I think really all you need to do is start using it. The HTMX api is minimal but powerful. At the end of the day you are requesting chunks of html that replace existing sections in your html.
You are building an HTML driven app so your auth workflow is going to be whatever HTML can handle.
In a world of either SPAs or SSRs, htmx straddles the two. That is what makes it so interesting and unique, but it is also a different enough model to not always cleanly map from my pre-existing skills/patterns.
For the app I'm doing, I'd really prefer to keep it an SPA. In that the URL never changes, and there is never a full page reload (outside of OIDC/sign-in). I'm sure I can get there, I just need to think about the architecture a bit.
I'm kind of obsessive in this app about loading the app being nearly instant (I personally use it countless times a day). I really love the idea of returning back a full rendered page of HTML that wouldn't require any other API calls before showing the content.
The current stack (which I hate, but it works, I wrote it, and so who am I going to blame?) - is PHP/MySQL. My concept for something that will be lower maintenance going forward (keeping up on PHP versions...ugh.) is to use Deno + oak + SQLite (running in Deno using WASM) on the server to render HTML fragments/pages, and htmx on the front end. With no build step and no database server to maintain.
On a side note about htmx. Really compelling concept, I wish there was a larger community around it. Chicken and the egg kind of a problem, but there just aren't a lot of resources (blog posts, videos, etc) compared to many other web frameworks. Basic stuff like how do integrate something like Auth0 auth into an htmx app.