Local Storage Card with local K8s

Hi, I'm currently running a k3s cluster locally, and I've been wanting to test a workflow that has a @card decorator to see some html and see it from the UI that is being served from K8s. The worfllow is ran locally. I am testing this without an s3 instance (that was giving me some issues with locating the /cards content when the UI was asking for them to retrieve from the ui service)

here is my ~/.metaflowconfig/config.json
{
"METAFLOW_DEFAULT_METADATA": "service",
"METAFLOW_SERVICE_URL": "http://localhost:8080/"
}
I've created a /.metaflow directory with no contents and chowned to my user to work

• the first run i make (Im testing this workflow ) has no .metaflow directory.
  This saves metadata about the run on /.metadata (a root folder owned by my user - not ideal but i want to mount this onto the ui service pod so that it can serve the contents later, i can do it from any folder but I haven't figured out how to have a consistent location for this yet
• second run: there is a .metaflow/ directory all of the sudden, and my data is split. in /.metaflow directory where i expect things to be has only a NBFlow sub directory, and my local ./.metaflow directory has the mf.cards I'm missing on the other one.

Now every subsequent run after this, metaflow defaults everything to my local ./.metaflow directory, but what i need is
I want to figure out a way to have a consistent way of setting my local storage directory and make sure the cards content also land here.
is there some combination of env variables or config i can add to my config.json?
also, if I were to instead just go back to my s3 setup, what configurations should i look at? I haven't had luck finding sources on this.

I must be something pretty simple here but I'm very new to metaflow, any help would be appreciated!

Start by reproducing the linked random_forest flow.py workflow with the shown ~/.metaflowconfig/config.json and inspect where /.metadata, ./.metaflow, and /.metaflow/NBFlow are created. Trace how mf.cards content is located by the UI service when running on local K8s. Done means the issue documents or implements a consistent local storage location for metadata and cards, with the relevant S3 configuration clarified if needed.