Jordan has used the R build environment (the one used to create the data). I think this is fine, especially given that I've got some conda environment YAMLs going for the model code. I also think meddle offers a way to populate the build-environment with your current environment automatically?