Hello all, My team was discussing hooks (API endpoints, webhooks, etc) that are available for integration with Metaflow. Essentially we want to build something in the future that pulls certain artifacts created during experimentation on Metaflow based on some selection by the user and put it in some external model registry. Is this the only way to do this? --> https://docs.metaflow.org/metaflow/client

you can easily retrieve datasets and models from the datastore and use them to serve predictions -> an example here of a streamlit app that visualize results from a metaflow pipeline ->

<https://github.com/jacopotagliabue/MLSys-NYU-2022/blob/733bff0c31422e8142c4ceeaf3abdb50b34d3de1/weeks/10/app/app.py#L37>

Metaflow also has native integration with experiment tracker (Comet, Neptune etc) if you prefer to use an additional service for assets and artifacts

yep, for pulling Client API is the way to go. Another option is to introduce your own custom decorator

@my_model_registry

which pushes the desired information to your systems

Thanks so much @worried-mechanic-36312 and @straight-shampoo-11124

you need to have a metaflow config setup - you can use

metaflow configure export

and

metaflow configure import

to save and load it across users (or just copy

~/.metaflowconfig