This repository contains maintenance workflows for the Ersilia Model Hub.
The inspect_model.yml file defines the GitHub action which is responsible for running the inspections. Within this file, we have several different code blocks which report the result of each method in the inspect.py file within ersilia/publish. For each check, we have two seperate functions: one of which returns True or False indicating if the check passed or failed, and the other containing the details of a failed check. If a check fails, we create an issue with the details printed. We access these values through the inspect.py script in ersilia/cli/commands, which stores the results of the publish/inspect.py file as a json.
For example, we have in the commands/inspect.py file an attribute called metadata_complete and an attribute called metadata_complete_details. These are stored by calling the same metadataComplete function with a flag of either 0 (for metadata_complete) or 1 (for metadata_complete_details) to indicate which type of output we desire from the function. You can see this function metadataComplete(self, flag) in the publish/inspect.py file and observe how it conducts several tests on the metadata, both calling the RepoMetadataFile function and conducting several other checks on URLs stored in the model metadata. At each stage, if a test fails, we verify the flag inputted and return a False result if the flag is 0, and add information to our details variable if the flag is 1.
If you wanted to add additional checks, you would need to add them three places: publish/inspect.py, commands/inspect.py, and inspect_model.yml. First, create your function with the intended functionality in publish/inspect.py. Then, call that function from commands/inspect.py and store it in the json file. Lastly, access that json file from inspect_model.yml and report the results. Ensure that when you add code to inspect_model.yml, you adjust the exit commands as needed so as to not prematurely end the script.
The structure of the functions in all three of these places are pretty formulaic, and you should be able to pretty directly copy structures, especially within commands/inspect.py and inspect_model.yml.
The repo_info.json file contains a list of all of the repositories in the ersilia organization. For each repository, we store the last time it was checked as well as the last time it was updated. In the pickRepo.py file, we choose among the repositories which have not been inspected since their last update, selecting the repository which was last inspected longest ago and printing it. This is called from the inspect_model.yml file, and determines the input for the automated inspect runs. Note that you can still also manually inspect a model by running an action with the model name as the input. To maintain the repo_info.json there are several other helper files that may be helpful. The makeRepoDoc.py script creates the json from scratch, setting all most_recent_date_checked values to a default date of 01/01/2000 to ensure all repositories are marked as needing to be inspected. Lastly, the updateRepoDoc.py script can be ran to update all of the last_updated values and add any additional repositories if there are repositories in the ersilia os database that aren't in the current repo_info.json file. I would suggest running updateRepoDoc.py in the event that you've run out of repositories to check in the automated workflow runs.