Hi, I find various different ways to load credentials and pass to executing nodes (e.g. using hooks, or custom input datasets). What’s best practice here? Any advice?

I use credentials in the catalog because I need them only to load or save data. Therefore, I define them in the credentials.yml file, which is taken by the catalog. I also usually use the oc.env omegaconf resolver when the credential is in a env variable. I do not know if it is a good practice to use it as arguments in a node, maybe you can create a custom dataset if your objective is load or save data in that node

cool, thank you. My use case is to authenticate with the datarobot client to interact with it inside nodes. A

after_context_created

hook seems to do the trick well

yeah the catalog hook is for enterprise solutions which provide credentials in environment variables