In a support thread titled “Can´t set ai embedding model,” Wurzelseppi reports being unable to select an embedding model in the Discourse AI Plugin while behind a corporate firewall, mistakenly believing connectivity to generativelanguage.googleapis.com was required. keegan clarifies that the URL is only a test placeholder and advises configuring the embedding model via /admin/plugins/discourse-ai/ai-embeddings using the correct API key. Falco asks if Wurzelseppi is self-hosting the model internally. In post 4, Wurzelseppi confirms the issue was user error — they misunderstood the setup process — and now have successfully configured the embedding model, though they anticipate further issues and plan to open a new topic for those. The thread concludes with gratitude and a positive note.
Hey guys,
I’m sitting behind a very thick firewall in a company with my Discourse instance and want to enable embedding in the Discourse AI Plugin.
I can’t choose any model for embedding, though, and now found this:
It seems that there is a check against generativelanguage.googleapis.com to see if a model is eligible for embedding.
My instance can’t reach this URL, though…
Am I damned (fubar) now? 
Greetings,
WS
It’s not actually checking against that URL. That URL was just used as an fake URL example in our tests when we test configuring embeddings in our testing environment.
What do you mean by this?
Have you tried configuring your Embedding model with the appropriate API key in: /admin/plugins/discourse-ai/ai-embeddings? After you have an embedding model configured here, it should be available for being selected.
Are you self hosting your embeddings model on the intranet?
Hi,
That was entirely my fault.
I didn’t get how to configure the embedding properly, but now I’m over this step (and have the next problem, but I’ll start a new port for that 
)
Many thanks and have a great day!
WS