL.md +230 -0
  319. mlmm_toolkit-0.3.4/skills/mlmm-hpc/dynamic-dispatch.md +194 -0
  320. mlmm_toolkit-0.3.4/skills/mlmm-install-backends/SKILL.md +174 -0
  321. mlmm_toolkit-0.3.4/skills/mlmm-install-backends/aimnet2.md +78 -0
  322. mlmm_toolkit-0.3.4/skills/mlmm-install-backends/core.md +120 -0
  323. mlmm_toolkit-0.3.4/skills/mlmm-install-backends/dft.md +93 -0
  324. mlmm_toolkit-0.3.4/skills/mlmm-install-backends/env-cuda.md +143 -0
  325. mlmm_toolkit-0.3.4/skills/mlmm-install-backends/mace.md +95 -0
  326. mlmm_toolkit-0.3.4/skills/mlmm-install-backends/orb.md +88 -0
  327. mlmm_toolkit-0.3.4/skills/mlmm-install-backends/uma.md +112 -0
  328. mlmm_toolkit-0.3.4/skills/mlmm-install-backends/xtb.md +24 -0
  329. mlmm_toolkit-0.3.4/skills/mlmm-mcp/SKILL.md +128 -0
  330. mlmm_toolkit-0.3.4/skills/mlmm-overview/SKILL.md +152 -0
  331. mlmm_toolkit-0.3.4/skills/mlmm-structure-io/SKILL.md +146 -0
  332. mlmm_toolkit-0.3.4/skills/mlmm-structure-io/charge-multiplicity.md +195 -0
  333. mlmm_toolkit-0.3.4/skills/mlmm-structure-io/cif.md +37 -0
  334. mlmm_toolkit-0.3.4/skills/mlmm-structure-io/gjf.md +87 -0
  335. mlmm_toolkit-0.3.4/skills/mlmm-structure-io/parm7.md +134 -0
  336. mlmm_toolkit-0.3.4/skills/mlmm-structure-io/pdb.md +193 -0
  337. mlmm_toolkit-0.3.4/skills/mlmm-ts-strategy/SKILL.md +145 -0
  338. mlmm_toolkit-0.3.4/skills/mlmm-workflows-output/SKILL.md +432 -0
  339. mlmm_toolkit-0.3.4/tests/README.md +40 -0
  340. mlmm_toolkit-0.3.4/tests/conftest.py +24 -0
  341. mlmm_toolkit-0.3.4/tests/test_active_map_invalidation.py +84 -0
  342. mlmm_toolkit-0.3.4/tests/test_add_elem_info_output.py +94 -0
  343. mlmm_toolkit-0.3.4/tests/test_advanced_help_dedup.py +149 -0
  344. mlmm_toolkit-0.3.4/tests/test_align_log.py +124 -0
  345. mlmm_toolkit-0.3.4/tests/test_all_calculator_template.py +266 -0
  346. mlmm_toolkit-0.3.4/tests/test_all_dry_run_semantics.py +145 -0
  347. mlmm_toolkit-0.3.4/tests/test_all_helpers.py +1109 -0
  348. mlmm_toolkit-0.3.4/tests/test_all_layer_honor.py +109 -0
  349. mlmm_toolkit-0.3.4/tests/test_all_ml_region_subset.py +226 -0
  350. mlmm_toolkit-0.3.4/tests/test_all_multiplicity_option.py +57 -0
  351. mlmm_toolkit-0.3.4/tests/test_all_path_tangent.py +173 -0
  352. mlmm_toolkit-0.3.4/tests/test_all_postprocessing_contract.py +135 -0
  353. mlmm_toolkit-0.3.4/tests/test_all_run_ownership.py +329 -0
  354. mlmm_toolkit-0.3.4/tests/test_all_scan_contract_checker.py +54 -0
  355. mlmm_toolkit-0.3.4/tests/test_all_summary_provenance.py +272 -0
  356. mlmm_toolkit-0.3.4/tests/test_array_shim_equivalence.py +157 -0
  357. mlmm_toolkit-0.3.4/tests/test_backend_model_load_announcement.py +110 -0
  358. mlmm_toolkit-0.3.4/tests/test_beza_example_contract.py +90 -0
  359. mlmm_toolkit-0.3.4/tests/test_bond_changes.py +149 -0
  360. mlmm_toolkit-0.3.4/tests/test_bool_compat_cli.py +119 -0
  361. mlmm_toolkit-0.3.4/tests/test_bundled_pysis_policy.py +163 -0
  362. mlmm_toolkit-0.3.4/tests/test_bundled_pysisyphus_regressions.py +1530 -0
  363. mlmm_toolkit-0.3.4/tests/test_calc_eval.py +65 -0
  364. mlmm_toolkit-0.3.4/tests/test_charge_prep_exact_model.py +276 -0
  365. mlmm_toolkit-0.3.4/tests/test_cli_feature_regressions.py +1243 -0
  366. mlmm_toolkit-0.3.4/tests/test_cli_help.py +303 -0
  367. mlmm_toolkit-0.3.4/tests/test_cli_utils.py +36 -0
  368. mlmm_toolkit-0.3.4/tests/test_colab_notebook.py +6776 -0
  369. mlmm_toolkit-0.3.4/tests/test_confirmed_runtime_boundaries.py +69 -0
  370. mlmm_toolkit-0.3.4/tests/test_custom_calc_file.py +268 -0
  371. mlmm_toolkit-0.3.4/tests/test_cycle_limit_defaults.py +43 -0
  372. mlmm_toolkit-0.3.4/tests/test_default_group.py +274 -0
  373. mlmm_toolkit-0.3.4/tests/test_defaults.py +169 -0
  374. mlmm_toolkit-0.3.4/tests/test_dependency_metadata.py +23 -0
  375. mlmm_toolkit-0.3.4/tests/test_determinism.py +55 -0
  376. mlmm_toolkit-0.3.4/tests/test_dft_model_parm7_lj_tables.py +325 -0
  377. mlmm_toolkit-0.3.4/tests/test_dft_output_boundary.py +231 -0
  378. mlmm_toolkit-0.3.4/tests/test_dft_result_payload.py +163 -0
  379. mlmm_toolkit-0.3.4/tests/test_dimer_constraints.py +380 -0
  380. mlmm_toolkit-0.3.4/tests/test_dmf_result_contract.py +252 -0
  381. mlmm_toolkit-0.3.4/tests/test_docs_command_contract.py +188 -0
  382. mlmm_toolkit-0.3.4/tests/test_docs_command_smoke.py +46 -0
  383. mlmm_toolkit-0.3.4/tests/test_docs_deploy_workflow.py +32 -0
  384. mlmm_toolkit-0.3.4/tests/test_docs_link_contract.py +57 -0
  385. mlmm_toolkit-0.3.4/tests/test_dry_run_footer_contract.py +77 -0
  386. mlmm_toolkit-0.3.4/tests/test_dynamic_dispatch_skill.py +35 -0
  387. mlmm_toolkit-0.3.4/tests/test_embedcharge_contract.py +68 -0
  388. mlmm_toolkit-0.3.4/tests/test_energy_plateau_stall.py +564 -0
  389. mlmm_toolkit-0.3.4/tests/test_error_handling.py +183 -0
  390. mlmm_toolkit-0.3.4/tests/test_example_scripts_contract.py +54 -0
  391. mlmm_toolkit-0.3.4/tests/test_extract.py +351 -0
  392. mlmm_toolkit-0.3.4/tests/test_freq_active_dof.py +60 -0
  393. mlmm_toolkit-0.3.4/tests/test_frequency_zero_cutoff.py +72 -0
  394. mlmm_toolkit-0.3.4/tests/test_gaussian_input.py +97 -0
  395. mlmm_toolkit-0.3.4/tests/test_generated_reference_contract.py +112 -0
  396. mlmm_toolkit-0.3.4/tests/test_gpu_kernel_parity.py +391 -0
  397. mlmm_toolkit-0.3.4/tests/test_heavy_ts_cycle_budget.py +37 -0
  398. mlmm_toolkit-0.3.4/tests/test_hessian_basis_reconciliation.py +186 -0
  399. mlmm_toolkit-0.3.4/tests/test_hessian_cache_handoff.py +767 -0
  400. mlmm_toolkit-0.3.4/tests/test_hessian_calc_boundaries.py +49 -0
  401. mlmm_toolkit-0.3.4/tests/test_hessian_file.py +457 -0
  402. mlmm_toolkit-0.3.4/tests/test_hessian_method_contract.py +279 -0
  403. mlmm_toolkit-0.3.4/tests/test_hessian_optimizer_internal_partial.py +42 -0
  404. mlmm_toolkit-0.3.4/tests/test_hessian_writeback.py +86 -0
  405. mlmm_toolkit-0.3.4/tests/test_import_graph_gate.py +132 -0
  406. mlmm_toolkit-0.3.4/tests/test_imports.py +40 -0
  407. mlmm_toolkit-0.3.4/tests/test_inline_xyz.py +26 -0
  408. mlmm_toolkit-0.3.4/tests/test_irc_active_space.py +98 -0
  409. mlmm_toolkit-0.3.4/tests/test_irc_endpoint_hessian_guard.py +157 -0
  410. mlmm_toolkit-0.3.4/tests/test_irc_never_stop.py +295 -0
  411. mlmm_toolkit-0.3.4/tests/test_layer_index_validation.py +32 -0
  412. mlmm_toolkit-0.3.4/tests/test_mace_anicc_factory.py +89 -0
  413. mlmm_toolkit-0.3.4/tests/test_mcp_runner.py +256 -0
  414. mlmm_toolkit-0.3.4/tests/test_mcp_tools.py +312 -0
  415. mlmm_toolkit-0.3.4/tests/test_mep_threshold_flags.py +186 -0
  416. mlmm_toolkit-0.3.4/tests/test_micro_force_equilibrium.py +128 -0
  417. mlmm_toolkit-0.3.4/tests/test_microiteration_contract.py +302 -0
  418. mlmm_toolkit-0.3.4/tests/test_microiteration_cycle_carry.py +1062 -0
  419. mlmm_toolkit-0.3.4/tests/test_ml_layer_precedence.py +204 -0
  420. mlmm_toolkit-0.3.4/tests/test_ml_region_outputs.py +95 -0
  421. mlmm_toolkit-0.3.4/tests/test_mlmm_calc_serial_gap.py +105 -0
  422. mlmm_toolkit-0.3.4/tests/test_mlmm_result_constraints.py +458 -0
  423. mlmm_toolkit-0.3.4/tests/test_mm_parm_preflight.py +394 -0
  424. mlmm_toolkit-0.3.4/tests/test_normal_modes_lowering.py +185 -0
  425. mlmm_toolkit-0.3.4/tests/test_oniom_export_regressions.py +409 -0
  426. mlmm_toolkit-0.3.4/tests/test_opt.py +120 -0
  427. mlmm_toolkit-0.3.4/tests/test_opt_flatten_mass_unweight.py +117 -0
  428. mlmm_toolkit-0.3.4/tests/test_optimizer_checkpoint.py +1115 -0
  429. mlmm_toolkit-0.3.4/tests/test_optimizer_trial_safeguards.py +1970 -0
  430. mlmm_toolkit-0.3.4/tests/test_orb_analytical_hessian_guard.py +35 -0
  431. mlmm_toolkit-0.3.4/tests/test_output_adapter.py +250 -0
  432. mlmm_toolkit-0.3.4/tests/test_path_logging_sections.py +22 -0
  433. mlmm_toolkit-0.3.4/tests/test_path_opt_output_boundary.py +158 -0
  434. mlmm_toolkit-0.3.4/tests/test_path_semantics.py +78 -0
  435. mlmm_toolkit-0.3.4/tests/test_pdb_fix.py +224 -0
  436. mlmm_toolkit-0.3.4/tests/test_pdb_indexing_resolver.py +243 -0
  437. mlmm_toolkit-0.3.4/tests/test_plotly_image_timeout.py +115 -0
  438. mlmm_toolkit-0.3.4/tests/test_prealign_activation_contract.py +86 -0
  439. mlmm_toolkit-0.3.4/tests/test_precision_dispatch.py +73 -0
  440. mlmm_toolkit-0.3.4/tests/test_preflight.py +32 -0
  441. mlmm_toolkit-0.3.4/tests/test_reject_uphill_flag.py +147 -0
  442. mlmm_toolkit-0.3.4/tests/test_release_metadata_contract.py +94 -0
  443. mlmm_toolkit-0.3.4/tests/test_release_workflow.py +57 -0
  444. mlmm_toolkit-0.3.4/tests/test_residue_catalog_parity.py +157 -0
  445. mlmm_toolkit-0.3.4/tests/test_restrained_pes_contract.py +450 -0
  446. mlmm_toolkit-0.3.4/tests/test_result_commit.py +228 -0
  447. mlmm_toolkit-0.3.4/tests/test_rfo_uphill_rejection.py +151 -0
  448. mlmm_toolkit-0.3.4/tests/test_run_session.py +336 -0
  449. mlmm_toolkit-0.3.4/tests/test_scan3d_csv_provenance.py +85 -0
  450. mlmm_toolkit-0.3.4/tests/test_scan_energy_accounting.py +70 -0
  451. mlmm_toolkit-0.3.4/tests/test_scan_oniom_smoke.py +552 -0
  452. mlmm_toolkit-0.3.4/tests/test_scan_precedence.py +160 -0
  453. mlmm_toolkit-0.3.4/tests/test_scan_validation_regressions.py +380 -0
  454. mlmm_toolkit-0.3.4/tests/test_scientific_outcomes.py +1420 -0
  455. mlmm_toolkit-0.3.4/tests/test_silent_noop_warnings.py +85 -0
  456. mlmm_toolkit-0.3.4/tests/test_skill_command_checker.py +68 -0
  457. mlmm_toolkit-0.3.4/tests/test_skill_drift_checker.py +47 -0
  458. mlmm_toolkit-0.3.4/tests/test_smoke_cli_contract.py +98 -0
  459. mlmm_toolkit-0.3.4/tests/test_sp_ml_region_contract.py +245 -0
  460. mlmm_toolkit-0.3.4/tests/test_stop_plateau_flag.py +219 -0
  461. mlmm_toolkit-0.3.4/tests/test_structure_formats.py +1812 -0
  462. mlmm_toolkit-0.3.4/tests/test_summary_log.py +680 -0
  463. mlmm_toolkit-0.3.4/tests/test_symmetrize.py +48 -0
  464. mlmm_toolkit-0.3.4/tests/test_thermo_policy.py +326 -0
  465. mlmm_toolkit-0.3.4/tests/test_thermo_symmetry_number.py +174 -0
  466. mlmm_toolkit-0.3.4/tests/test_thermochemistry_golden.py +375 -0
  467. mlmm_toolkit-0.3.4/tests/test_toy_system_contract.py +70 -0
  468. mlmm_toolkit-0.3.4/tests/test_tr_projection.py +528 -0
  469. mlmm_toolkit-0.3.4/tests/test_tr_projection_default.py +72 -0
  470. mlmm_toolkit-0.3.4/tests/test_tsopt_microiter_modes.py +282 -0
  471. mlmm_toolkit-0.3.4/tests/test_tsopt_output_boundary.py +104 -0
  472. mlmm_toolkit-0.3.4/tests/test_tsopt_probe_core_lifetime.py +107 -0
  473. mlmm_toolkit-0.3.4/tests/test_tsopt_reference_mode.py +65 -0
  474. mlmm_toolkit-0.3.4/tests/test_tsopt_soft_imaginary_warning.py +127 -0
  475. mlmm_toolkit-0.3.4/tests/test_tsopt_terminal_phva.py +168 -0
  476. mlmm_toolkit-0.3.4/tests/test_utils.py +458 -0
  477. mlmm_toolkit-0.3.4/tests/test_workers_option.py +116 -0
  478. mlmm_toolkit-0.3.4/tests/test_write_result_json.py +92 -0
  479. mlmm_toolkit-0.3.4/tests/test_xtb_embedcharge_parsers.py +67 -0
  480. mlmm_toolkit-0.3.4/tests/test_xyz_trajectory.py +147 -0
  481. mlmm_toolkit-0.3.4/thermoanalysis/QCData.py +293 -0
  482. mlmm_toolkit-0.3.4/thermoanalysis/config.py +84 -0
  483. mlmm_toolkit-0.3.4/thermoanalysis/constants.py +20 -0
  484. mlmm_toolkit-0.3.4/thermoanalysis/thermo.py +1020 -0
  485. mlmm_toolkit-0.3.1/.github/scripts/check_all_scan_contract.py +0 -133
  486. mlmm_toolkit-0.3.1/.github/scripts/check_engineering_markers.py +0 -89
  487. mlmm_toolkit-0.3.1/.github/scripts/check_help_registry.py +0 -98
  488. mlmm_toolkit-0.3.1/.github/scripts/check_intro_template.py +0 -118
  489. mlmm_toolkit-0.3.1/.github/scripts/check_markdown_links.py +0 -122
  490. mlmm_toolkit-0.3.1/.github/scripts/check_skill_commands.py +0 -140
  491. mlmm_toolkit-0.3.1/.github/scripts/check_skill_drift.py +0 -186
  492. mlmm_toolkit-0.3.1/.github/scripts/generate_reference.py +0 -310
  493. mlmm_toolkit-0.3.1/.github/scripts/run_docs_quality.py +0 -79
  494. mlmm_toolkit-0.3.1/.github/scripts/smoke_docs_commands.py +0 -300
  495. mlmm_toolkit-0.3.1/.github/scripts/smoke_dump_trajectories.py +0 -253
  496. mlmm_toolkit-0.3.1/.github/workflows/docs.yml +0 -51
  497. mlmm_toolkit-0.3.1/.github/workflows/docs_quality.yml +0 -42
  498. mlmm_toolkit-0.3.1/.github/workflows/markers.yml +0 -39
  499. mlmm_toolkit-0.3.1/.github/workflows/pytest.yml +0 -44
  500. mlmm_toolkit-0.3.1/.github/workflows/release.yml +0 -50
  501. mlmm_toolkit-0.3.1/.gitignore +0 -73
  502. mlmm_toolkit-0.3.1/CHANGELOG.md +0 -563
  503. mlmm_toolkit-0.3.1/CITATION.cff +0 -32
  504. mlmm_toolkit-0.3.1/CONTRIBUTING.md +0 -232
  505. mlmm_toolkit-0.3.1/MANIFEST.in +0 -10
  506. mlmm_toolkit-0.3.1/PKG-INFO +0 -926
  507. mlmm_toolkit-0.3.1/README.md +0 -181
  508. mlmm_toolkit-0.3.1/THIRD_PARTY_NOTICES.txt +0 -41
  509. mlmm_toolkit-0.3.1/docs/Makefile +0 -27
  510. mlmm_toolkit-0.3.1/docs/_static/custom.css +0 -110
  511. mlmm_toolkit-0.3.1/docs/_static/theme_toggle.js +0 -34
  512. mlmm_toolkit-0.3.1/docs/_templates/page.html +0 -339
  513. mlmm_toolkit-0.3.1/docs/add-elem-info.md +0 -72
  514. mlmm_toolkit-0.3.1/docs/all.md +0 -284
  515. mlmm_toolkit-0.3.1/docs/architecture.md +0 -389
  516. mlmm_toolkit-0.3.1/docs/backends.md +0 -193
  517. mlmm_toolkit-0.3.1/docs/bond-summary.md +0 -67
  518. mlmm_toolkit-0.3.1/docs/cli-conventions.md +0 -205
  519. mlmm_toolkit-0.3.1/docs/concepts.md +0 -328
  520. mlmm_toolkit-0.3.1/docs/conf.py +0 -145
  521. mlmm_toolkit-0.3.1/docs/define-layer.md +0 -76
  522. mlmm_toolkit-0.3.1/docs/device-hpc.md +0 -197
  523. mlmm_toolkit-0.3.1/docs/dft.md +0 -138
  524. mlmm_toolkit-0.3.1/docs/energy-diagram.md +0 -58
  525. mlmm_toolkit-0.3.1/docs/extract.md +0 -228
  526. mlmm_toolkit-0.3.1/docs/fix-altloc.md +0 -121
  527. mlmm_toolkit-0.3.1/docs/freq.md +0 -163
  528. mlmm_toolkit-0.3.1/docs/getting-started.md +0 -232
  529. mlmm_toolkit-0.3.1/docs/glossary.md +0 -147
  530. mlmm_toolkit-0.3.1/docs/index.md +0 -159
  531. mlmm_toolkit-0.3.1/docs/irc.md +0 -144
  532. mlmm_toolkit-0.3.1/docs/ja/add-elem-info.md +0 -64
  533. mlmm_toolkit-0.3.1/docs/ja/all.md +0 -311
  534. mlmm_toolkit-0.3.1/docs/ja/architecture.md +0 -379
  535. mlmm_toolkit-0.3.1/docs/ja/backends.md +0 -182
  536. mlmm_toolkit-0.3.1/docs/ja/cli-conventions.md +0 -345
  537. mlmm_toolkit-0.3.1/docs/ja/concepts.md +0 -318
  538. mlmm_toolkit-0.3.1/docs/ja/define-layer.md +0 -78
  539. mlmm_toolkit-0.3.1/docs/ja/device-hpc.md +0 -207
  540. mlmm_toolkit-0.3.1/docs/ja/dft.md +0 -139
  541. mlmm_toolkit-0.3.1/docs/ja/energy-diagram.md +0 -58
  542. mlmm_toolkit-0.3.1/docs/ja/extract.md +0 -242
  543. mlmm_toolkit-0.3.1/docs/ja/fix-altloc.md +0 -134
  544. mlmm_toolkit-0.3.1/docs/ja/freq.md +0 -167
  545. mlmm_toolkit-0.3.1/docs/ja/getting-started.md +0 -564
  546. mlmm_toolkit-0.3.1/docs/ja/glossary.md +0 -151
  547. mlmm_toolkit-0.3.1/docs/ja/index.md +0 -305
  548. mlmm_toolkit-0.3.1/docs/ja/irc.md +0 -188
  549. mlmm_toolkit-0.3.1/docs/ja/json-output.md +0 -209
  550. mlmm_toolkit-0.3.1/docs/ja/mcp_server.md +0 -110
  551. mlmm_toolkit-0.3.1/docs/ja/mlmm-calc.md +0 -138
  552. mlmm_toolkit-0.3.1/docs/ja/mm-parm.md +0 -83
  553. mlmm_toolkit-0.3.1/docs/ja/oniom-export.md +0 -78
  554. mlmm_toolkit-0.3.1/docs/ja/oniom-gaussian.md +0 -68
  555. mlmm_toolkit-0.3.1/docs/ja/oniom-import.md +0 -73
  556. mlmm_toolkit-0.3.1/docs/ja/oniom-orca.md +0 -70
  557. mlmm_toolkit-0.3.1/docs/ja/opt.md +0 -288
  558. mlmm_toolkit-0.3.1/docs/ja/output-layout.md +0 -85
  559. mlmm_toolkit-0.3.1/docs/ja/path-opt.md +0 -125
  560. mlmm_toolkit-0.3.1/docs/ja/path-search.md +0 -152
  561. mlmm_toolkit-0.3.1/docs/ja/python-api.md +0 -159
  562. mlmm_toolkit-0.3.1/docs/ja/quickstart-all.md +0 -37
  563. mlmm_toolkit-0.3.1/docs/ja/quickstart-scan-spec.md +0 -62
  564. mlmm_toolkit-0.3.1/docs/ja/quickstart-tsopt-freq.md +0 -50
  565. mlmm_toolkit-0.3.1/docs/ja/recipes-common-errors.md +0 -87
  566. mlmm_toolkit-0.3.1/docs/ja/reproducibility.md +0 -47
  567. mlmm_toolkit-0.3.1/docs/ja/scan.md +0 -269
  568. mlmm_toolkit-0.3.1/docs/ja/scan2d.md +0 -215
  569. mlmm_toolkit-0.3.1/docs/ja/scan3d.md +0 -194
  570. mlmm_toolkit-0.3.1/docs/ja/sp.md +0 -94
  571. mlmm_toolkit-0.3.1/docs/ja/trj2fig.md +0 -71
  572. mlmm_toolkit-0.3.1/docs/ja/troubleshooting.md +0 -577
  573. mlmm_toolkit-0.3.1/docs/ja/tsopt.md +0 -322
  574. mlmm_toolkit-0.3.1/docs/ja/yaml-reference.md +0 -656
  575. mlmm_toolkit-0.3.1/docs/json-output.md +0 -279
  576. mlmm_toolkit-0.3.1/docs/make.bat +0 -35
  577. mlmm_toolkit-0.3.1/docs/mcp_server.md +0 -131
  578. mlmm_toolkit-0.3.1/docs/mlmm-calc.md +0 -127
  579. mlmm_toolkit-0.3.1/docs/mlmm_toolkit_overview.png +0 -0
  580. mlmm_toolkit-0.3.1/docs/mm-parm.md +0 -83
  581. mlmm_toolkit-0.3.1/docs/oniom-export.md +0 -78
  582. mlmm_toolkit-0.3.1/docs/oniom-gaussian.md +0 -70
  583. mlmm_toolkit-0.3.1/docs/oniom-import.md +0 -74
  584. mlmm_toolkit-0.3.1/docs/oniom-orca.md +0 -72
  585. mlmm_toolkit-0.3.1/docs/opt.md +0 -156
  586. mlmm_toolkit-0.3.1/docs/output-layout.md +0 -86
  587. mlmm_toolkit-0.3.1/docs/path-opt.md +0 -125
  588. mlmm_toolkit-0.3.1/docs/path-search.md +0 -139
  589. mlmm_toolkit-0.3.1/docs/python-api.md +0 -161
  590. mlmm_toolkit-0.3.1/docs/quickstart-all.md +0 -40
  591. mlmm_toolkit-0.3.1/docs/quickstart-scan-spec.md +0 -62
  592. mlmm_toolkit-0.3.1/docs/quickstart-tsopt-freq.md +0 -50
  593. mlmm_toolkit-0.3.1/docs/recipes-common-errors.md +0 -94
  594. mlmm_toolkit-0.3.1/docs/reference/commands/add_elem_info.md +0 -22
  595. mlmm_toolkit-0.3.1/docs/reference/commands/all.md +0 -250
  596. mlmm_toolkit-0.3.1/docs/reference/commands/define_layer.md +0 -32
  597. mlmm_toolkit-0.3.1/docs/reference/commands/dft.md +0 -104
  598. mlmm_toolkit-0.3.1/docs/reference/commands/energy_diagram.md +0 -28
  599. mlmm_toolkit-0.3.1/docs/reference/commands/extract.md +0 -61
  600. mlmm_toolkit-0.3.1/docs/reference/commands/fix_altloc.md +0 -31
  601. mlmm_toolkit-0.3.1/docs/reference/commands/freq.md +0 -148
  602. mlmm_toolkit-0.3.1/docs/reference/commands/irc.md +0 -138
  603. mlmm_toolkit-0.3.1/docs/reference/commands/mm_parm.md +0 -40
  604. mlmm_toolkit-0.3.1/docs/reference/commands/oniom_export.md +0 -53
  605. mlmm_toolkit-0.3.1/docs/reference/commands/oniom_import.md +0 -24
  606. mlmm_toolkit-0.3.1/docs/reference/commands/opt.md +0 -171
  607. mlmm_toolkit-0.3.1/docs/reference/commands/path_opt.md +0 -142
  608. mlmm_toolkit-0.3.1/docs/reference/commands/path_search.md +0 -151
  609. mlmm_toolkit-0.3.1/docs/reference/commands/scan.md +0 -152
  610. mlmm_toolkit-0.3.1/docs/reference/commands/scan2d.md +0 -139
  611. mlmm_toolkit-0.3.1/docs/reference/commands/scan3d.md +0 -142
  612. mlmm_toolkit-0.3.1/docs/reference/commands/sp.md +0 -115
  613. mlmm_toolkit-0.3.1/docs/reference/commands/trj2fig.md +0 -31
  614. mlmm_toolkit-0.3.1/docs/reference/commands/tsopt.md +0 -175
  615. mlmm_toolkit-0.3.1/docs/reference/yaml.md +0 -122
  616. mlmm_toolkit-0.3.1/docs/reproducibility.md +0 -77
  617. mlmm_toolkit-0.3.1/docs/requirements_docs.txt +0 -5
  618. mlmm_toolkit-0.3.1/docs/scan.md +0 -264
  619. mlmm_toolkit-0.3.1/docs/scan2d.md +0 -215
  620. mlmm_toolkit-0.3.1/docs/scan3d.md +0 -202
  621. mlmm_toolkit-0.3.1/docs/sp.md +0 -92
  622. mlmm_toolkit-0.3.1/docs/trj2fig.md +0 -76
  623. mlmm_toolkit-0.3.1/docs/troubleshooting.md +0 -175
  624. mlmm_toolkit-0.3.1/docs/tsopt.md +0 -245
  625. mlmm_toolkit-0.3.1/docs/yaml-reference.md +0 -680
  626. mlmm_toolkit-0.3.1/examples/methyltransferase/run_all.sh +0 -4
  627. mlmm_toolkit-0.3.1/examples/methyltransferase/run_stepwise.sh +0 -40
  628. mlmm_toolkit-0.3.1/examples/toy_system/ml_region_p.pdb +0 -63
  629. mlmm_toolkit-0.3.1/examples/toy_system/opt_ase.py +0 -22
  630. mlmm_toolkit-0.3.1/examples/toy_system/p_complex_layered.pdb +0 -125
  631. mlmm_toolkit-0.3.1/examples/toy_system/run.sh +0 -100
  632. mlmm_toolkit-0.3.1/examples/toy_system/test_core.py +0 -25
  633. mlmm_toolkit-0.3.1/hessian_ff/README.md +0 -65
  634. mlmm_toolkit-0.3.1/hessian_ff/analytical_hessian.py +0 -609
  635. mlmm_toolkit-0.3.1/hessian_ff/forcefield.py +0 -339
  636. mlmm_toolkit-0.3.1/hessian_ff/loaders.py +0 -608
  637. mlmm_toolkit-0.3.1/hessian_ff/native/Makefile +0 -8
  638. mlmm_toolkit-0.3.1/hessian_ff/native/bonded_ext.cpp +0 -640
  639. mlmm_toolkit-0.3.1/hessian_ff/native/loader.py +0 -360
  640. mlmm_toolkit-0.3.1/hessian_ff/native/nonbonded_ext.cpp +0 -1150
  641. mlmm_toolkit-0.3.1/hessian_ff/terms/angle.py +0 -73
  642. mlmm_toolkit-0.3.1/hessian_ff/terms/dihedral.py +0 -141
  643. mlmm_toolkit-0.3.1/hessian_ff/terms/nonbonded.py +0 -209
  644. mlmm_toolkit-0.3.1/hessian_ff/tests/data/small/complex.parm7 +0 -1346
  645. mlmm_toolkit-0.3.1/hessian_ff/tests/data/small/complex.pdb +0 -125
  646. mlmm_toolkit-0.3.1/hessian_ff/tests/data/small/complex.rst7 +0 -63
  647. mlmm_toolkit-0.3.1/hessian_ff/tests/data/small/p_complex.parm7 +0 -1346
  648. mlmm_toolkit-0.3.1/hessian_ff/tests/data/small/p_complex_layered.pdb +0 -125
  649. mlmm_toolkit-0.3.1/hessian_ff/tests/test_energy_force.py +0 -49
  650. mlmm_toolkit-0.3.1/hessian_ff/tests/test_hessian.py +0 -137
  651. mlmm_toolkit-0.3.1/hessian_ff/tests/test_validation.py +0 -40
  652. mlmm_toolkit-0.3.1/hessian_ff/workflows.py +0 -889
  653. mlmm_toolkit-0.3.1/mlmm/__init__.py +0 -37
  654. mlmm_toolkit-0.3.1/mlmm/_version.py +0 -24
  655. mlmm_toolkit-0.3.1/mlmm/backends/__init__.py +0 -161
  656. mlmm_toolkit-0.3.1/mlmm/backends/_determinism.py +0 -109
  657. mlmm_toolkit-0.3.1/mlmm/backends/custom.py +0 -144
  658. mlmm_toolkit-0.3.1/mlmm/backends/mlmm_calc.py +0 -2623
  659. mlmm_toolkit-0.3.1/mlmm/backends/xtb_embedcharge_correction.py +0 -485
  660. mlmm_toolkit-0.3.1/mlmm/cli/app.py +0 -948
  661. mlmm_toolkit-0.3.1/mlmm/cli/common_options.py +0 -377
  662. mlmm_toolkit-0.3.1/mlmm/cli/decorators.py +0 -177
  663. mlmm_toolkit-0.3.1/mlmm/cli/default_group.py +0 -441
  664. mlmm_toolkit-0.3.1/mlmm/cli/help_pages.py +0 -75
  665. mlmm_toolkit-0.3.1/mlmm/cli/preflight.py +0 -88
  666. mlmm_toolkit-0.3.1/mlmm/core/__init__.py +0 -14
  667. mlmm_toolkit-0.3.1/mlmm/core/calc_eval.py +0 -51
  668. mlmm_toolkit-0.3.1/mlmm/core/defaults.py +0 -428
  669. mlmm_toolkit-0.3.1/mlmm/core/logging.py +0 -58
  670. mlmm_toolkit-0.3.1/mlmm/core/residue_data.py +0 -137
  671. mlmm_toolkit-0.3.1/mlmm/core/utils.py +0 -3194
  672. mlmm_toolkit-0.3.1/mlmm/domain/add_elem_info.py +0 -359
  673. mlmm_toolkit-0.3.1/mlmm/domain/bond_changes.py +0 -260
  674. mlmm_toolkit-0.3.1/mlmm/domain/bond_summary.py +0 -150
  675. mlmm_toolkit-0.3.1/mlmm/io/__init__.py +0 -13
  676. mlmm_toolkit-0.3.1/mlmm/io/energy_diagram.py +0 -269
  677. mlmm_toolkit-0.3.1/mlmm/io/hessian_cache.py +0 -44
  678. mlmm_toolkit-0.3.1/mlmm/io/hessian_calc.py +0 -142
  679. mlmm_toolkit-0.3.1/mlmm/io/pdb_fix.py +0 -407
  680. mlmm_toolkit-0.3.1/mlmm/io/summary.py +0 -859
  681. mlmm_toolkit-0.3.1/mlmm/io/trj2fig.py +0 -446
  682. mlmm_toolkit-0.3.1/mlmm/mcp/__init__.py +0 -25
  683. mlmm_toolkit-0.3.1/mlmm/mcp/_runner.py +0 -208
  684. mlmm_toolkit-0.3.1/mlmm/mcp/_tools.py +0 -882
  685. mlmm_toolkit-0.3.1/mlmm/workflows/_all_helpers.py +0 -463
  686. mlmm_toolkit-0.3.1/mlmm/workflows/align_freeze.py +0 -586
  687. mlmm_toolkit-0.3.1/mlmm/workflows/all.py +0 -4425
  688. mlmm_toolkit-0.3.1/mlmm/workflows/define_layer.py +0 -560
  689. mlmm_toolkit-0.3.1/mlmm/workflows/dft.py +0 -1341
  690. mlmm_toolkit-0.3.1/mlmm/workflows/extract.py +0 -2259
  691. mlmm_toolkit-0.3.1/mlmm/workflows/freq.py +0 -1684
  692. mlmm_toolkit-0.3.1/mlmm/workflows/irc.py +0 -896
  693. mlmm_toolkit-0.3.1/mlmm/workflows/mm_parm.py +0 -982
  694. mlmm_toolkit-0.3.1/mlmm/workflows/oniom_export.py +0 -2027
  695. mlmm_toolkit-0.3.1/mlmm/workflows/oniom_import.py +0 -457
  696. mlmm_toolkit-0.3.1/mlmm/workflows/opt.py +0 -1908
  697. mlmm_toolkit-0.3.1/mlmm/workflows/path_opt.py +0 -1601
  698. mlmm_toolkit-0.3.1/mlmm/workflows/path_search.py +0 -2381
  699. mlmm_toolkit-0.3.1/mlmm/workflows/restraints.py +0 -133
  700. mlmm_toolkit-0.3.1/mlmm/workflows/scan.py +0 -1126
  701. mlmm_toolkit-0.3.1/mlmm/workflows/scan2d.py +0 -1185
  702. mlmm_toolkit-0.3.1/mlmm/workflows/scan3d.py +0 -1245
  703. mlmm_toolkit-0.3.1/mlmm/workflows/scan_common.py +0 -179
  704. mlmm_toolkit-0.3.1/mlmm/workflows/sp.py +0 -473
  705. mlmm_toolkit-0.3.1/mlmm/workflows/tsopt.py +0 -3192
  706. mlmm_toolkit-0.3.1/mlmm_toolkit.egg-info/PKG-INFO +0 -926
  707. mlmm_toolkit-0.3.1/mlmm_toolkit.egg-info/SOURCES.txt +0 -459
  708. mlmm_toolkit-0.3.1/mlmm_toolkit.egg-info/entry_points.txt +0 -3
  709. mlmm_toolkit-0.3.1/mlmm_toolkit.egg-info/requires.txt +0 -55
  710. mlmm_toolkit-0.3.1/pyproject.toml +0 -128
  711. mlmm_toolkit-0.3.1/pysisyphus/Geometry.py +0 -1755
  712. mlmm_toolkit-0.3.1/pysisyphus/README.md +0 -73
  713. mlmm_toolkit-0.3.1/pysisyphus/_array.py +0 -79
  714. mlmm_toolkit-0.3.1/pysisyphus/calculators/Calculator.py +0 -844
  715. mlmm_toolkit-0.3.1/pysisyphus/calculators/Dimer.py +0 -681
  716. mlmm_toolkit-0.3.1/pysisyphus/config.py +0 -133
  717. mlmm_toolkit-0.3.1/pysisyphus/cos/ChainOfStates.py +0 -732
  718. mlmm_toolkit-0.3.1/pysisyphus/cos/GrowingChainOfStates.py +0 -71
  719. mlmm_toolkit-0.3.1/pysisyphus/cos/GrowingString.py +0 -508
  720. mlmm_toolkit-0.3.1/pysisyphus/cos/__init__.py +0 -15
  721. mlmm_toolkit-0.3.1/pysisyphus/helpers.py +0 -624
  722. mlmm_toolkit-0.3.1/pysisyphus/helpers_pure.py +0 -649
  723. mlmm_toolkit-0.3.1/pysisyphus/intcoords/BondedFragment.py +0 -32
  724. mlmm_toolkit-0.3.1/pysisyphus/intcoords/CartesianCoords.py +0 -140
  725. mlmm_toolkit-0.3.1/pysisyphus/intcoords/DLC.py +0 -197
  726. mlmm_toolkit-0.3.1/pysisyphus/intcoords/DistanceFunction.py +0 -34
  727. mlmm_toolkit-0.3.1/pysisyphus/intcoords/DummyImproper.py +0 -70
  728. mlmm_toolkit-0.3.1/pysisyphus/intcoords/DummyTorsion.py +0 -72
  729. mlmm_toolkit-0.3.1/pysisyphus/intcoords/LinearDisplacement.py +0 -80
  730. mlmm_toolkit-0.3.1/pysisyphus/intcoords/PrimTypes.py +0 -286
  731. mlmm_toolkit-0.3.1/pysisyphus/intcoords/RedundantCoords.py +0 -659
  732. mlmm_toolkit-0.3.1/pysisyphus/intcoords/Torsion.py +0 -101
  733. mlmm_toolkit-0.3.1/pysisyphus/intcoords/augment_bonds.py +0 -126
  734. mlmm_toolkit-0.3.1/pysisyphus/intcoords/setup.py +0 -965
  735. mlmm_toolkit-0.3.1/pysisyphus/intcoords/setup_fast.py +0 -176
  736. mlmm_toolkit-0.3.1/pysisyphus/intcoords/update.py +0 -272
  737. mlmm_toolkit-0.3.1/pysisyphus/io/hessian.py +0 -72
  738. mlmm_toolkit-0.3.1/pysisyphus/io/pdb.py +0 -301
  739. mlmm_toolkit-0.3.1/pysisyphus/io/qcschema.py +0 -34
  740. mlmm_toolkit-0.3.1/pysisyphus/irc/DWI.py +0 -108
  741. mlmm_toolkit-0.3.1/pysisyphus/irc/EulerPC.py +0 -362
  742. mlmm_toolkit-0.3.1/pysisyphus/irc/IRC.py +0 -917
  743. mlmm_toolkit-0.3.1/pysisyphus/irc/Instanton.py +0 -307
  744. mlmm_toolkit-0.3.1/pysisyphus/linalg.py +0 -411
  745. mlmm_toolkit-0.3.1/pysisyphus/modefollow/lanczos.py +0 -95
  746. mlmm_toolkit-0.3.1/pysisyphus/optimizers/HessianOptimizer.py +0 -1270
  747. mlmm_toolkit-0.3.1/pysisyphus/optimizers/LBFGS.py +0 -228
  748. mlmm_toolkit-0.3.1/pysisyphus/optimizers/Optimizer.py +0 -1130
  749. mlmm_toolkit-0.3.1/pysisyphus/optimizers/RFOptimizer.py +0 -181
  750. mlmm_toolkit-0.3.1/pysisyphus/optimizers/StringOptimizer.py +0 -173
  751. mlmm_toolkit-0.3.1/pysisyphus/optimizers/closures.py +0 -301
  752. mlmm_toolkit-0.3.1/pysisyphus/optimizers/gdiis.py +0 -281
  753. mlmm_toolkit-0.3.1/pysisyphus/optimizers/hessian_updates.py +0 -417
  754. mlmm_toolkit-0.3.1/pysisyphus/optimizers/poly_fit.py +0 -285
  755. mlmm_toolkit-0.3.1/pysisyphus/optimizers/restrict_step.py +0 -145
  756. mlmm_toolkit-0.3.1/pysisyphus/stocastic/Pipeline.py +0 -303
  757. mlmm_toolkit-0.3.1/pysisyphus/tsoptimizers/RSIRFOptimizer.py +0 -56
  758. mlmm_toolkit-0.3.1/pysisyphus/tsoptimizers/RSPRFOptimizer.py +0 -190
  759. mlmm_toolkit-0.3.1/pysisyphus/tsoptimizers/TRIM.py +0 -95
  760. mlmm_toolkit-0.3.1/pysisyphus/tsoptimizers/TSHessianOptimizer.py +0 -546
  761. mlmm_toolkit-0.3.1/pysisyphus/xyzloader.py +0 -168
  762. mlmm_toolkit-0.3.1/skills/README.md +0 -49
  763. mlmm_toolkit-0.3.1/skills/mlmm-architecture/SKILL.md +0 -78
  764. mlmm_toolkit-0.3.1/skills/mlmm-cli/SKILL.md +0 -157
  765. mlmm_toolkit-0.3.1/skills/mlmm-cli/add-elem-info.md +0 -62
  766. mlmm_toolkit-0.3.1/skills/mlmm-cli/all-endpoint-mep.md +0 -126
  767. mlmm_toolkit-0.3.1/skills/mlmm-cli/all-scan-list.md +0 -150
  768. mlmm_toolkit-0.3.1/skills/mlmm-cli/all-ts-only.md +0 -185
  769. mlmm_toolkit-0.3.1/skills/mlmm-cli/all.md +0 -170
  770. mlmm_toolkit-0.3.1/skills/mlmm-cli/bond-summary.md +0 -77
  771. mlmm_toolkit-0.3.1/skills/mlmm-cli/define-layer.md +0 -102
  772. mlmm_toolkit-0.3.1/skills/mlmm-cli/dft.md +0 -145
  773. mlmm_toolkit-0.3.1/skills/mlmm-cli/extract.md +0 -113
  774. mlmm_toolkit-0.3.1/skills/mlmm-cli/fix-altloc.md +0 -64
  775. mlmm_toolkit-0.3.1/skills/mlmm-cli/freq.md +0 -134
  776. mlmm_toolkit-0.3.1/skills/mlmm-cli/irc.md +0 -138
  777. mlmm_toolkit-0.3.1/skills/mlmm-cli/mm-parm.md +0 -144
  778. mlmm_toolkit-0.3.1/skills/mlmm-cli/oniom-export.md +0 -102
  779. mlmm_toolkit-0.3.1/skills/mlmm-cli/oniom-import.md +0 -85
  780. mlmm_toolkit-0.3.1/skills/mlmm-cli/opt.md +0 -108
  781. mlmm_toolkit-0.3.1/skills/mlmm-cli/path-opt.md +0 -100
  782. mlmm_toolkit-0.3.1/skills/mlmm-cli/path-search.md +0 -128
  783. mlmm_toolkit-0.3.1/skills/mlmm-cli/scan.md +0 -106
  784. mlmm_toolkit-0.3.1/skills/mlmm-cli/scan2d.md +0 -90
  785. mlmm_toolkit-0.3.1/skills/mlmm-cli/scan3d.md +0 -69
  786. mlmm_toolkit-0.3.1/skills/mlmm-cli/sp.md +0 -92
  787. mlmm_toolkit-0.3.1/skills/mlmm-cli/trj2fig.md +0 -61
  788. mlmm_toolkit-0.3.1/skills/mlmm-cli/tsopt.md +0 -149
  789. mlmm_toolkit-0.3.1/skills/mlmm-env-detect/SKILL.md +0 -178
  790. mlmm_toolkit-0.3.1/skills/mlmm-hpc/SKILL.md +0 -214
  791. mlmm_toolkit-0.3.1/skills/mlmm-hpc/dynamic-dispatch.md +0 -179
  792. mlmm_toolkit-0.3.1/skills/mlmm-install-backends/SKILL.md +0 -177
  793. mlmm_toolkit-0.3.1/skills/mlmm-install-backends/aimnet2.md +0 -81
  794. mlmm_toolkit-0.3.1/skills/mlmm-install-backends/core.md +0 -120
  795. mlmm_toolkit-0.3.1/skills/mlmm-install-backends/dft.md +0 -98
  796. mlmm_toolkit-0.3.1/skills/mlmm-install-backends/env-cuda.md +0 -162
  797. mlmm_toolkit-0.3.1/skills/mlmm-install-backends/mace.md +0 -95
  798. mlmm_toolkit-0.3.1/skills/mlmm-install-backends/orb.md +0 -95
  799. mlmm_toolkit-0.3.1/skills/mlmm-install-backends/uma.md +0 -112
  800. mlmm_toolkit-0.3.1/skills/mlmm-install-backends/xtb.md +0 -81
  801. mlmm_toolkit-0.3.1/skills/mlmm-mcp/SKILL.md +0 -123
  802. mlmm_toolkit-0.3.1/skills/mlmm-overview/SKILL.md +0 -153
  803. mlmm_toolkit-0.3.1/skills/mlmm-structure-io/SKILL.md +0 -139
  804. mlmm_toolkit-0.3.1/skills/mlmm-structure-io/charge-multiplicity.md +0 -176
  805. mlmm_toolkit-0.3.1/skills/mlmm-structure-io/gjf.md +0 -88
  806. mlmm_toolkit-0.3.1/skills/mlmm-structure-io/parm7.md +0 -118
  807. mlmm_toolkit-0.3.1/skills/mlmm-structure-io/pdb.md +0 -176
  808. mlmm_toolkit-0.3.1/skills/mlmm-ts-strategy/SKILL.md +0 -122
  809. mlmm_toolkit-0.3.1/skills/mlmm-workflows-output/SKILL.md +0 -379
  810. mlmm_toolkit-0.3.1/tests/README.md +0 -28
  811. mlmm_toolkit-0.3.1/tests/conftest.py +0 -25
  812. mlmm_toolkit-0.3.1/tests/smoke/check_frozen_atoms.py +0 -42
  813. mlmm_toolkit-0.3.1/tests/smoke/p_complex.parm7 +0 -1346
  814. mlmm_toolkit-0.3.1/tests/smoke/p_complex.pdb +0 -125
  815. mlmm_toolkit-0.3.1/tests/smoke/p_complex_layered.pdb +0 -125
  816. mlmm_toolkit-0.3.1/tests/smoke/pocket_r.pdb +0 -48
  817. mlmm_toolkit-0.3.1/tests/smoke/r_complex.pdb +0 -122
  818. mlmm_toolkit-0.3.1/tests/smoke/r_complex_layered.pdb +0 -122
  819. mlmm_toolkit-0.3.1/tests/test_all_helpers.py +0 -263
  820. mlmm_toolkit-0.3.1/tests/test_all_layer_honor.py +0 -81
  821. mlmm_toolkit-0.3.1/tests/test_all_multiplicity_option.py +0 -20
  822. mlmm_toolkit-0.3.1/tests/test_array_shim_equivalence.py +0 -156
  823. mlmm_toolkit-0.3.1/tests/test_bond_changes.py +0 -127
  824. mlmm_toolkit-0.3.1/tests/test_bool_compat_cli.py +0 -117
  825. mlmm_toolkit-0.3.1/tests/test_cli_feature_regressions.py +0 -166
  826. mlmm_toolkit-0.3.1/tests/test_cli_help.py +0 -201
  827. mlmm_toolkit-0.3.1/tests/test_cli_utils.py +0 -53
  828. mlmm_toolkit-0.3.1/tests/test_custom_calc_file.py +0 -99
  829. mlmm_toolkit-0.3.1/tests/test_default_group.py +0 -204
  830. mlmm_toolkit-0.3.1/tests/test_defaults.py +0 -78
  831. mlmm_toolkit-0.3.1/tests/test_error_handling.py +0 -107
  832. mlmm_toolkit-0.3.1/tests/test_extract.py +0 -220
  833. mlmm_toolkit-0.3.1/tests/test_freq_active_dof.py +0 -35
  834. mlmm_toolkit-0.3.1/tests/test_imports.py +0 -34
  835. mlmm_toolkit-0.3.1/tests/test_mcp_runner.py +0 -36
  836. mlmm_toolkit-0.3.1/tests/test_mlmm_calc_serial_gap.py +0 -60
  837. mlmm_toolkit-0.3.1/tests/test_mm_parm_preflight.py +0 -81
  838. mlmm_toolkit-0.3.1/tests/test_opt.py +0 -97
  839. mlmm_toolkit-0.3.1/tests/test_precision_dispatch.py +0 -58
  840. mlmm_toolkit-0.3.1/tests/test_preflight.py +0 -62
  841. mlmm_toolkit-0.3.1/tests/test_scan_oniom_smoke.py +0 -416
  842. mlmm_toolkit-0.3.1/tests/test_summary_log.py +0 -67
  843. mlmm_toolkit-0.3.1/tests/test_symmetrize.py +0 -49
  844. mlmm_toolkit-0.3.1/tests/test_utils.py +0 -290
  845. mlmm_toolkit-0.3.1/tests/test_write_result_json.py +0 -62
  846. mlmm_toolkit-0.3.1/thermoanalysis/QCData.py +0 -244
  847. mlmm_toolkit-0.3.1/thermoanalysis/README.md +0 -46
  848. mlmm_toolkit-0.3.1/thermoanalysis/config.py +0 -3
  849. mlmm_toolkit-0.3.1/thermoanalysis/constants.py +0 -20
  850. mlmm_toolkit-0.3.1/thermoanalysis/thermo.py +0 -1011
  851. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/LICENSE +0 -0
  852. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/docs/ja/bond-summary.md +0 -0
  853. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/docs/reference/commands/bond_summary.md +0 -0
  854. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/docs/reference/commands/index.md +0 -0
  855. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/examples/mcp_client_config.json +0 -0
  856. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/examples/methyltransferase/complex.parm7 +0 -0
  857. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/examples/methyltransferase/complex.pdb +0 -0
  858. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/examples/methyltransferase/complex.rst7 +0 -0
  859. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/examples/methyltransferase/pocket.pdb +0 -0
  860. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/examples/methyltransferase/r_layered.pdb +0 -0
  861. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/examples/toy_system/ml_region_r.pdb +0 -0
  862. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/examples/toy_system/p_complex.pdb +0 -0
  863. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/examples/toy_system/p_complex.rst7 +0 -0
  864. /mlmm_toolkit-0.3.1/examples/toy_system/p_complex.parm7 → /mlmm_toolkit-0.3.4/examples/toy_system/p_toy.parm7 +0 -0
  865. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/examples/toy_system/r_complex.pdb +0 -0
  866. /mlmm_toolkit-0.3.1/examples/toy_system/r_complex_layered.pdb → /mlmm_toolkit-0.3.4/examples/toy_system/r_toy.pdb +0 -0
  867. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/hessian_ff/__init__.py +0 -0
  868. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/hessian_ff/constants.py +0 -0
  869. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/hessian_ff/native/__init__.py +0 -0
  870. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/hessian_ff/native/analytical_hessian.py +0 -0
  871. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/hessian_ff/native/analytical_hessian_ext.cpp +0 -0
  872. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/hessian_ff/native/bonded.py +0 -0
  873. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/hessian_ff/native/nonbonded.py +0 -0
  874. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/hessian_ff/prmtop_parmed.py +0 -0
  875. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/hessian_ff/system.py +0 -0
  876. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/hessian_ff/terms/__init__.py +0 -0
  877. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/hessian_ff/terms/bond.py +0 -0
  878. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/hessian_ff/terms/cmap.py +0 -0
  879. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/hessian_ff/tests/__init__.py +0 -0
  880. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/hessian_ff/tests/conftest.py +0 -0
  881. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/hessian_ff/tests/test_coords_input.py +0 -0
  882. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/hessian_ff/tests/test_smoke.py +0 -0
  883. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/mlmm/__main__.py +0 -0
  884. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/mlmm/cli/__init__.py +0 -0
  885. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/mlmm/cli/bool_compat.py +0 -0
  886. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/mlmm/domain/__init__.py +0 -0
  887. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/mlmm/mcp/server.py +0 -0
  888. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/mlmm/py.typed +0 -0
  889. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/mlmm/workflows/__init__.py +0 -0
  890. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/mlmm_toolkit.egg-info/dependency_links.txt +0 -0
  891. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/mlmm_toolkit.egg-info/top_level.txt +0 -0
  892. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/LICENSE +0 -0
  893. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/TablePrinter.py +0 -0
  894. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/__init__.py +0 -0
  895. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/calculators/__init__.py +0 -0
  896. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/constants.py +0 -0
  897. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/elem_data.py +0 -0
  898. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/intcoords/Bend.py +0 -0
  899. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/intcoords/Bend2.py +0 -0
  900. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/intcoords/Cartesian.py +0 -0
  901. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/intcoords/Coords.py +0 -0
  902. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/intcoords/LinearBend.py +0 -0
  903. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/intcoords/OutOfPlane.py +0 -0
  904. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/intcoords/Primitive.py +0 -0
  905. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/intcoords/RobustTorsion.py +0 -0
  906. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/intcoords/Rotation.py +0 -0
  907. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/intcoords/Stretch.py +0 -0
  908. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/intcoords/Torsion2.py +0 -0
  909. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/intcoords/Translation.py +0 -0
  910. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/intcoords/__init__.py +0 -0
  911. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/intcoords/derivatives.py +0 -0
  912. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/intcoords/eval.py +0 -0
  913. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/intcoords/exceptions.py +0 -0
  914. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/intcoords/helpers.py +0 -0
  915. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/intcoords/logging_conf.py +0 -0
  916. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/intcoords/valid.py +0 -0
  917. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/io/__init__.py +0 -0
  918. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/io/cjson.py +0 -0
  919. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/io/crd.py +0 -0
  920. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/io/hdf5.py +0 -0
  921. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/io/mol2.py +0 -0
  922. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/io/xyz.py +0 -0
  923. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/io/zmat.py +0 -0
  924. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/irc/__init__.py +0 -0
  925. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/modefollow/NormalMode.py +0 -0
  926. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/modefollow/__init__.py +0 -0
  927. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/optimizers/BacktrackingOptimizer.py +0 -0
  928. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/optimizers/__init__.py +0 -0
  929. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/optimizers/exceptions.py +0 -0
  930. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/optimizers/guess_hessians.py +0 -0
  931. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/stocastic/FragmentKick.py +0 -0
  932. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/stocastic/Kick.py +0 -0
  933. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/stocastic/__init__.py +0 -0
  934. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/stocastic/align.py +0 -0
  935. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/pysisyphus/tsoptimizers/__init__.py +0 -0
  936. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/setup.cfg +0 -0
  937. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/skills/mlmm-cli/energy-diagram.md +0 -0
  938. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/skills/mlmm-install-backends/ambertools.md +0 -0
  939. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/skills/mlmm-structure-io/xyz.md +0 -0
  940. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/tests/test_e2e_minimal.py +0 -0
  941. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/tests/test_error_envelope.py +0 -0
  942. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/tests/test_freeze_noncart_geometry.py +0 -0
  943. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/tests/test_help_grouping.py +0 -0
  944. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/thermoanalysis/LICENSE +0 -0
  945. {mlmm_toolkit-0.3.1 → mlmm_toolkit-0.3.4}/thermoanalysis/__init__.py +0 -0
@@ -0,0 +1,1182 @@
1
+ # Changelog
2
+
3
+ All notable changes to **mlmm-toolkit** will be documented in this file.
4
+
5
+ The format follows [Keep a Changelog](https://keepachangelog.com/).
6
+
7
+ ## [Unreleased]
8
+
9
+ _No changes yet._
10
+
11
+ ## [0.3.4] — 2026-09-02
12
+
13
+ ### Added
14
+ - Add full-system COMT and refined-path BezA Endpoint examples to the Colab GUI.
15
+
16
+ ### Changed
17
+ - Improve the Colab workflow guidance, built-in example walkthroughs, and
18
+ linked setup and runtime help.
19
+
20
+ ## [0.3.3] — 2026-08-29
21
+
22
+ > Upgrade warning: unchanged inputs can produce different geometries, energies/barriers,
23
+ > vibrational classifications, thermochemistry, and scientific/terminal status. Users of
24
+ > `result.json`/`summary.json` must review the Breaking changes and output-schema updates below.
25
+
26
+ ### Breaking changes
27
+ - Remove the public `--tr-projection` option and the `legacy-active` treatment.
28
+ Frozen-boundary PHVA now always uses the constrained treatment; stale YAML
29
+ values fail explicitly.
30
+ - Restrict `path-opt` / `path-search` / `all` `--thresh` to single-structure
31
+ optimizations. It no longer sets the GSM string-optimizer preset, which now
32
+ has its own `--thresh-gsm`. A command that relied on `--thresh` to tighten or
33
+ loosen the string optimizer must pass `--thresh-gsm` as well.
34
+ - **CMAP now defaults to the force-field-faithful policy in both MM layers.**
35
+ Parm7 CMAP terms are preserved in both REAL and MODEL calculations, so complete
36
+ model-internal terms cancel in `E_real_low + E_high - E_model_low` while boundary
37
+ terms remain part of the low-level coupling. `--no-cmap` is an explicit
38
+ modified-force-field opt-out and removes CMAP from both layers. Earlier releases
39
+ removed CMAP only from MODEL by default, which mixed two low-level Hamiltonians.
40
+ - **`add-elem-info` no longer overwrites its input by default.** Omitting `-o`
41
+ now writes `<input>_add_elem.pdb`. Pass `--inplace` to replace the input;
42
+ `--overwrite` continues to mean re-infer existing element fields.
43
+ - **JSON schema 2.0 (breaking).** UMA-specific summary keys became backend-neutral MLIP keys
44
+ (`post_segments[].uma` → `.mlip`, `gibbs_uma` → `gibbs_mlip`, `gibbs_dft_uma` → `gibbs_dft_mlip`);
45
+ the old keys were removed. This is `schema_version: "2.0"`. Update parsers before upgrading.
46
+ - **Removed the `add_uma_precision_option` Python alias** (`mlmm.cli.common_options`). It previously
47
+ aliased `add_precision_option`; external scripts importing the old name break with `ImportError`.
48
+ Migration: import `add_precision_option`.
49
+ - **MCP tools now reject `extra_args` that override a managed output option.** Passing `-o`/`--out-dir`
50
+ (or another MCP-managed output switch) through `extra_args` now raises `ValueError` instead of being
51
+ forwarded verbatim. Migration: use the tool's own output parameters.
52
+ - **MCP `search_paths` now requires `product_pdb` (keyword-only).** The tool previously
53
+ sent one structure to a CLI that requires at least two. Migration: pass the product
54
+ endpoint explicitly; optional `intermediate_pdbs` are inserted in reaction order.
55
+ - **Hessian handoffs now verify electronic state.** `freq --dump-hess` writes
56
+ schema 2 with model charge and multiplicity. `irc --read-hess` rejects a
57
+ state mismatch; schema-1 files require explicit
58
+ `--allow-unverified-hess-state` after independent verification.
59
+ - **ONIOM reference imports now verify atom order.** Exported inputs carry an
60
+ atom-order digest. `oniom-import --ref-pdb` rejects malformed or mismatched
61
+ markers; markerless files with repeated elements require
62
+ `--allow-unverified-ref-order`, which cannot bypass a known mismatch.
63
+ - **The energy-plateau stop is now opt-in and off by default**
64
+ (`opt.energy_plateau: false`), and it never applies to the MM micro
65
+ iterations of a `--microiter` run. A flat energy is not evidence of a
66
+ stationary point, so a run that stopped there reported `stalled` while
67
+ leaving the geometry unconverged; in the micro step a flat MM energy with
68
+ forces still above threshold ended the macro/micro alternation with the
69
+ environment unrelaxed, which left extra imaginary modes in the TS search.
70
+ `--max-cycles` (macro) and `microiter.micro_max_cycles` (micro) remain the
71
+ real bounds. Turn the macro stop back on with `--stop-plateau` (see Added);
72
+ YAML `opt.energy_plateau: true` also still enables it.
73
+
74
+ ### Added
75
+ - Add `--gsm-param {equi,energy}` to `all`, `path-opt`, and `path-search` as an advanced GSM node-parameterization control; `equi` remains the default.
76
+ - Add `--stop-plateau/--no-stop-plateau`, `--stop-plateau-thresh`, and
77
+ `--stop-plateau-window` to `opt`, `tsopt`, and `all`, exposing the
78
+ energy-plateau stop and its two tuning values on the command line.
79
+ - Add `--thresh-gsm` and `--thresh-dmf` to `all`, `path-opt`, and
80
+ `path-search`. The former selects the GSM string-optimizer preset; the latter
81
+ selects the DMF IPOPT dual-infeasibility tolerance (`tight`, `middle`,
82
+ `loose`, or a positive float).
83
+ - Add `all --tsopt-from-mep-tan/--no-tsopt-from-mep-tan`. The MEP tangent guides the
84
+ initial TS root by default; when disabled, TSOPT selects from the initial-structure
85
+ Hessian modes.
86
+ - Add a runnable full-system BezA example with reactant, intermediate, and
87
+ product structures plus endpoint-MEP and staged-scan workflows.
88
+ - Report citations for the methods actually used at the end of `summary.log`
89
+ and final stdout, and expose the same `{method, citation, doi}` records as
90
+ `summary.json.references`.
91
+ - Write directly inspectable ML-region structures both before and after link
92
+ hydrogens generated from parm7 ML/MM boundary bonds, including PDB companions
93
+ for PDB input.
94
+ - Install AmberTools in the Colab notebook through an isolated Miniforge, so
95
+ `mm-parm` and the DMF path mode run there; the Colab Python environment is
96
+ left untouched.
97
+ - Cancel a running Colab job from the interface, and send an occupied output
98
+ directory to `result(1)`, `result(2)`, … instead of writing into it.
99
+ - Add an mmCIF/large-PDB bridge (atom-identity–preserving; multi-model input keeps the first model,
100
+ with a warning), exact selectors, safe duplicate atom names, and root/segment CIF companions with
101
+ original identifiers.
102
+ - Add `tsopt --ref-mode`, opt-in IRC never-stop traversal, and analytical
103
+ Hessians for ORB, MACE, and AIMNet2.
104
+ - Add a release-pinned, keyboard-accessible Colab GUI that preserves full-system
105
+ coordinate/topology identity, accepts a separately prepared model, builds workflow-aware
106
+ commands, reaps interrupted jobs, and limits result views/downloads to the current run.
107
+ Its compact Input, Viewer, Options, and Results workspace co-locates workflow
108
+ selection with a 4:3 molecular view. Exact 3D picks use Mol*'s native focus and
109
+ selection behavior without rebuilding the viewer. The GUI keeps primary-input
110
+ selections separate from view-only secondary structures and rolls back failed structure
111
+ switches. Center selectors retain chain and insertion-code identity;
112
+ incompatible secondary views suppress primary selection and measurement overlays.
113
+ Validation follows the exact command and content hashes for every existing-file input
114
+ declared by the selected Click command, while its transcript remains separate from
115
+ current-run diagnostics. Private runtime files avoid upload collisions; Results links one
116
+ trajectory control to the 3D frame and energy-profile cursor, while keeping structures,
117
+ diagrams, SVG, HTML, CSV, PDF, and other generated files in a separate preview. Missing
118
+ energies, failures, cancellations, and diagnostic bundles remain explicit.
119
+ The input queue appends uploads as removable rows, keeps water visible on request,
120
+ and derives searchable per-option controls and click-to-open help from the selected
121
+ live CLI. Key and advanced controls remain collapsed until needed, while the
122
+ generated command line stays visible.
123
+ - Add `opt`/`all --reject-uphill/--no-reject-uphill` (default off). On `all`,
124
+ it is forwarded to the post-IRC endpoint re-optimization child only.
125
+ - Add `all --irc-step-size` so the end-to-end workflow can forward a smaller
126
+ EulerPC step to every post-TS IRC branch.
127
+ - Detect the molecular point group and external rotational symmetry number for
128
+ every `freq` structure and include the rotational `1/sigma` correction
129
+ automatically. `thermo.symmetry_number` remains an advanced YAML override.
130
+ - Add selectable UMA/ORB/MACE/AIMNet2 frame rescoring to `trj2fig`, with model,
131
+ precision, and provenance controls. Comment-energy mode remains
132
+ calculator-free, and rescoring is a pure MLIP calculation rather than ONIOM.
133
+ - Announce the first load of each ML backend model with
134
+ `[backend] Preparing MLIP model (...)...` and `[backend] Done.`.
135
+
136
+ ### Changed
137
+ - Remove the repeated status/image-count/backend footer from the Colab summary table; the compact run context owns status and artifact details.
138
+ - Remove the redundant `Path with N moving images.` startup line; tagged GSM sections and `String=...` records identify progress.
139
+ - Rename the prepared Toy inputs to `r_toy.pdb`, `p_toy.pdb`, and `p_toy.parm7` and label the notebook route as MEP mode.
140
+ - Enable repeated trajectory playback by default, select stitched `finished_irc_trj` for IRC profiles, and place the trajectory/energy view before result status and generated-file details in Colab.
141
+ - Show Hessian cache-reuse notices at `-v 2`; cache identity and rejection details remain at `-v 3`.
142
+ - Map the `hess` TS-optimizer preset to RS-P-RFO. Standalone `tsopt` retains RS-I-RFO through `--opt-mode rsirfo`.
143
+ - Remove the legacy `light` and `heavy` optimizer aliases; use `grad`/`hess` or the algorithm names exposed by each subcommand.
144
+ - Keep finite product cycle defaults: 100000 for ordinary optimization,
145
+ 300 for GSM/DMF, 125 for IRC, 100000 for ML/MM microiterations, and 100 for
146
+ DFT SCF. An explicit YAML `null` remains the uncapped engine value.
147
+ - Detect B-factor ML/MM layers by default and expose
148
+ `--detect-layer/--no-detect-layer` consistently across commands and `all`
149
+ child stages. Explicit `--model-pdb` / `--model-indices` membership remains
150
+ supported.
151
+ - Treat `--embedcharge` as an experimental, opt-in, computationally expensive
152
+ feature. MLIP/MM commands apply the xTB point-charge delta correction; `dft`
153
+ adds MM point charges directly to the PySCF QM Hamiltonian.
154
+ - Report every option's effective default. Options whose real default lives in
155
+ a config block are declared `None` so an explicit value stays distinguishable
156
+ from an omission; each now carries that default as a display string, so
157
+ `--help`, `--help-advanced`, the generated reference and the Colab Options
158
+ pane stop reading as unset. The Colab controls -- dropdowns included -- label
159
+ and select `default: <value>`, or `default: None` when there genuinely is none.
160
+ - Drop the notice that replaced the preparation panel for workflows that
161
+ extract internally; the panel is simply hidden.
162
+ - Name the custom-calculator entry point `--calc-file-func-name`. It names a
163
+ callable inside `--calc-file`, which the previous spelling `--calc-factory`
164
+ left unsaid. `--calc-factory` is removed rather than aliased.
165
+ - Head the stdout citation block `====== Citations & References ======` like
166
+ every other console section. `summary.log` keeps its numbered
167
+ `[6] Methods and citations`; the two share one renderer, so the log file's
168
+ section index was leaking into the console. Each method and its numbered
169
+ reference are now printed on separate lines.
170
+ - Reconcile a completed Colab run from the browser heartbeat when its final
171
+ widget update is lost, restoring the controls and Results without rerunning
172
+ the GUI cell.
173
+ - Keep one execution path in the Colab GUI. An unreachable async twin of the
174
+ whole run path (`_start_async_task`, `_do_validate_async`, `_do_run_guarded`,
175
+ `_do_run_async`, `_stream_async`, `_stop_async_process`,
176
+ `_validate_command_async`, `_async_task_done`) was never called, so a reader
177
+ had two implementations to reason about and only one ever ran. Removed with
178
+ three unreferenced one-liners; the notebook cell is 264 lines shorter.
179
+ - Correct the `baker` description in the `opt` skill page: convergence requires
180
+ all four force/step thresholds AND the energy change, not the published
181
+ force-and-either form. The contract gate pinned the stale sentence.
182
+ - Rename the Colab preparation button to "Extract ML region & use it", and put a
183
+ one-line hint where the panel used to vanish: `all` extracts internally and
184
+ `extract` is itself the extraction command, so the panel is hidden for both
185
+ and the hint names the route to a standalone `--model-pdb`.
186
+ - Make uphill-trial rejection opt-in for L-BFGS and RFO minimization. When
187
+ explicitly enabled, its energy tolerance is `1e-4` Hartree. TS optimization
188
+ always leaves the rejection disabled.
189
+ - Restore ordinary IRC's strict energy-rise stop: any positive one-step rise
190
+ stops that direction. `--irc-never-stop` bypasses this and the other physical
191
+ endpoint criteria until the cycle cap.
192
+ - Derive the ML-region charge in the Colab notebook from `--ligand-charge`
193
+ instead of requiring a confirmed `-q`; a ticked charge box is now an explicit
194
+ override.
195
+ - Install the DFT extra by default in the Colab notebook and verify plot export
196
+ by rendering a PNG. A first run takes several minutes and needs a GPU
197
+ runtime.
198
+ - Label thermochemistry as `E + G_corr = G`, force uphill rejection off for
199
+ transition-state optimization, and keep its toggle limited to minimum and
200
+ post-IRC endpoint optimization.
201
+ - Apply the `baker` preset as a deliberately tightened variant of the published
202
+ criterion (Bakken and Helgaker, J. Chem. Phys. 117, 9160 (2002)): maximum
203
+ force, RMS force, maximum step, RMS step and the energy change must all hold,
204
+ including the final check of a retained lower-energy geometry at the
205
+ uphill-rejection trust floor. A zero-length step settles the energy criterion,
206
+ because the geometry cannot move.
207
+ - Stop forcing one reparametrization pass per growing-string cycle. An image
208
+ already inside the parametrization tolerance is no longer displaced, and
209
+ coincident parameter densities are rejected instead of divided.
210
+ - Pin backend setup recipes to the official PyTorch 2.8 wheel matrix, install
211
+ dedicated-environment MACE only after removing `fairchem-core`, and make HPC
212
+ templates fail fast unless the `hessian_ff` JIT compiler prerequisites are
213
+ present.
214
+ - Remove the unused internal `AllContext` parameter mirror and break the product
215
+ import cycles (`core.utils`↔`extract`, `freq`↔`opt`) by relocating the shared
216
+ charge/spin preparation and layer helpers; the relocation itself makes no CLI or
217
+ JSON contract change (see the `sp` ML-region resolution change below, which does
218
+ move numbers).
219
+ - Reject explicit analytical Hessians with `workers > 1`.
220
+ - Project only rigid modes that are an actual null space of the frozen system for
221
+ PHVA, IRC, Dimer, and TS validation, and record the effective mode. The former
222
+ active-fragment projection could hide a real imaginary mode, so `n_imag`, ZPE
223
+ and ΔG‡ move on frozen-boundary systems. The superseded public
224
+ `--tr-projection` option and `legacy-active` treatment are removed.
225
+ - Require `n_imag = 1` for TS success, preserve rejected optimizer state, and
226
+ report resolved backend/model/precision and the highest common rate-limiting method.
227
+ - Keep the default Hessian TS search on standard restricted-step root
228
+ following. Trial mode-loss rejection, intermediate eigenvalue-structure
229
+ gating, automatic saddle recovery, and automatic displaced multistarts are
230
+ disabled by default. Exact PHVA remains the terminal authority and requires
231
+ `n_imag = 1`; explicit `--flatten` remains available for extra modes.
232
+ - Double the default segment path resolution (`max_nodes_segment` 10 → 20), which
233
+ changes the MEP, its highest-energy image, and therefore the reported barrier.
234
+ - Unify the residue/ion/water catalog so charge inference recognizes the same
235
+ ions as element inference; charge summaries now include monatomic ions that
236
+ were previously unrecognized.
237
+ - Populate missing element columns in the topology-matched PDB exported by
238
+ `mm-parm`, preserving LEaP atom records and order; extraction recovery text
239
+ now names the non-destructive output or explicit `--inplace` route.
240
+ - Keep finite-difference Hessian assembly, low-rank Bofill updates, and mass
241
+ scaling device-resident on GPU runs, avoiding per-step host round-trips.
242
+ - Make the `tsopt` Hessian-radius contract explicit: an omitted
243
+ `--radius-hessian` includes the movable MM atoms required by the default
244
+ `partial` final-frequency basis, while an explicit narrower cutoff is rejected
245
+ before optimization unless `--active-dof-mode` requests only evaluated atoms.
246
+ - Make periodic IRC HDF5 trajectory checkpointing opt-in
247
+ (`irc.dump_every: null` by default). Enabled checkpoints contain coordinates,
248
+ energies, and gradients, but never a dense Hessian.
249
+ - Honor YAML `opt.thresh`, the `dft` method, and `--func-basis` in `all`: these are
250
+ forwarded to the post-MEP TSOPT and R/P endpoint-optimization children only when set
251
+ explicitly, so a YAML value is no longer clobbered by a hardcoded default. Endpoint
252
+ geometry/energy and DFT numbers change when configured.
253
+ - Build the align/refine and `--no-tsopt` HEI-probe calculators from the resolved
254
+ calculator template, so `uma_precision`/`uma_model`/`workers`/`cutoff` and
255
+ supported YAML `calc` settings now reach them; aligned geometry and HEI
256
+ energy/barrier move when these are set.
257
+ - Skip the BFGS Hessian update when the curvature `s·y ≤ 0` (was applied), and force
258
+ `use_active=False` on internal-coordinate frequency analysis, fixing a partial-Hessian
259
+ index mismatch reachable via `--coord-type` with frozen atoms.
260
+ - Decide IRC endpoint minimality on an exact Cartesian Hessian under the opt-in
261
+ `--irc-pos-def`, replacing the integrator's quasi-Newton Hessian.
262
+ - Mask the harmonic-restraint Hessian on a separate `hessian_constrained_atoms` set
263
+ from the forces, changing frequencies where restraints and frozen atoms interact.
264
+ - Seed `scan2d`/`scan3d` grid points only from explicitly converged relaxations,
265
+ reference relative and minimum energies to seed-eligible points only, and honor YAML
266
+ `opt.thresh`/`opt.max_cycles` (previously clobbered by the CLI default); relative
267
+ energies shift when any grid point fails.
268
+ - Isolate Direct-Max-Flux configuration per invocation (`fresh_dmf_config` deep-copy).
269
+ - Load `path-opt --ref-pdb` geometry from the original unrounded `.xyz` (was: a PDB
270
+ rounded to 0.001 Å, then reloaded), removing a coordinate round-trip.
271
+ - Resolve the `sp` ML region through the shared validator instead of a bare
272
+ `except`-and-use-the-whole-system fallback: a resolution failure now raises rather
273
+ than silently treating the full input as the ML layer (which changed the ONIOM energy).
274
+ - Take `dft` QM-region atom indices from the one-based file ordinal (was the deposited
275
+ PDB serial and `name[0]` element), so the QM region and link-hydrogen pairs shift on
276
+ gapped serials or two-letter elements.
277
+ - Drop the ±1 terminal-charge correction for Amber-capped C-/N-terminal residues in the
278
+ `extract` charge summary, changing the net protein/active-site charge for capped termini.
279
+ - Attach the built wheel and source distribution to each GitHub Release before
280
+ trusted PyPI publication.
281
+ - `result.json`/`summary.json` gained additive field families — a `run_id` (from
282
+ `MLMM_RUN_ID`; a conflicting id raises `RunIdentityError`); `execution_status` and
283
+ `scientific_status` with reasons and item/expected/observed ids; per-stage and
284
+ per-point outcomes; per-segment `converged`, `irc`, `endpoint_opt`, `ts_imag`, and
285
+ `dft_status`; scan `energy_reference`/`n_points_usable`; a serialized
286
+ `microiteration` object in `opt`/`tsopt` output; a `thermo_policy` block in the
287
+ frequency YAML/`result.json`; and resolved provenance. These are additive for
288
+ consumers that tolerate unknown fields. `scientific_status` also participates in
289
+ usability/promotion decisions, not only provenance. Aggregate success requires
290
+ every applicable producer convergence signal; direct-TS segments do not invent
291
+ an MEP gate.
292
+ - `key_output_files` now lists only the artifacts claimed by the current
293
+ invocation's run manifest rather than files discovered under the output tree, so a
294
+ reused `-o/--out-dir` no longer reports stale files from an earlier run.
295
+ - Frequency JSON/YAML records `symmetry_number` and
296
+ `symmetry_number_source`; `all` copies complete child provenance into each
297
+ post-segment's `thermo_symmetry` map for R/TS/P. IRC result JSON records
298
+ `electronic_state_verified` for a reused frequency Hessian.
299
+ - The MCP tool-return envelope moved from `schema_version` `1.0` to `1.1`, adding a
300
+ per-invocation `run_id`, a `summary_run_mismatch` status, and a run-id byte check
301
+ (distinct from the summary `schema_version: "2.0"`).
302
+ - Commit `result.json`/`summary.json` and converted structures by staged atomic
303
+ replace and raise on a write failure that was previously swallowed; the mmCIF/PDB
304
+ bridge hard-fails on out-of-range or non-finite coordinates and unresolved elements
305
+ rather than emitting a corrupted fixed-column record. Valid-input output bytes are
306
+ unchanged.
307
+
308
+ ### Fixed
309
+ - Route the first Colab `Load results` action through the native browser bridge.
310
+ - Organize Colab Results into `Energy profile & Trajectory` and
311
+ `Imaginary frequency`, with a contextual `View` or `Mode` selector.
312
+ - Install declared wheel dependencies during release validation and allow a
313
+ published release tag to be retried safely after a workflow failure.
314
+ - Restore the intended Mol* expanded-view defaults, clear the manual-command
315
+ notice on Rebuild, and keep combined-IRC endpoint labels below their markers.
316
+ - Initialize generated Colab numeric controls from each option's effective CLI
317
+ default instead of its Click range minimum, and omit the flag again when the
318
+ control is restored to that default.
319
+ - Keep energy-weighted GSM parametrization finite across minimum-energy and flat intervals while rejecting invalid energies.
320
+ - Correct CLI/YAML/default documentation, runnable output recipes, TS-only guidance, and canonical toggle names.
321
+ - Pin Torch 2.8.0 in every minimal no-deps CI lane and match the primary citation title.
322
+ - Bound Plotly/Kaleido static-image export to ten minutes in an isolated
323
+ renderer process so a stalled Chrome startup or shutdown cannot block a
324
+ scientific workflow indefinitely.
325
+ - Disable polynomial line search inside the Hessian Dimer and reject attempts
326
+ to enable it, because the effective Dimer force is not conjugate to the
327
+ reported physical energy.
328
+ - Keep IRC branches diagnostic when energy-based displacement cannot find a
329
+ downhill departure; later small-gradient termination can no longer certify
330
+ or cache such a branch.
331
+ - Preserve distance-derived frozen atoms when a calculator is attached, and
332
+ constrain default Hessian targets to atoms inside `movable_cutoff`.
333
+ - Propagate requested endpoint-preoptimization convergence into standalone and
334
+ `all` path scientific status while retaining diagnostic path continuation.
335
+ - Make `all --dry-run` validate extraction/scan syntax and the ML-region
336
+ charge/multiplicity state before reporting success.
337
+ - Forward the resolved embedding state and cutoff together when YAML enables
338
+ embedding and CLI overrides only the cutoff.
339
+ - Add `all --freeze-atoms`, merge it with YAML `geom.freeze_atoms` and the
340
+ Frozen-MM layer across every child stage, expose it in Colab `all`, and keep
341
+ the unsupported selection hidden only for `dft`.
342
+ - Honor an explicit `all --opt-mode` for both TSOPT and post-IRC endpoint
343
+ optimization when `--opt-mode-post` is omitted.
344
+ - Separate the Colab workspace-path and example actions, restore `opt --dump`
345
+ trajectories from the `[command]` entry in `run.log`, and persist `run.log`
346
+ for ordinary output-directory CLI runs.
347
+ - Keep `dft --embedcharge` as direct PySCF electrostatic embedding and do not
348
+ add the optional MLIP/xTB point-charge correction to the embedded DFT energy.
349
+ - Number TS candidates, validated TS structures, and intermediates across linked
350
+ Colab MEP/IRC trajectories, and label combined IRC endpoints by direction.
351
+ - Label the Colab advanced-control disclosure as `Show All options`.
352
+ - Load an existing Colab-workspace file through the validated input queue and
353
+ restore `./result_all/` whenever a built-in example is loaded.
354
+ - Show workflow warnings from `scientific_status_reasons` above Colab result tables while
355
+ keeping Run details limited to execution metadata.
356
+ - Keep the complete MEP summary visible at default pipeline verbosity,
357
+ including no-change results and every segment's barrier and reaction energy.
358
+ - Reconcile reported frequency modes against active DOF, rigid modes, and the
359
+ near-zero window, and render reproduced commands shell-safely.
360
+ - Give generated text and log previews a bounded, visible scroll region in the
361
+ Colab Results panel.
362
+ - Report `all` summary modes as `MEP` / `Scan` / `TS-only` and show absolute
363
+ root and internal path-module output directories in `summary.log`.
364
+ - Match the interactive Colab energy-level segment and connector proportions
365
+ to the CLI diagram and shrink overlong level annotations to fit their bars.
366
+ - Keep the scan2d/scan3d starting reference as the `-1` row in `surface.csv`
367
+ while excluding it from energy baselines, interpolation, and plots.
368
+ - Compact the Colab run/output and Results status surfaces and enlarge
369
+ scan-grid pick markers while retaining the scan3d color scale.
370
+ - Offset the scan2d coloured base plane above the z-axis floor to prevent
371
+ coplanar rendering flicker.
372
+ - Activate the Colab Results tab before rendering completed results, keep the
373
+ result selectors reusable, and reserve plot space for MEP/IRC labels.
374
+ - Keep cancelled Colab runs on the active tab and out of completed-result views.
375
+ - Report the exact MLIP model and UMA task in logs, JSON, and Colab Results, and
376
+ include model-specific UMA, Orb-v3, MACE, and OMol25 citations.
377
+ - Use pause semantics for Colab trajectory playback.
378
+ - Restore the reviewed Colab selection UI: persistent removable chips, concise
379
+ workflow labels and center summaries, explicit frozen-atom completion, and a
380
+ compact staged-scan editor. The upload panel now treats parm7 as optional for
381
+ `all`, which generates it when omitted.
382
+ - Fail closed on unknown execution, failed segment DFT, and non-converged path bridges, and clear stale scientific-status reasons.
383
+ - Preserve internal result identity while presenting aggregate structure-linked MEP/IRC profiles, reset prior result state before reload, and recover only declared or current-run artifacts.
384
+ - Preserve the TSOPT stop cause, restrict the output tree to current-run artifacts, and report the resolved frequency zero cutoff in `summary.log`.
385
+ - Label subtractive energies as ML/MM in human output and include compact three-layer counts.
386
+ - Keep optimizer verbosity monotonic and show Hessian cache reuse at `-v 2` without exposing raw DFT child diagnostics below `-v 3`.
387
+ - Translate internal partial-result codes into concise, actionable warnings in `summary.log` and final stdout.
388
+ - Preserve the blank line before the first MLIP model-load announcement even when an stderr warning immediately precedes it.
389
+ - Preserve the MM micro-optimizer until its non-convergence diagnostics have been collected, preventing a post-macro microiteration failure from raising `UnboundLocalError`.
390
+ - Record terminal PHVA as `skipped`, rather than `unavailable`, when a non-converged Hessian TS optimizer does not authorize the analysis.
391
+ - Print a section heading before each tagged recursive GSM segment and before standalone Growing String optimization.
392
+ - Bound the complete minimum-RFO line-search/GDIIS displacement by scaling it
393
+ to the trust radius while preserving its accelerated direction.
394
+ - Enforce the same trust radius on hard-case Newton and reference RFO steps,
395
+ preventing near-degenerate solvers from returning oversized displacements.
396
+ - Keep an explicit finite cycle cap when resuming a checkpoint written by an
397
+ uncapped optimizer.
398
+ - Retain endpoint-optimization diagnostics when `--dump` is enabled.
399
+ - Gate MEP `--ref-mode` handoff to Hessian TS optimizers; Dimer records the handoff as not applicable instead of receiving an option it rejects.
400
+ - Preserve a non-converged or stalled TS final structure and stop before terminal PHVA.
401
+ - Use the configurable `freq.zero_cutoff_cm` value for standalone frequency analysis, flattening, and TS saddle classification.
402
+ - Restrict reference-aligned reaction-mode selection to negative exact-PHVA modes and validate the selected frequency before IRC. Invalid or missing selections use an explicit lowest-imaginary root-0 fallback with unverified reaction identity.
403
+ - Keep `--skip-final-freq` artifact-preserving but stop composite `all` before IRC because the negative reaction direction was not validated.
404
+ - Separate numerical optimization status from saddle order. A converged higher-order stationary point is not a first-order TS, but `all` may perform warning-labelled diagnostic IRC when a valid negative root exists.
405
+ - Synchronize EN/JA docs, skills, live help, generated command references, and checked-in contract tests with the current behavior.
406
+ - Remove the replaced path-tangent/single-mode helpers and one unreferenced mass-weighted-frequency wrapper; larger workflow and Notebook refactors remain deferred.
407
+ - Keep Hessian status and timing lines adjacent to the surrounding optimizer
408
+ cycle rows instead of inserting blank lines before and after each Hessian
409
+ evaluation.
410
+ - Run exact PHVA after every configured Hessian-TS convergence criterion passes
411
+ for the proposed step, or after an opt-in energy-plateau stop. A run that
412
+ instead exhausts `max_cycles` skips final PHVA and mode export, reports both
413
+ imaginary-mode fields as `null`, and remains `not_converged`. A plateau runs
414
+ PHVA but retains `stalled`.
415
+ - Keep RS-P-RFO line searches off by default while honoring explicit YAML
416
+ values. RS-I-RFO and TRIM discard these unused kwargs, and Hessian TS searches
417
+ require exactly one root.
418
+ - Apply the shared `opt` block to the `tsopt` macro optimizer. `rms_force`,
419
+ `rms_force_only`, `max_force_only`, `force_only`, `overachieve_factor`,
420
+ `min_step_norm`, `assert_min_step`, `converge_to_geom_rms_thresh`,
421
+ `check_eigval_structure` and `line_search` were read from `opt` and then
422
+ dropped, so a YAML convergence control was silently ignored by TS
423
+ optimization while `opt` honored it. Both the microiteration and ordinary
424
+ paths now build their macro kwargs through the same helper, and only values
425
+ you changed are forwarded, so `rsirfo.*` stays authoritative for untouched
426
+ keys and a default configuration is byte-identical to before.
427
+ - Read the `lbfgs` section in `tsopt`, so the microiteration MM relaxation's
428
+ L-BFGS controls (`keep_last`, `max_step`, `line_search`, ...) are reachable
429
+ from YAML. They were settable from neither YAML nor CLI.
430
+ - Report why a micro relaxation stopped. The micro runs with its stdout
431
+ redirected, so the log said only `status=not_converged`; it now names the
432
+ executed cycles against the bound, the last max force and step, and — when
433
+ the bound is what ended it — `microiter.micro_max_cycles` as the setting to
434
+ raise.
435
+ - Raise `microiter.micro_max_cycles` from 10000 to 100000 so the MM relaxation
436
+ is bounded by its convergence criterion rather than by the cycle cap. The
437
+ micro step is held to the same preset as the macro step, and under `baker`
438
+ the few transient relaxations that follow a large rearrangement need more
439
+ than 10000 L-BFGS cycles: measured on the release-smoke ML/MM TS lane, 791
440
+ relaxations with a median of 56 cycles and three transients at 16815 / 13167
441
+ / 10624. At the old cap those three reported `not_converged`, whose
442
+ fail-closed handling aborted the TS search after 45 of 791 macro steps and
443
+ left four imaginary modes; with the new bound every relaxation converges and
444
+ the search reaches a certified first-order saddle. No convergence threshold
445
+ changed, and the 788 relaxations that already converged are unaffected.
446
+ - Apply the whole shared `opt` block to the macro step of an `opt --microiter`
447
+ run. It previously forwarded only `thresh`, `max_cycles`, `dump` and
448
+ `out_dir`, so YAML convergence controls (`rms_force`, `rms_force_only`,
449
+ `max_force_only`, `force_only`, `overachieve_factor`, `min_step_norm`,
450
+ `assert_min_step`, `converge_to_geom_rms_thresh`, `check_eigval_structure`)
451
+ and `line_search` / `print_every` were silently ignored in the default `hess`
452
+ path while `--no-microiter` honored them. The macro step now uses the same
453
+ merge rule as an ordinary run, so only values you actually changed are passed
454
+ and a default configuration reaches the optimizer unchanged.
455
+ - Reject single-class B-factor metadata such as an all-zero PDB as an ML/MM
456
+ layer partition instead of silently treating the full system as ML.
457
+ - Preserve valid B-factor movable/frozen MM layers when `sp` uses an explicit
458
+ `--model-pdb` or `--model-indices`; invalid layer metadata no longer remains enabled.
459
+ - Honor `dmf.ipopt_options.dual_inf_tol` when it is set in YAML instead of
460
+ replacing it with a fixed preset.
461
+ - Reject dependent Amber virtual sites before allocating the `hessian_ff`
462
+ backend, and direct users to OpenMM or a three-point-water topology.
463
+ - Apply the inclusive B-factor tolerance consistently when reading ML,
464
+ movable-MM, and frozen-MM layers; leave missing or malformed B-factors
465
+ unassigned.
466
+ - Resolve the legacy `MACE-OFF23_small`, `_medium`, and `_large` aliases to the
467
+ corresponding upstream MACE-OFF model sizes.
468
+ - Preserve selected-root diagnostics in non-Cartesian coordinates and convert
469
+ torch eigenvectors explicitly for reference-mode overlap checks.
470
+ - Apply the tightened `baker` criterion on every evaluable cycle, including the
471
+ first retained geometry.
472
+ - Convert Cartesian coordinates from Bohr to Angstrom before the
473
+ thermochemistry hand-off; the moments of inertia, rotational entropy and
474
+ absolute Gibbs energies were computed from Bohr values in an Angstrom contract.
475
+ - Require every internal-coordinate component to satisfy the back-transformation
476
+ tolerance before the iteration is accepted; a single converged component used
477
+ to end it.
478
+ - Apply the thermochemistry policy's `zpe_scale` (passed on as
479
+ `zpe_scale_factor`) exactly once to the reported zero-point energy and to the
480
+ vibrational internal energy. Values other than the default `1.0` were scaled
481
+ twice.
482
+ - Collect invalid primitive indices from the actual dihedral and bend index
483
+ lists instead of assuming contiguous ranges, keeping both sets.
484
+ - Correct the diagnostic QRRHO free-rotor partition function (frequency instead
485
+ of wavenumber, plus the missing factor of pi); reported QRRHO entropy and Gibbs
486
+ energy come from the Grimme interpolation and are unchanged.
487
+ - Evaluate the vibrational heat capacity through a series expansion for small
488
+ exponents and `expm1` elsewhere, removing the cancellation error near zero.
489
+ - Accept equivalent leading-digit PDB and trailing-digit Amber hydrogen names
490
+ while preserving strict atom-order checks for non-hydrogen atoms.
491
+ - Route `dft` through `MLMMCore` for topology preparation, MM calculators, and
492
+ subtractive recombination. DFT now replaces only the high-level model energy
493
+ and requests no forces; there is no separate model-parm7 builder or MM path.
494
+ - Keep the IRC running when the EulerPC corrector oscillates. The corrector
495
+ descends the two-point interpolated surface rather than the real potential,
496
+ so a reversal there is an interpolation artifact; it now warns and keeps the
497
+ last non-oscillating point instead of aborting the whole run.
498
+ - Stop `scan2d` after writing `surface.csv` with a clear diagnostic when fewer
499
+ than three non-collinear converged grid points remain, instead of passing an
500
+ underdetermined data set to SciPy's RBF interpolator.
501
+ - Read the bundled pysisyphus optimizer's legacy comma-delimited Hartree XYZ comments in
502
+ `trj2fig` and other strict trajectory consumers without treating unrelated
503
+ numeric comments as energies.
504
+ - Expose the shared `--allow-charge-mult-mismatch` escape hatch on `tsopt`,
505
+ matching the other ML/MM compute commands that run the same ML-region
506
+ electron-parity validation.
507
+ - Require `scan2d --scan-lists` during Click parsing, reject unknown
508
+ `energy-diagram` options instead of silently accepting them, and keep
509
+ detail-only pipeline messages out of verbosity level 1.
510
+ - Reject unknown options and orphan arguments in legacy grouped-value commands,
511
+ while accepting grouped or repeated path-search inputs and topology references
512
+ consistently in shell and in-process Click invocations.
513
+ - Reject `all --scan-lists` with multiple inputs instead of silently selecting
514
+ the endpoint-path route and ignoring the staged scan.
515
+ - Honor `tsopt --dump` when the initial microiteration MM relaxation fails closed
516
+ before the first macro step; the combined trajectory now contains that executed
517
+ micro relaxation instead of being absent.
518
+ - Write `thermoanalysis.yaml` from the `all` pipeline's `freq` stages by default, so
519
+ `--thermo` actually yields thermochemistry. The child `freq` inherited its own
520
+ `--dump` default (off), so `all --thermo` computed and printed the thermochemistry
521
+ but never persisted the file the Gibbs assembly reads — every run reported
522
+ `thermochemistry result is missing` and produced no Gibbs diagram. An explicit
523
+ `--no-dump` still suppresses the file.
524
+ - Keep Hessian-Dimer orientations and off-center images on the frozen Cartesian
525
+ constraint manifold, refreshing constraint-compatible rigid null modes at
526
+ each central image.
527
+ - Keep scan energies on the unbiased PES and prevent stale Hessian reuse across
528
+ TS, IRC, frequency, and endpoint optimization.
529
+ - Roll back rejected RFO/L-BFGS state, recover TS searches from `n_imag=0`, and
530
+ keep path-guided flattening explicitly opt-in and mode-safe.
531
+ - Keep interpolated/GDIIS RFO displacements on the reference step's tensor
532
+ device, preventing a CUDA-to-NumPy conversion failure during Hessian scans.
533
+ - Report a microiterated optimization's real terminal convergence. The macro
534
+ loop's verdict was computed and then dropped, so a converged TS — including one
535
+ whose exact-PHVA validation found a single imaginary mode — was written as
536
+ `not_converged`, and `all` refused to start its IRC.
537
+ - Release the transition-state probe calculator before returning, so the IRC
538
+ phase's leased core is the only heavy ML/MM core alive across the handoff.
539
+ - Run the `opt --flatten` loop when the optimizer stalls on an energy plateau.
540
+ The loop rebuilds the Hessian and displaces along the remaining imaginary
541
+ modes, so a stall is exactly when it is wanted; only a flatten *retry* that
542
+ stalls again stops the loop.
543
+ - Keep optimizing when `opt.dump_restart` is set on an optimizer class whose
544
+ restart state is not declared: the unsupported checkpoint is refused once,
545
+ further dumping is disabled, and the run continues instead of aborting.
546
+ - Resolve the rigid-mode default from one place, so the `freq`/IRC/TS-optimizer
547
+ fallbacks can no longer drift apart from the documented default.
548
+ - Correct IRC endpoint labels, `all` worker propagation, YAML custom-factory
549
+ provenance, and CIF publication at pipeline root and segment level.
550
+ - Keep custom `--calc-file` selection consistent across child and in-process
551
+ `all` stages when `--backend` is also supplied.
552
+ - Select one coherent altLoc per residue and identity-check frequency-to-IRC
553
+ Hessian handoff against geometry, atom order, and active-DOF basis.
554
+ - Honor EulerPC's normalized IRC filename prefix in conversion, endpoint checks,
555
+ and JSON; ship the third-party/OpenMM CMAP notices in built distributions.
556
+ - Apply the ZPE scale factor exactly once in thermochemistry, so a non-unity
557
+ factor no longer enters the reported ZPE and `U`/`H`/`G` quadratically.
558
+ - Isolate each Dimer's random state from the process-global NumPy RNG, and raise
559
+ a clear error for an invalid rotation method.
560
+ - Fingerprint the native `hessian_ff` build and runtime identity before loading a
561
+ prebuilt extension, refusing a stale or host-incompatible binary.
562
+ - Stop reporting an electronic energy as a Gibbs free energy: `all` builds the
563
+ per-segment MLIP and DFT//MLIP/MM Gibbs diagrams only when every state's frequency
564
+ free energy (and DFT thermal correction) is finite, otherwise it skips the diagram
565
+ and warns, instead of substituting the MLIP/DFT electronic energy or a `0.0`
566
+ thermal correction. Reported ΔG changes wherever a thermochemistry value was
567
+ missing.
568
+ - Compare the full per-atom signature when guarding `extract`'s multi-input atom
569
+ order, so a swapped mid-chain atom now raises instead of passing a first/last-atom
570
+ spot check.
571
+
572
+ ### Documentation
573
+ - Rebuild the CLI references and skills for the current ML/MM input contract,
574
+ validate required parm7 and XYZ topology references in runnable skill examples,
575
+ and document JSON 2.0 truth/provenance plus the Colab workflow.
576
+ - Clarify that recognized monatomic ions use the internal charge table and that
577
+ `-l` entries such as `MG:3` are ignored rather than overriding that table.
578
+
579
+ ## [0.3.2] — 2026-07-10
580
+
581
+ ### Changed
582
+ - **`--precision` now defaults per backend instead of globally to fp32: ORB runs fp64
583
+ when no precision is given (MACE already did), UMA keeps fp32.** ORB's fp32 is the
584
+ reduced `float32-high` (TF32) matmul mode. Pass `--precision fp32`
585
+ explicitly to select the reduced-precision screening configuration.
586
+ - **Behavior change (default): the default UMA model is now `uma-s-1p2`** (was
587
+ `uma-s-1p1`). Other models (`uma-s-1p1`,
588
+ `uma-m-1p1`, MACE-OMOL, Orb-v3-omol) remain selectable via `-b` / `--backend-model` / config.
589
+ - **Centralized the default UMA model** in a single constant `DEFAULT_UMA_MODEL`
590
+ (`mlmm/core/defaults.py`); the `uma-s-1p1` defaults previously hardcoded across backends and
591
+ `io/trj2fig.py` now all reference it. CLI help and docs were updated for consistency.
592
+
593
+ ### Added
594
+ - **`--dry-run` on `scan2d` and `scan3d`.** Validates options and prints the execution
595
+ plan without running the scan, pairing with `scan` and pdb2reaction.
596
+ - **`--workers` / `--workers-per-node` on every MLIP subcommand** (`sp`, `opt`, `tsopt`, `freq`,
597
+ `irc`, `scan` / `scan2d` / `scan3d`, `path-opt`, `path-search`, `all`), pairing with pdb2reaction.
598
+ `--workers > 1` routes the UMA backend through fairchem's `ParallelMLIPPredictUnit` (needs
599
+ `fairchem-core[extras]`); the parallel predictor exposes no autograd model, so analytical Hessians
600
+ are unavailable and an `Analytical` request is rejected with an error. The default
601
+ `--workers 1` keeps the in-process predictor and is byte-for-byte the previous behavior.
602
+ - **Microiteration now works with every Hessian TS optimizer.** `--microiter` (default on)
603
+ previously engaged only with RS-I-RFO (`--opt-mode hess` / `rsirfo`); it now also drives the
604
+ RS-P-RFO (`--opt-mode rsprfo`) and TRIM (`--opt-mode trim`) macro step, alternating a 1-step
605
+ macro TS move with MM-only L-BFGS relaxation. All three are `TSHessianOptimizer` subclasses that
606
+ share the `optimize()`/`prepare_opt()`/Bofill-update contract the macro loop drives. The default
607
+ TS optimizer (RS-I-RFO) is unchanged.
608
+
609
+ ### Fixed
610
+ - **MACE backend could not load its own default model.** The `MACE-OMOL-0` default was
611
+ routed to `mace_off()`, which treats any non-preset, non-URL string as a local file
612
+ path (raising `FileNotFoundError`). It now uses the dedicated `mace_omol` factory.
613
+ - **An analytical-Hessian request the backend cannot honour now raises instead of
614
+ degrading silently.** A backend build without an autograd model rejects
615
+ `--hessian-calc-mode analytical` with an error naming `FiniteDifference`, rather than
616
+ quietly changing the Hessian method.
617
+ - **`make_is_param_explicit` logs on a failed parameter-source query.** Behaviour is
618
+ unchanged (still treats the param as not explicit); the debug log makes a genuine
619
+ typo diagnosable, matching pdb2reaction's `cli_param_overridden`.
620
+ - **A `--config` YAML `calc.precision` is now dispatched on every subcommand.** The
621
+ per-subcommand call sites guarded the dispatch helper on the CLI value being
622
+ non-`None`, so a YAML-only precision was never applied (ORB silently ran at its
623
+ default precision); `all` was already correct. The guard is dropped on all 9
624
+ subcommands, re-enabling the helper's YAML path and its invalid-token / aimnet2-fp64
625
+ validators.
626
+ - **`--backend-model` supplied via a `--config` YAML is now honoured** on every
627
+ subcommand (same guard as above; only `all` was correct).
628
+ - **The deterministic-compute comparison now reuses one MM topology.** Topology
629
+ generation is outside `--deterministic`; pass a fixed `--parm` when comparing
630
+ exact artifacts.
631
+ - **Run summary recorded the default UMA model and `mlip_backend: "unknown"`, ignoring `--backend-model` / `-b`.**
632
+ The `all` workflow's summary payload never populated `uma_model`, so `summary.py` fell back to the default
633
+ `MLMM_CALC_KW` model (now `uma-s-1p2`): a run launched with e.g. `--backend-model uma-s-1p1` recorded
634
+ `UMA model: uma-s-1p2` in `summary.log` and `"uma_model": null` / `"mlip_backend": "unknown"` in `summary.json`.
635
+ The computation honored `--backend-model`; the summary now records the
636
+ resolved model (`--backend-model` or `DEFAULT_UMA_MODEL`) and backend (`-b`
637
+ or the `uma` default).
638
+
639
+ ## [0.3.1] — 2026-07-05
640
+
641
+ ### Fixed
642
+ - **Charge/spin were silently dropped on the UMA MLIP path.** `AtomicData.from_ase` was
643
+ called without `r_data_keys`, so fairchem-core ≥2.x ran the UMA/ORB/MACE backends at
644
+ charge=0/spin=0 regardless of the requested charge/multiplicity. Now passes
645
+ `r_data_keys=["spin","charge"]`, and `atoms.info["spin"]` is the spin multiplicity (2S+1)
646
+ for the OMol backends — it previously used the unpaired-electron count (`mult-1`), which sent
647
+ closed-shell singlets to the null spin token.
648
+ - `mm-parm` used a local copy of the residue/ion charge tables that had drifted from
649
+ `mlmm.core.residue_data` (it lacked the phosphorylated S1P/T1P/Y1P and several His tautomers,
650
+ assigning them charge 0, and used a different disulfide cutoff). It now imports the canonical
651
+ tables.
652
+
653
+ ### Changed
654
+ - Documentation corrections to match the code (`--precision` per-backend default; `pdbfixer`
655
+ is optional; backend/architecture notes).
656
+
657
+ ## [0.3.0] — 2026-06-28
658
+
659
+ ### Changed
660
+ - **Behavior change (default):** `all --refine-path/--no-refine-path` now
661
+ defaults to `--no-refine-path`. The `all` pipeline's MEP stage runs a
662
+ single-pass `path-opt` by default; pass `--refine-path` to run the recursive
663
+ `path-search` (automatic multi-step bond-change segmentation), which was the
664
+ previous default. The default MEP work directory is now `_work/path_opt/`
665
+ (was `_work/path_search/`). The standalone `path-search` subcommand is
666
+ unchanged. Docs/skills updated throughout.
667
+ - `--dft-func-basis` is now surfaced in the primary `mlmm <subcmd> --help`
668
+ (previously only under `--help-advanced`), so the DFT//MLIP/MM functional/basis
669
+ is discoverable without the advanced listing.
670
+ - Standalone `path-opt` / `path-search` now default to `--preopt` (each MEP
671
+ endpoint is pre-optimized before the search); the previous default was
672
+ `--no-preopt`. The `all` pipeline forwards the flag explicitly and is
673
+ unaffected.
674
+ - `--precision fp64` now also forces the Hessian to fp64 (`H_double`) so the
675
+ optimiser / eigen linear algebra cannot silently run in a lower precision
676
+ than the model; a config that set `H_double=False` under fp64 is overridden
677
+ with a warning.
678
+ - AIMNet2 now rejects both `--precision fp64` (its model inputs are cast to
679
+ float32 upstream) and `--deterministic` (its forces come from a custom CUDA
680
+ kernel outside torch's deterministic-algorithms control), with clear errors
681
+ instead of running misleadingly.
682
+ - The `all`-pipeline determinism comparison uses a fixed smoke stack and input;
683
+ exact repeatability for other stacks requires separate verification.
684
+ - `--help` of `mlmm` now groups subcommands under semantic sections
685
+ ("Pipelines" / "Pipeline stages" / "Inputs & topology" / "Analysis")
686
+ in a configurable, deterministic order; subcommands not listed in any
687
+ section fall through to a trailing "Other" bucket so we never hide an
688
+ entry silently.
689
+ - CLI exception renderer appends `Try 'mlmm <subcmd> -h' for help.` to
690
+ every user-input-style error so first-time users see a recovery path,
691
+ and routes the full traceback through `logging.getLogger(...).exception`
692
+ so log scrapers / `-v` users get the structured record alongside the
693
+ terminal echo.
694
+ - `_calc_energy` deduplicated into `mlmm.core.calc_eval`; both
695
+ `workflows/opt.py` and `workflows/tsopt.py` now re-export the helper.
696
+ - Repo-wide ruff `F401` sweep: removed 66 unused imports and two
697
+ orphaned helpers (`_get_masses`, `_build_tr_basis`) in
698
+ `mlmm/io/hessian_calc.py`.
699
+ - 4 mis-typed parameters annotated `Optional[...]` where the default is
700
+ `None` (`MLMMCore.__init__: input_pdb / real_parm7 / model_pdb` and
701
+ `build_model_pdb_from_bfactors: tolerance`).
702
+ - `docs/cli-conventions.md` now spells out the four permanent boolean
703
+ forms (`--flag` / `--no-flag` / `--flag True/yes/1/on` /
704
+ `--flag False/no/0/off`) and adds a "Contributing a new bool flag"
705
+ section pointing at the `add_*_option` factory + `_COMMAND_BOOL_*`
706
+ registries.
707
+
708
+ ### Fixed
709
+ - Bond-change detection (`domain/bond_changes.compare_structures`) is now
710
+ row-chunked instead of building dense N×N distance matrices, removing a CUDA
711
+ out-of-memory failure on large solvated clusters (~20k+ atoms) during
712
+ `path-search` / `scan` kink detection on 16–24 GB GPUs.
713
+ - mlmm-toolkit's default MM backend now declares `ninja` as a dependency. Its C++ kernels are
714
+ JIT-compiled through `torch.utils.cpp_extension`, which on modern torch requires
715
+ Ninja and no longer has a distutils fallback, so a clean-env install (e.g. a
716
+ fresh HPC conda env) could fail at runtime with "native bonded extension is
717
+ unavailable". Ninja ships as a pip wheel on every platform (incl. linux-aarch64),
718
+ so the extensions now build out of the box; the dead `use_ninja=False` fallback
719
+ was removed.
720
+ - The `hessian_ff` native extensions now build into a local-filesystem
721
+ directory by default (system temp dir, or `TORCH_EXTENSIONS_DIR`), since
722
+ `torch`'s cpp_extension build lock deadlocks on network filesystems and
723
+ previously hung the first build on NFS/Lustre. Requires GCC ≥ 9
724
+ (`conda install -c conda-forge gxx_linux-64`).
725
+ - Dropped `torch_geometric` / `torch_scatter` from the package dependencies:
726
+ neither is imported by mlmm and current `fairchem-core` does not require them,
727
+ while `torch_scatter` (sdist only) broke a clean `pip install mlmm-toolkit`
728
+ under PEP 517 build isolation.
729
+ - OPC / TIP4P 4-point water with a virtual site (Amber `EPW`, element `EP`) is
730
+ now read and parameterised correctly through the PDB/ASE input layer and the
731
+ OpenMM MM backend (`computeVirtualSites`), instead of mishandling the massless
732
+ extra point. (mlmm-toolkit's default MM backend still needs a 3-point model —
733
+ see `docs/mm-parm.md`.)
734
+ - `orb_precision` now reaches `_OrbBackend` (the kwarg was being silently
735
+ dropped via `**_kwargs`, so `--precision fp64` on the ORB backend
736
+ always ran at the default `float32-high`). The legacy alias
737
+ `"float32"` is rewritten to `"float32-high"` for backward compatibility.
738
+ - Degenerate ML/MM link distances (|r(MM) − r(ML)| < 1e-6 Å) now raise
739
+ `ValueError` instead of silently dropping the link H. The previous
740
+ `continue` left the link slot at its template (0,0,0) position,
741
+ corrupting every downstream energy / force / Hessian. Same fix
742
+ applied in `dft._append_link_hydrogens` for the ONIOM export path.
743
+ - `uma_precision` is now hidden from the echoed config for non-`uma`
744
+ backends (it was missing from `_BACKEND_KEY_PREFIXES`).
745
+ - `_COMMAND_BOOL_SINGLE_FLAG_OPTIONS` is now wired through to the
746
+ DefaultGroup, so `--auto-mm-keep-temp` is correctly classified as a
747
+ single-flag bool by the `bool_compat` shim.
748
+ - `dataset_list` log filter is scoped to the `fairchem` logger subtree
749
+ instead of the root logger (the previous code silently suppressed any
750
+ log record across the process whose message contained `dataset_list`).
751
+ - `workflows/tsopt.py` heavy- / light-mode fallback no longer swallows
752
+ energy + imaginary-mode failures with `except Exception: pass`. Narrow
753
+ to the actual classes raised by `_calc_energy` /
754
+ `_frequencies_cm_and_modes`, log via `logger.warning`, and emit a NaN
755
+ sentinel for energy so downstream consumers can distinguish a missing
756
+ value from a zero.
757
+ - `convert_and_annotate_xyz_to_pdb` docstring corrected: defaults are
758
+ 0 / 10 / 20 (ML / movable / frozen) per `annotate_pdb_bfactors_inplace`,
759
+ not the 100 / 50 / 150 values the docstring previously claimed.
760
+ - ML-region charge/spin parity check hoisted to `MLMMCore.__init__`
761
+ preflight, with a corrected off-by-one in `selection_indices` (the
762
+ hoist initially treated them as 1-based, but `_mk_model_parm7`
763
+ returns parmed 0-based atom indices; the off-by-one produced sum_Z
764
+ values for a shifted atom slice, so e.g. `-q -1 -m 1` for a layered
765
+ ARG-bearing pocket was rejected even though the actual ML electron
766
+ parity was valid). A bad `--charge` / `--multiplicity` combination
767
+ now fails in O(ms) instead of after the multi-second ML model load.
768
+
769
+ ### Added
770
+ - `--calc-file PATH` (with `--calc-factory NAME`): drive the ML region with an
771
+ arbitrary ASE Calculator loaded from a user Python file (a `custom` backend) —
772
+ usable on every subcommand and forwarded through the `all` pipeline. Couple
773
+ GFN-xTB, DFTB+, ORCA, or any ASE-compatible engine without modifying the
774
+ package; the MM side and ONIOM coupling are unchanged and Hessians use the
775
+ finite-difference path. See `docs/backends.md`.
776
+ - `--backend-model NAME` flag on every backend-using subcommand (`opt`,
777
+ `tsopt`, `freq`, `irc`, `scan` / `scan2d` / `scan3d`, `path-opt`,
778
+ `path-search`, `sp`, `all`) to override the model variant for the selected
779
+ `--backend` (e.g. `--backend uma --backend-model uma-m-1p1`), routed to the
780
+ backend's model kwarg via `apply_backend_model_to_calc_cfg`. Previously the
781
+ model variant was settable only through `--config` YAML.
782
+ - `--deterministic` flag on every compute subcommand (`opt`, `tsopt`,
783
+ `freq`, `irc`, `scan`, `scan2d`, `scan3d`, `path-opt`, `path-search`,
784
+ `all`, `sp`) requests deterministic algorithms and an `index_reduce_` shim.
785
+ `MLMM_STRICT_DETERMINISTIC=1` is the environment-variable equivalent;
786
+ backend/model/SDK and target-stack repeatability must be verified.
787
+ - `docs/reproducibility.md` documents backend support and verification guidance.
788
+ - `tests/test_help_grouping.py` locks the four-bucket `--help` section
789
+ rendering + order.
790
+ - `MLMMCore.compute` / `mlmm` Calculator class / MLMMCore.__init__
791
+ workspace setup all gained docstrings (previously empty).
792
+ - `result.json` / `summary.json` envelope now carries
793
+ `schema_version: "1.0"` (bumped when the structure changes) and
794
+ `write_result_json` mirrors every per-stage `result.json` payload to
795
+ a sibling `summary.json` so MCP clients and agents can converge on a
796
+ single filename across every subcommand. `result.json` is preserved
797
+ for back-compat. `RESULT_JSON_STATUS_VALUES` enumerates the allowed
798
+ `status` strings (`success` / `partial` / `error` / `unknown`).
799
+ - Structured error envelope when a subcommand fails: the JSON envelope
800
+ now carries `error_class_chain` (MRO names so agents can match the
801
+ exception hierarchy without parsing text), `error_module`, and
802
+ `error_label` alongside the legacy `error` / `error_type` / `status`
803
+ keys.
804
+ - `mlmm.mcp._runner` exposes `SubcmdResultDict` (TypedDict matching the
805
+ runtime tool payload), `MCP_SUBCMD_RESULT_SCHEMA_VERSION = "1.0"`,
806
+ and `MCP_SUBCMD_RESULT_STATUSES` enum. `SubcmdResult.to_dict` now
807
+ emits `schema_version` so MCP clients can pin the contract.
808
+ - `docs/output-layout.md` (new): single-page reference for the filename
809
+ conventions per subcommand + agent recipe for reading `summary.json`
810
+ with class-chain pattern matching. `docs/json-output.md` and
811
+ `docs/mcp_server.md` updated with the schema_version, summary.json
812
+ mirror, and error envelope semantics.
813
+ - `tests/test_write_result_json.py`, `tests/test_error_envelope.py`,
814
+ `tests/test_mcp_runner.py` (~9 tier-1 assertions) lock in the new
815
+ envelope contracts so regressions show up at pytest time.
816
+ - `mlmm.workflows._all_helpers` (new module) hosts module-level helpers
817
+ extracted from `workflows/all.py:cli()`:
818
+ * `AllContext` (frozen dataclass) bundling the 72 `mlmm all` CLI
819
+ parameters in declaration order.
820
+ * `copy_path_outputs_to_root` / `promote_diag_for_root` (replace
821
+ nested closures).
822
+ * `build_energy_level_dict` factors the 5-way R/TS/P energy-level
823
+ dict pattern (UMA / Gibbs / DFT / Gibbs-DFT) into one helper.
824
+ * `build_pipeline_summary_payload` factors the summary-log payload
825
+ assembly so the dict construction is unit-testable separately
826
+ from the I/O wrapper.
827
+ * `build_tsopt_overrides` / `build_freq_overrides` /
828
+ `build_dft_overrides` factor the inline override-dict assembly
829
+ (`if x is not None: cfg[k] = ...` ladder) for the post-MEP
830
+ TSOPT / FREQ / DFT stage calls.
831
+ * `append_backend_forwarding_args` consolidates the 8 copies of
832
+ the backend / charge-embedding / link / mm / cmap / args-yaml
833
+ argv-build ladder shared by every cli()-internal subcommand
834
+ forwarder. Preserves the "only emit `--no-embedcharge` when
835
+ the user explicitly typed it" subtlety so a YAML
836
+ `calc.embedcharge: true` is not silently overridden.
837
+ - `mlmm.core.utils.resolve_ml_layer_assignment` (new helper) collapses
838
+ the ~30-LOC layer-resolution block that was duplicated verbatim
839
+ across `scan.py`, `scan2d.py`, and `scan3d.py` (movable_cutoff
840
+ gating, detect_layer fallback to explicit ML region, model_pdb /
841
+ model_indices resolution, calc_cfg mutation). Single source of
842
+ truth for the layer-assignment behavior.
843
+ - `tests/test_all_helpers.py` (9 cases) pins the extracted helper
844
+ contracts (copy-outputs no-op, copy-outputs canonical artefact set,
845
+ promote-diag none/rewrite, energy-level kcal projection +
846
+ no-input-mutation, AllContext frozen + 72 fields, pipeline summary
847
+ payload shape, AllContext signature drift guard).
848
+ - TS-opt microiteration now supports internal coordinate macro steps
849
+ (`--coord-type dlc|redund|tric`) by running the MM-only micro relax
850
+ on a fresh cart twin Geometry and projecting the converged positions
851
+ back onto the macro geometry via `Geometry.reset_coords()`. Cart-only
852
+ macro path is byte-equivalent to the previous behaviour. Removes the
853
+ pre-existing pysisyphus cart<->internal roundtrip drift that aborted
854
+ microiter runs with `assert_allclose` after a few macro cycles.
855
+ - Smoke `tests/smoke/run.sh` expanded with per-stage `--coord-type
856
+ {dlc,redund,tric}` + `--precision fp64` + `--mm-backend openmm` +
857
+ `--link-atom-method fixed` test coverage (test50a/d/g/j/k/m/n/o);
858
+ test50 itself capped at `--max-cycles 5 --no-tsopt/thermo/dft` so
859
+ the DLC code path is exercised without the multi-hour GSM
860
+ convergence the uncapped run requires.
861
+ - `--precision fp32|fp64` accepted on every calculator-constructing
862
+ subcommand. The flag was previously available only on `tsopt / freq /
863
+ irc / sp`; it now also covers `opt / all / path-opt / path-search /
864
+ scan / scan2d / scan3d`. For `all`, the value propagates to every
865
+ child stage through the shared args YAML, so a single top-level
866
+ switch covers the full pipeline.
867
+ - `--irc-pos-def` (IRC convergence guard requiring PSD mass-weighted
868
+ Hessian) is opt-in on `irc`; blocks the IRC "shoulder" false
869
+ convergence where the rms-only criterion calls success before
870
+ reaching the local minimum.
871
+ - `mlmm.core.residue_data` (new module) hosts the `AMINO_ACIDS` / `ION`
872
+ / `WATER_RES` tables shared between `workflows/extract` and
873
+ `domain/add_elem_info`; removes the L3 -> L2 import inversion that
874
+ previously had `domain/add_elem_info` reaching back into
875
+ `workflows/extract`. The legacy `from mlmm.workflows.extract import
876
+ AMINO_ACIDS, ...` form is preserved via re-export.
877
+
878
+ ### Removed
879
+ - **BREAKING:** `mlmm pysis` subcommand and the `mlmm.pysis_runner` module.
880
+ This was a thin wrapper that registered the `mlmm` calculator into
881
+ pysisyphus's `CALC_DICT` and shelled out to the pysisyphus YAML runner,
882
+ providing v0.1.x YAML-workflow compatibility (`mlmm opt.yaml`). That
883
+ compatibility surface is dropped: drive runs through the `mlmm`
884
+ subcommands (with `--config` for YAML-supplied defaults) instead. Using
885
+ the `mlmm` calculator directly from Python (`from mlmm import mlmm`)
886
+ is unaffected.
887
+ - **BREAKING:** Flat-top compatibility shim layer removed. The package now
888
+ lives under 6 layer directories (`cli/`, `workflows/`, `domain/`,
889
+ `backends/`, `io/`, `core/`); the shims at `mlmm/<file>.py` that
890
+ re-exported the new locations have been deleted in this release. External
891
+ code must migrate dotted imports to the layered paths:
892
+
893
+ | Old (removed) | New |
894
+ |----------------------------|------------------------------------|
895
+ | `mlmm.{all,opt,tsopt,freq,irc,scan,scan2d,scan3d,path_opt,path_search,extract,dft,mm_parm,oniom_export,oniom_import,define_layer}` | `mlmm.workflows.<same>` |
896
+ | `mlmm.align_freeze_atoms` | `mlmm.workflows.align_freeze` |
897
+ | `mlmm.scan_common` | `mlmm.workflows.scan_common` |
898
+ | `mlmm.{defaults,utils}` | `mlmm.core.<same>` |
899
+ | `mlmm.{mlmm_calc,xtb_embedcharge_correction}` | `mlmm.backends.<same>` |
900
+ | `mlmm.{bond_changes,bond_summary,add_elem_info}` | `mlmm.domain.<same>` |
901
+ | `mlmm.{energy_diagram,trj2fig,hessian_cache,hessian_calc}` | `mlmm.io.<same>` |
902
+ | `mlmm.harmonic_constraints` | `mlmm.workflows.restraints` |
903
+ | `mlmm.fix_altloc` | `mlmm.io.pdb_fix` |
904
+ | `mlmm.summary_log` | `mlmm.io.summary` |
905
+ | `mlmm.cli_utils` | `mlmm.cli.decorators` |
906
+ | `mlmm.{bool_compat,default_group,preflight}` | `mlmm.cli.<same>` |
907
+ | `mlmm.advanced_help` | `mlmm.cli.help_pages` |
908
+
909
+ All console-script subcommands are preserved across this move **except**
910
+ `mlmm pysis` (removed, see below); `mlmm sp` was added. Other Python
911
+ imports change as tabulated above.
912
+ - `--trust-band` / `--hessian-window` / `--weighted-trust` CLI flags
913
+ (and their `add_*_option` factories). The vendored
914
+ pysisyphus `HessianOptimizer` kwargs are left dormant; no
915
+ behaviour change since defaults were always legacy.
916
+
917
+ ### Documentation
918
+ - Documented that the `[orb]` extra's `torch_scatter` has no PyPI binary wheel
919
+ (sdist only, fails under PEP517 build isolation): install from PyG's
920
+ prebuilt-wheel index, e.g.
921
+ `pip install "mlmm-toolkit[orb]" -f https://data.pyg.org/whl/torch-2.8.0+cu129.html`.
922
+ - Documented automatic vs manual ML-region definition (concepts) and clarified
923
+ that `extract --add-linkh` is distance-based and unnecessary when preparing a
924
+ `--model-pdb` for mlmm (the ML/MM calculator caps the boundary from the `--parm`
925
+ topology). Genericized the `json-output` environment example to placeholders,
926
+ and switched the README overview image to an absolute URL so it renders on PyPI.
927
+ - Documented the default MM backend's 4-point-water (OPC/TIP4P) limitation and the
928
+ 3-point (OPC3/TIP3P) / `--mm-backend openmm` workarounds in `mm-parm.md`.
929
+ ## [0.2.9] — unreleased
930
+
931
+ Consolidated changes since v0.2.4, pending the `v0.2.9` tag. Highlights: end-to-end
932
+ JSON output, MM-only optimization mode, GPU memory headroom for 16 GB
933
+ consumer cards via Hessian-update CPU offload, 1-based atom indices in
934
+ all user-facing I/O, energy-plateau convergence fallback, and a
935
+ comprehensive documentation overhaul (EN/JA).
936
+
937
+ ### Added — CLI features
938
+ - `mlmm bond-summary`: detect bond changes between two structures
939
+ (positional args supported, R/P sanity check companion to `mlmm all`).
940
+ - `mlmm opt --mm-only`: skip the MLIP component and minimize on the MM
941
+ force field only. Layers are still honored; `--opt-mode hess` rejected
942
+ (MM-only calculator does not provide a Hessian); microiteration is
943
+ auto-disabled. Useful as a cheap MM pre-relaxation before ML/MM
944
+ ONIOM optimization.
945
+ - `--out-json` across all MLIP subcommands (`opt`, `tsopt`, `freq`,
946
+ `irc`, `scan`, `scan2d`, `scan3d`, `path-opt`, `dft`): emits a
947
+ `result.json` record per run, including backend, charge,
948
+ spin, and timing. `mlmm all` migrates `summary.yaml` → `summary.json`.
949
+ - `--link-atom-method scaled|fixed` on all computation subcommands.
950
+ - `--cmap / --no-cmap` to control one CMAP policy across both MM layers.
951
+ - `--hess-device`, `--read-hess`, `--dump-hess`, `--skip-final-freq`
952
+ for explicit Hessian device control and serialization.
953
+ - `--engine gpu|cpu` for `mlmm dft`; `--lowmem` selects
954
+ `gpu4pyscf.rks_lowmem.RKS` as the closed-shell default.
955
+ - `--modified-residue` for `mlmm extract` / `mlmm all` (non-standard
956
+ amino acid handling).
957
+ - `--freeze-atoms` for `mlmm irc`.
958
+ - ML-region charge/multiplicity sanity validator that runs before the
959
+ first MLIP evaluation, surfacing wavefunction-charge mismatches early.
960
+
961
+ ### Added — Convergence and stability
962
+ - Energy-plateau convergence fallback for optimizers (range-based,
963
+ `1e-4 au` threshold; skipped for chain-of-states optimizers).
964
+ - Pre-MEP global alignment of reactant / product structures.
965
+ - Auto ECP for def2 basis sets in `mlmm dft`.
966
+
967
+ ### Added — Documentation & infrastructure
968
+ - Agent skill set under `skills/` covering install, structure
969
+ I/O, CLI subcommands, workflows / outputs, and HPC submission.
970
+ - ChemRxiv preprint added as `preferred-citation` in `CITATION.cff`
971
+ (DOI placeholder until Zenodo release).
972
+ - xTB install documentation and improved not-found error message.
973
+ - JSON Output Reference page (EN + JA).
974
+ - GPU4PySCF Blackwell-class OOM workaround note in `dft.md` (EN + JA).
975
+ - `mm-parm` antechamber sqm odd-electron pitfall.
976
+
977
+ ### Changed
978
+ - Atom indices unified to **1-based** for all user-facing I/O
979
+ (`--freeze-atoms`, layer atoms, scan target display, etc.) with
980
+ corresponding EN/JA documentation updates.
981
+ - `--refine-path` introduced (default `True`, recursive `path-search`; the
982
+ default was later flipped to `--no-refine-path` / single-pass `path-opt` in 0.3.0).
983
+ - `--exclude-backbone` default `False`.
984
+ - RFO / RS-I-RFO `trust_max` reduced from `0.20` to `0.10` for MLIP
985
+ stability.
986
+ - `--max-nodes` default `20` for `path-search` / `path-opt`.
987
+ - `--thresh` default `gau_loose` (was `gau`).
988
+ - Unified `ml_hessian_mode` and `hessian_calc_mode` under a single
989
+ `hessian_calc_mode` setting.
990
+ - Imaginary mode filenames shortened (`final_imag_mode` → `imag`).
991
+ - Shorter / cleaner CLI logging with deduplicated blank lines, relative
992
+ paths in echo, and suppressed empty config blocks.
993
+
994
+ ### Fixed — GPU memory
995
+ - `pysisyphus/optimizers/hessian_updates.py`: `bofill_update` runs on
996
+ CPU for `torch.Tensor` input, avoiding a ~5 GB GPU peak (sr1 / psb /
997
+ mix temporaries each ~1.35 GB for ~4000-atom active DOF). Result is
998
+ transferred back to the Hessian's device. Pairs with the IRC fix
999
+ below; together they bring active-DOF 4000–5000 IRC into the 16 GB
1000
+ consumer-GPU envelope.
1001
+ - `pysisyphus/irc/IRC.py`: stash `forward_mw_hessian` on CPU during
1002
+ backward integration (consumer transfers it back to numpy at the end
1003
+ anyway). Frees ~`N_dof² × 8 B` of VRAM for the entire backward run.
1004
+ - All workflow CLIs (`opt`, `tsopt`, `freq`, `irc`, `scan`,
1005
+ `path-search`): explicit `del` of heavy locals (`calc`, `optimizer`,
1006
+ `H_t`, `modes`) in `finally` before `gc.collect()` and
1007
+ `torch.cuda.empty_cache()`. Fixes the post-tsopt → freq OOM observed
1008
+ in `mlmm all` on bezA-class systems with 16 GB GPU (cyclic-gc could
1009
+ not break torch.nn.Module hook / closure cycles).
1010
+
1011
+ ### Fixed — Correctness
1012
+ - ML-region `model_charge` / `model_mult` now follow the CLI-resolved
1013
+ value (`-q` or `-l` derivation) across `opt`, `tsopt`, `freq`,
1014
+ `irc`, `scan`, `path-search`, `path-opt`. Previously the
1015
+ `MLMM_CALC_KW` default `0` silently overrode user input because
1016
+ `dict.get(key, fallback)` returned the present default rather than
1017
+ the CLI value; wavefunctions were computed against charge 0
1018
+ regardless of `--ligand-charge` (UMA's robustness masked downstream
1019
+ symptoms). Single-line fix replaces the get-with-fallback dance with
1020
+ direct assignment from CLI-resolved `charge` / `spin`.
1021
+ - `mlmm/all.py`: honor input PDB B-factor layers when active-site
1022
+ extraction is skipped (`-c` not supplied with pre-layered input).
1023
+ - `mlmm/mlmm_calc.py`: use 1-based ATOM/HETATM file position as `idx`,
1024
+ not raw PDB serial. PDBs with serial gaps (e.g. 3411→3418) no longer
1025
+ trigger `IndexError` in `_mk_model_parm7`. Regression test:
1026
+ `tests/test_mlmm_calc_serial_gap.py`.
1027
+ - Microiteration oscillation with scaled link atoms.
1028
+ - IRC initial-displacement bisection (in-place mutation + missing
1029
+ numpy import).
1030
+ - EulerPC corrector Euler integration safety guards (zero / NaN
1031
+ gradient norms).
1032
+ - Tangent normalization and SVD alignment NaN / zero guards.
1033
+ - DFT failure path: graceful skip of energy diagrams + status in
1034
+ summary; error `result.json` written on subcommand failure for all
1035
+ CLI commands.
1036
+ - `bond-summary` PDB loading (`geom_from_pdb_str` → `geom_from_pdb`).
1037
+ - `_resolve_device`: handle `'auto'` → `cuda` / `cpu`.
1038
+ - PDB element inference: text-based coord replacement + robust atom
1039
+ name parsing for protein hydrogens.
1040
+ - `mlmm/all.py` modified-residue handling: `AMINO_ACIDS` restore
1041
+ wrapped in `try / finally`.
1042
+ - MODEL / ENDMDL missing on first frame in PDB trajectory conversion.
1043
+
1044
+ ### Fixed — Documentation hygiene
1045
+ - `skills/`: factual cleanup of trajectory file names
1046
+ (`opt_trj.xyz` → `optimization_trj.xyz`), default-dict references
1047
+ (`UMA_CALC_KW` → `MLMM_CALC_KW`), TS Hessian flag (`--hessian-init`
1048
+ → `--hessian-calc-mode`), output-tree paths, status enum values, and
1049
+ removal of phantom `.log` rows. Cosmetic-only changes to the AI-side
1050
+ cheatsheets; the docs/source are the canonical references.
1051
+
1052
+ ### Documentation
1053
+ - Bilingual documentation overhaul (EN/JA) covering terminology,
1054
+ accuracy, and consistency: net charge (not total), pocket → ML
1055
+ region determination, total → net charge wording, link-atom and
1056
+ microiteration concepts, EN/JA alignment for 1-based atom indices.
1057
+ - "At a glance" 5-bullet block (Use when / Method / Outputs
1058
+ / Defaults / Next step) added to the 10 core subcommand pages
1059
+ (`opt`, `tsopt`, `freq`, `irc`, `scan`, `scan2d`, `scan3d`,
1060
+ `path-opt`, `path-search`, `dft`).
1061
+ - `docs/yaml-reference.md` (EN+JA): document `mlmm:` block as alias
1062
+ for `calc:` (10 callsites verified); add `dft.engine`, `dft.ecp`,
1063
+ `freq.active_dof_mode`; correct lbfgs/rfo Used-by columns; drop
1064
+ dead `stopt.rfo` example.
1065
+ - `docs/cli-conventions.md` (EN+JA): drop `< override-yaml` from
1066
+ precedence chain (slot kept in `cli_utils.py` for legacy compat but
1067
+ no CLI subcommand exposes the flag); `--opt-mode` alias coverage
1068
+ (`grad / hess / light / heavy / lbfgs / rfo`); em-dash convention.
1069
+ - `docs/json-output.md` (EN+JA): replace "every MLIP-based subcommand"
1070
+ `--out-json` claim with explicit list (excludes `path-search`,
1071
+ `all`, `bond-summary`, `define-layer`, `mm-parm`); enumerate the
1072
+ 6-alias `opt_mode` Choice; add `n_segments_reactive` field for
1073
+ `mlmm all`.
1074
+ - Reference docs regenerated from current CLI; markdown filenames
1075
+ normalized from underscores to hyphens; theme toggle simplified to
1076
+ light ↔ dark.
1077
+
1078
+ ## [0.2.4] — 2026-03-18
1079
+
1080
+ ### Added
1081
+ - GitHub Pages deployment and PyPI release workflows.
1082
+ - Bidirectional scan (4-tuple) documentation (EN/JA).
1083
+
1084
+ ### Fixed
1085
+ - Regenerated CLI reference docs to match current `--help` output.
1086
+ - Documentation accuracy: defaults, YAML examples, EN/JA alignment.
1087
+ - Generalized UMA-specific wording to MLIP where applicable.
1088
+ - Improved hessian_ff JIT build error message.
1089
+
1090
+ ## [0.2.0] — 2026-03-16
1091
+
1092
+ **Complete rewrite from `mlmm` (v0.1.1) to `mlmm-toolkit` (v0.2.0).**
1093
+ This release replaces the previous pysisyphus-wrapper architecture with a unified Click-based CLI toolkit. The package name, CLI interface, dependency stack, and internal architecture have all changed.
1094
+
1095
+ ### Breaking Changes
1096
+
1097
+ - **Package name**: `mlmm-toolkit` on PyPI. Install with `pip install mlmm-toolkit`.
1098
+ - **CLI completely redesigned**: The old entry points (`mlmm`, `def_ml_region`, `bond_scan`, `ts_search`, `energy_summary`, `trj2fig`, `add_elem_info`, `get_freeze_indices`, `xyz_geom2pdb`) are replaced by a single `mlmm <subcommand>` interface with 21 subcommands.
1099
+ - **OpenMM removed**: MM calculations now use `hessian_ff`, a bundled C++ native extension for Amber force fields. OpenMM and OpenMM-CUDA-12 are no longer dependencies.
1100
+ - **RDKit removed**: No longer required.
1101
+ - **pysisyphus bundled**: No longer installed from a separate git repository; a modified fork is included in the package.
1102
+ - **fairchem-core from PyPI**: No longer installed from a custom git fork.
1103
+ - **numpy constraint relaxed**: `numpy<2.0` → `numpy>=1.24` (NumPy 2.x compatible).
1104
+ - **Configuration**: Shared runtime defaults were centralized in `defaults.py`; command-local CLI defaults remain with their Click options.
1105
+
1106
+ ### Added — CLI & Workflow
1107
+
1108
+ - **`mlmm all`**: End-to-end workflow command. Given PDB files (R → P), automatically extracts active-site pockets, generates MM parameters, assigns ONIOM layers, runs MEP search, and optionally performs TS optimization, IRC, vibrational analysis, and single-point DFT — all in one invocation.
1109
+ - **`mlmm opt`**: Single-structure geometry optimization with LBFGS (grad mode) or RFO (hess mode) with microiteration support.
1110
+ - **`mlmm scan` / `scan2d` / `scan3d`**: 1D, 2D, and 3D constrained distance scans along user-specified atom pairs.
1111
+ - **`mlmm path-search`**: Recursive minimum-energy path search with GSM (Growing String Method) and DMF (Direct Max Flux).
1112
+ - **`mlmm path-opt`**: Single-pass MEP optimization (GSM or DMF).
1113
+ - **`mlmm tsopt`**: Transition-state optimization (dimer method with partial Hessian, or RS-I-RFO with full Hessian).
1114
+ - **`mlmm irc`**: Intrinsic reaction coordinate calculation from a TS geometry. Outputs `forward_last.pdb` and `backward_last.pdb` endpoint structures.
1115
+ - **`mlmm freq`**: Vibrational analysis and thermochemistry (partial or full Hessian).
1116
+ - **`mlmm dft`**: GPU-accelerated single-point DFT via PySCF / gpu4pyscf (`pip install "mlmm-toolkit[dft]"`).
1117
+ - **`mlmm extract`**: Active-site pocket extraction from full protein-ligand PDB structures.
1118
+ - **`mlmm mm-parm`**: Automatic Amber parm7/rst7 generation via AmberTools (tleap + GAFF2/AM1-BCC).
1119
+ - **`mlmm define-layer`**: Assign 3-layer ML/MM partitioning via B-factor encoding (ML=0, MovableMM=10, FrozenMM=20).
1120
+ - **`mlmm oniom-export` / `oniom-import`**: Gaussian/ORCA ONIOM input/output interoperability.
1121
+ - **`mlmm add-elem-info`**: Fix missing or incorrect PDB element columns.
1122
+ - **`mlmm fix-altloc`**: Resolve alternate location indicators in PDB files.
1123
+ - **`mlmm energy-diagram`**: Plot energy profiles with Plotly (interactive HTML + static PNG).
1124
+ - **`mlmm trj2fig`**: Trajectory visualization (PNG snapshots from XYZ trajectories).
1125
+
1126
+ ### Added — Core Features
1127
+
1128
+ - **Multi-backend MLIP support**: UMA (default), ORB, MACE, AIMNet2 — selectable via `-b/--backend` on all subcommands.
1129
+ - **ONIOM-like ML/MM decomposition**: `E = E_MM_real + E_ML_model − E_MM_model` with link-atom Jacobian transformation.
1130
+ - **hessian_ff**: Bundled C++ native extension for analytical Amber force field energies, forces, and Hessians (replaces OpenMM).
1131
+ - **Microiteration scheme**: Efficient optimization of large ML/MM systems (~10,000 atoms) by separating ML and MM degrees of freedom.
1132
+ - **xTB embed-charge correction**: Optional point-charge embedding for the ML region (`--embedcharge`).
1133
+ - **Partial Hessian approach**: Compute Hessians only for the ML region + boundary atoms, enabling TS optimization and frequency analysis on large systems.
1134
+ - **Automatic versioning**: setuptools-scm replaces hard-coded `__version__`.
1135
+ - **Progressive help**: `--help` shows primary options; `--help-advanced` shows the full option set.
1136
+ - **DefaultGroup**: Lazy-loading Click subcommand architecture with automatic boolean normalization (`--flag/--no-flag` and `--flag True/False` both supported).
1137
+
1138
+ ### Added — Dependencies & Extras
1139
+
1140
+ - **Optional backends**: `pip install "mlmm-toolkit[orb]"`, `"mlmm-toolkit[aimnet]"`. MACE (`mace-torch`) conflicts with `fairchem-core` (UMA) due to incompatible `e3nn` versions; use separate conda environments if both are needed.
1141
+ - **Optional DFT**: `pip install "mlmm-toolkit[dft]"` for PySCF + gpu4pyscf-cuda12x + CuPy.
1142
+ - **Optional PDBFixer**: `pip install "mlmm-toolkit[pdbfixer]"` for hydrogen addition.
1143
+ - **New core dependencies**: `click`, `torch_geometric`, `torch_scatter`, `pyparsing`, `tabulate`.
1144
+ - **Bundled packages**: `pysisyphus` (modified fork), `thermoanalysis`, `hessian_ff`.
1145
+
1146
+ ### Added — Documentation & Testing
1147
+
1148
+ - Bilingual documentation (English + Japanese) under `docs/` and `docs/ja/`.
1149
+ - Per-subcommand reference docs with CLI tables and YAML configuration examples.
1150
+ - `CONTRIBUTING.md` with contributor guidelines.
1151
+ - GitHub Actions CI: `pytest.yml`, `smoke_test.yml`, `docs_quality.yml`.
1152
+ - 198 unit tests with `pytest --timeout=120`.
1153
+ - Smoke test suite: 34 end-to-end tests covering all subcommands.
1154
+ - Working examples in `examples/toy_system/` and `examples/methyltransferase/`.
1155
+
1156
+ ### Fixed (relative to v0.1.1)
1157
+
1158
+ - IRC energy-based initial displacement: Added `step_length` clamp (max 0.5 au) to prevent divergence when `min_eigval ≈ 0`.
1159
+ - Hessian cache: Fixed `set_calculator()` clearing `within_partial_hessian` before `cart_hessian` assignment.
1160
+ - HessianOptimizer: Fixed GPU/CPU device mismatch in `hessian_recalc`.
1161
+ - Frequency analysis: Added float64 enforcement and explicit `(H+Hᵀ)/2` symmetrization.
1162
+ - PDB element inference: Fixed `_infer_element_from_pdb_atom_name()` misidentifying protein hydrogens (e.g., `HG2` → `Hg`). Now uses residue-context-aware `guess_element()`.
1163
+ - Silent `except Exception: pass` blocks converted to `logger.debug(...)` across all modules.
1164
+ - Charge derivation: Uses `--model-pdb` instead of full input when provided.
1165
+ - TypeError in `_pseudo_irc_and_match` when mapping value is `None`.
1166
+
1167
+ ### Architecture — v0.1.1 → v0.2.0
1168
+
1169
+ | Aspect | v0.1.1 (`mlmm`) | v0.2.0 (`mlmm-toolkit`) |
1170
+ |--------|------------------|--------------------------|
1171
+ | Package name | `mlmm` | `mlmm-toolkit` |
1172
+ | Codebase size | ~3,700 lines (13 files) | ~35,000 lines (40+ files) |
1173
+ | CLI framework | 8 separate entry points (argparse) | 1 entry point, 21 Click subcommands |
1174
+ | MM backend | OpenMM (finite difference) | hessian_ff (analytical C++) |
1175
+ | ML backends | UMA, AIMNet2 | UMA, ORB, MACE, AIMNet2 |
1176
+ | pysisyphus | Git dependency | Bundled (modified fork) |
1177
+ | fairchem-core | Git fork | PyPI release |
1178
+ | Configuration | Inline kwargs | Centralized `defaults.py` |
1179
+ | ONIOM support | Basic link atoms | Full Jacobian + 3-layer B-factor |
1180
+ | Documentation | README only | Full bilingual docs (EN/JA) |
1181
+ | CI/CD | None | GitHub Actions (3 workflows) |
1182
+ | Tests | None | 198 unit + 34 smoke tests |