Conda CI is failing

[j-wags]

Conda CI has been failing: https://github.com/openforcefield/openff-toolkit/actions/workflows/conda.yml

[j-wags]

I think the root cause has already been fixed in the QCArchive example notebook change in #2187. But while that fixed it on main, the conda CI tests the example notebooks as they exist in the most recent conda package. Since we haven't released the notebook change, conda CI is still failing. So I'll plan to make a release to fix this shortly.

[mattwthompson]

Re-started after the release: https://github.com/openforcefield/openff-toolkit/actions/runs/27344706045

[j-wags]

3/4 Conda test builds pass after I made the 0.18.1 Toolkit release ~14 hours ago, up from 0/4 before then.

The last failing test appears to be persistent (Matt ran it 3 times). It's ubuntu/py3.12/openeye=false. It reports installingambertools 24.8 cuda_None_nompi_py312hbe1986e_101 and finding antechamber but the charge assignment step fails in a generic non informative way..

I'll look a bit more into how we've resolved this sort of failure before (it looks very familiar) but if I can't get results in an hour or two I'll put this issue aside to work on higher priority things.

[j-wags]

I'm unable to reproduce this error in a linux docker container on my M1 macbook (using the same AmberTools package). I'm going to put this issue aside for at least a few days.

Here was my attempt to reproduce in a docker container.

docker run --platform linux/amd64 -it mambaorg/micromamba

micromamba create -n myenv -c conda-forge openff-toolkit-examples python=3.12 ambertools nbval
micromamba activate myenv
openff-toolkit-examples
cd examples/toolkit_showcase
python -m pytest -r fE -v -x --tb=short --nbval-lax toolkit_showcase.ipynb

(Later addendum: I also fail to reproduce the error if I add micromamba install "numpy<2.3" "pytest=7.4", which are the other dep version pins that I see in the CI log)

returns

=================================================== test session starts ====================================================
platform linux -- Python 3.12.13, pytest-9.0.3, pluggy-1.6.0 -- /opt/conda/envs/myenv/bin/python
cachedir: .pytest_cache
rootdir: /tmp/examples/toolkit_showcase
plugins: anyio-4.13.0, nbval-0.11.0
collected 14 items                                                                                                         

toolkit_showcase::ipynb::Cell 0 PASSED                                                                               [  7%]
toolkit_showcase::ipynb::Cell 1 PASSED                                                                               [ 14%]
toolkit_showcase::ipynb::Cell 2 PASSED                                                                               [ 21%]
toolkit_showcase::ipynb::Cell 3 PASSED                                                                               [ 28%]
toolkit_showcase::ipynb::Cell 4 PASSED                                                                               [ 35%]
toolkit_showcase::ipynb::Cell 5 PASSED                                                                               [ 42%]
toolkit_showcase::ipynb::Cell 6 PASSED                                                                               [ 50%]
toolkit_showcase::ipynb::Cell 7 PASSED                                                                               [ 57%]
toolkit_showcase::ipynb::Cell 8 PASSED                                                                               [ 64%]
toolkit_showcase::ipynb::Cell 9 PASSED                                                                               [ 71%]
toolkit_showcase::ipynb::Cell 10 PASSED                                                                              [ 78%]
toolkit_showcase::ipynb::Cell 11 PASSED                                                                              [ 85%]
toolkit_showcase::ipynb::Cell 12 PASSED                                                                              [ 92%]
toolkit_showcase::ipynb::Cell 13 PASSED                                                                              [100%]

============================================== 14 passed in 624.01s (0:10:24) ==============================================

Version info:

(myenv) mambauser@f058ad7040a2:/tmp/examples/toolkit_showcase$ micromamba list ambertools 
List of packages in environment: "/opt/conda/envs/myenv"

  Name        Version  Build                              Channel    
───────────────────────────────────────────────────────────────────────
  ambertools  24.8     cuda_None_nompi_py312hbe1986e_101  conda-forge
(myenv) mambauser@f058ad7040a2:/tmp/examples/toolkit_showcase$ micromamba list openff
List of packages in environment: "/opt/conda/envs/myenv"

  Name                     Version    Build         Channel    
─────────────────────────────────────────────────────────────────
  openff-amber-ff-ports    2025.09.0  pyhd8ed1ab_0  conda-forge
  openff-forcefields       2026.01.0  pyhd8ed1ab_0  conda-forge
  openff-interchange       0.5.3      pyhd8ed1ab_3  conda-forge
  openff-interchange-base  0.5.3      pyhd8ed1ab_3  conda-forge
  openff-nagl              0.5.5      pyhd8ed1ab_0  conda-forge
  openff-nagl-base         0.5.5      pyhd8ed1ab_0  conda-forge
  openff-nagl-models       2025.9.0   pyhd8ed1ab_0  conda-forge
  openff-packmol           0.1        pyhc364b38_1  conda-forge
  openff-toolkit           0.18.1     pyhd8ed1ab_0  conda-forge
  openff-toolkit-base      0.18.1     pyhd8ed1ab_0  conda-forge
  openff-toolkit-examples  0.18.1     pyhd8ed1ab_0  conda-forge
  openff-units             0.3.2      pyhd8ed1ab_0  conda-forge
  openff-utilities         0.1.16     pyhd8ed1ab_1  conda-forge

[mattwthompson]

IMHO the tradeoffs around using AmberTools in those notebooks are negative