I wanted a quick way to view stats across my cluster running sparkrun (currently with minimax 2.7) and it evolved into a little more.
On the front page, unauthenticated, you can see all the statistics for each node in the cluster and what the cluster is running. If you log into the admin section you can
-Start/stop/restart the current recipe
-Load a new recipe (it pulls in a list of all the populated community recipes that your current cluster can run)
-Pre-load a model from a recipe (or something else) from hugging face and pre-sync it across the cluster
-Do a basic chat with the model that is loaded as a test
-Setup an api token to control it across the network
-Setup your own TLS cert
It’s not as full featured as some of the things I see here but I wanted something super lightweight that literally just controlled spark run across my nodes.
https://github.com/brainchillz/sparkdash
and optionally, if you setup the homelab fleet manager that I made
https://github.com/brainchillz/NexusController
you can monitor/manage/get alerts for a bunch of clusters
or lots of other things like Synology/truenas/zimaos/openmediavault/proxmox/vmware esxi/vcenter or any host running my personal dashboard or any windows or linux VM running
the tiny python monitoring agent that ships with with the controller.