equinor
Align ert and everest installation and validation of forward models
Currently ert and everest have slightly different ways of validating the forward models, in everest there is an option for the forward model to supply their pydantic schema, which can be validated by everest: https://github.com/equinor/ert/blob/398ce6ad2861eb3489dfce7965e58df30d0d2379/src/everest/config/validation_utils.py#L284, while in ert, we have two optional functions which can be implemented to the forward model: https://github.com/equinor/ert/blob/9f8d73a6b4809a0384bda60632350d2123df7c0c/src/ert/config/ert_config.py#L603.
This might be an epic, in which case we should produce sub issues and track them in this issue.
The best approach is probably to modify the everest hooks to use the relatively new functionality in ert.
To make use of the ERT validation code, it seems that forward models need to be added as preinstalled forward model steps. Everest uses currently the INSTALL_JOB keyword, hence the first step is to make everest-models use the plugin system to add its jobs to ERT.
It turns out that the validation of the forward models requires knowledge of the data that Everest wants to install while running the steps of the forward model. This knowledge is available in the Everest configuration (install_data), but is lost when the ERT configuration is created. Since the ERT forward model classes don't have access to the Everest configuration, they cannot run the validation which requires access to installed files.
It seems therefore that validation of the forward models at the level of the Everest configuration is in fact appropriate, and should probably remain as is.
Will close this issue, as the methods are different enough that this is not really possible.