Machine Learning URL is not a URL
When I attempt to save the administration settings I receive the error
machineLearning.url must be a URL addressI have the URL set to
http://machine_learning:3003
via an env var.
This happens if I hit save from any section.9 Replies
Can you use the service name instead of the container name?
I'm using docker swarm so the service name is
<namespace>_<service>
. I could make a network alias and try referencing that.Underscores aren't technically valid urls, which is why it is failing. We could probably loosen the validation, but ideally there is a proper url you could just use
You are saying the namespace is machine and the service name is learning? Did
machine-learning: 3003
not work?Okay, let me start over, I misspoke.
The namespace is immich so the service is
immich_machine_learning
, but docker inspect shows it has an alias of machine_learning
.
I'll try the full service name
Same errorYou can't use underscores, but I think using "machine-learning" will work still
The settings accepted it, but then microservices died.
Machine learning was working with the underscore.
I bet I could rename the service so it would be
immich_machine-learning
which would have the machine-learning
alias.Cool
Are underscores going to be unsupported going forward?
I mean, like I said we could make it less restrictive. You can open a PR to remove the url validation and have it be free form (string) basically.