Release 2026.2.0

CHANGELOG.md

@@ -1,21 +1,20 @@
 Changelog
 =========
 
-[2026.2.0](https://github.com/openbiosim/ghostly/compare/2026.1.0...2026.2.0) - ********
--------------------------------------------------------------------------------------
+[2026.2.0](https://github.com/openbiosim/ghostly/compare/2026.1.0...2026.2.0) - Sep 2026
+----------------------------------------------------------------------------------------
 
-* Please add an item to this CHANGELOG for any new features or bug fixes when creating a PR.
 * Revert auto-zeroing of anchor dihedrals for ring-constrained immediate ghosts: a singly-attached ring still has one unconstrained rotational degree of freedom about the bridge-ghost bond, so removing the anchor reintroduces flapping risk.
 
 [2026.1.0](https://github.com/openbiosim/ghostly/compare/2025.2.0...2026.1.0) - Jun 2026
--------------------------------------------------------------------------------------
+----------------------------------------------------------------------------------------
 
 * Add linear spacer modification for ring-breaking ghost bridges.
 * Fixed missing removal of bridge-extension dihedrals (`real–ghost–ghost–ghost`) that arise when a ghost group contains a ring, e.g. cyclopropyl, where the ring topology creates spurious torsional coupling between the real scaffold and the ghost ring interior.
 * Auto-zero anchor dihedrals when the immediate ghost atom lies on a ring within the ghost subgraph. The ring topology already constrains the ghost orientation relative to the bridge, making the anchor redundant.
 
 [2025.2.0](https://github.com/openbiosim/ghostly/compare/2025.1.0...2025.2.0) - Mar 2026
--------------------------------------------------------------------------------------
+----------------------------------------------------------------------------------------
 
 * Ensure that there are independent, per-state, physical neighbours. This removes the spurious cross-state ghost filter that resulted in a downgrading of junction types when multiple junctions were present in the molecule.
 * Added a fallback for failed angle optimisation to ensure that the original equilibrium value is preserved when convergence fails.
@@ -29,6 +28,6 @@ Changelog
 * Changed the default for ``--optimise-angles`` to ``False``. Angle optimisation is conformer-dependent: different input geometries can yield different equilibrium angles for the same perturbation, adding variability to the resulting force field. The original force field theta0 is sufficient given the small ``k_soft`` value. Optimisation can still be enabled explicitly to follow Boresch et al. (JCTC 2021) strictly.
 
 [2025.1.0](https://github.com/OpenBioSim/ghostly/releases/tag/2025.1.0) - Nov 2025
--------------------------------------------------------------------------------
+----------------------------------------------------------------------------------
 
 * Initial public release.

README.md

@@ -161,7 +161,7 @@ ghostly --system system.bss --log-level debug
 When finished, the program will output a [BioSimSpace](https://biosimspace.openbiosim.org)
 stream file for the perturbable molecule, or AMBER or GROMACS files for the two end states.
 The format can be specified using the the `--output-format` option. If you require input
-for a free-energy perturbation simulation, e.g. a hybrid GROMACS toplogy file, the you can
+for a free-energy perturbation simulation, e.g. a hybrid GROMACS topology file, then you can
 use the stream file with [BioSimSpace](https://biosimspace.openbiosim.org) to generate the
 required input files. Additionally, the program will output a JSON file summarising the
 modifications that were made.

pixi.toml

@@ -9,25 +9,25 @@ loguru = "*"
 
 [target.linux-64.dependencies]
 # main
-biosimspace = ">=2026.1.0,<2026.2.0"
+biosimspace = ">=2026.2.0,<2026.3.0"
 # devel
-#biosimspace = "==2026.2.0.dev"
+#biosimspace = "==2026.3.0.dev"
 
 [target.linux-aarch64.dependencies]
 # biosimspace/sire not available as conda packages on linux-aarch64;
 # build from source first
 
 [target.osx-arm64.dependencies]
 # main
-biosimspace = ">=2026.1.0,<2026.2.0"
+biosimspace = ">=2026.2.0,<2026.3.0"
 # devel
-#biosimspace = "==2026.2.0.dev"
+#biosimspace = "==2026.3.0.dev"
 
 [target.win-64.dependencies]
 # main
-biosimspace = ">=2026.1.0,<2026.2.0"
+biosimspace = ">=2026.2.0,<2026.3.0"
 # devel
-#biosimspace = "==2026.2.0.dev"
+#biosimspace = "==2026.3.0.dev"
 
 [feature.test.dependencies]
 pytest = "*"

recipes/ghostly/recipe.yaml

@@ -21,16 +21,20 @@ requirements:
     - versioningit
   run:
     # main
-    - biosimspace >=2026.1.0,<2026.2.0
+    - biosimspace >=2026.2.0,<2026.3.0
     # devel
-    #- biosimspace ==2026.2.0.dev
+    #- biosimspace ==2026.3.0.dev
     - loguru
     - python
 
 tests:
   - python:
       imports:
         - ghostly
+      # AmberTools can be pulled in transitively, and installs tools into
+      # site-packages whose metadata still pins numpy <2, so 'pip check'
+      # fails even though nothing here uses them.
+      pip_check: false
   - script:
       - pytest -vvv --color=yes --import-mode=importlib ./tests
     files:

src/ghostly/_ghostly.py

@@ -2584,14 +2584,6 @@ def _check_rotamer_anchors(
 
     which has a single minimum at phi = 0 (trans) and a barrier of 2k.
 
-    .. note::
-
-        Stiffening is not currently enabled. When wiring in, add
-        ``stiffen_rotamers`` and ``k_rotamer`` parameters to ``modify()``
-        and expose them through the CLI. The ``modifications`` dict will
-        also need a ``"stiffened_dihedrals"`` key initialised to an empty
-        list for each end state.
-
     Parameters
     ----------