Okay, after some quick searching, Stingray gained a ProceduralMesh class/type in 1.6
So if thats there, then this should do everything I need.

Ref: https://forums.autodesk.com/t5/stingray-forum/version-1-6-and-procedural-meshes/td-p/6718859

So wait, is the lua library just hte interface for stingray with some modifications?

The big questions for me is mostly what api calls are available in the space.
Are we going to get documentation on the exposed api, even if its a simple list of calls and what they basically do?

Okay, thats hopeful! It's unfortunate, but ultimately not surprising that I can't use AO, I figured there had to be atleast some amount of baking being done in the level building process. If its only AO, it should be easy enough to cheat by limiting the types of scenes being made to avoid obvious loss of AO.

I can't compete with Umbra!! *cries*
Umbra is conceptually simple however, so yeah I could probably work something out thats a probably functional (though almost certainly less efficient)

This is all very promising, it sounds like I wouldn't lose much.
Are VT levels all just models, is there no special level geometry being built via BSP or anything like that?

Also, you said "not at the mesh level very efficiently" do I take it that a mesh could be constructed at run-time, but it probably isn't usable for this endeavor?

Well, I suppose I'm being a little presumptious. Perhaps the level system is all based upon models, and if so, I'd imagine the lighting system works fine with it. I didn't get that impression though.

While you could certainly approach it the way you suggest, I suspect there will be loss of visual quality.
Vermintide 2 is a beautiful game, the models, textures, and lighting are all on point.

It would be ashame to lose that quality for dynamic levels is all.

Also, consider this my official request that you guys investigate if it would be possible to allow run-time level generation as part of the modding api. I understand if its not possible though, its not small undertaking, and even in engines I play in there are limits for what you can do with level generation.

I wouldn't be surprised if the level generation required a series of baking steps, which wouldn't be accessible from the game engine itself.

If its simply not possible to ever allow it, then I am sad, but I can look at these other lesser approaches and see how much of it would be viable.

It's a good idea, the question about efficiency is an important one.

Whether or not the AI can handle a dynamic environment for pathing is a good question.

The problem with level randomization like L4D2 is that it still felt the same, I never noticed the different paths to any significant degree.
It's certainly worth investigating.

So you could in theory then, use that to build simplified levels, but I'm guessing you'd be severely limited on things like lighting which is probably a show stopper in itself. It would certainly defeat some of the advantages of trying to generate levels for VT I was hoping for, like the excellent rendering and lighting systems it has.

Well, this dims my hopes a little, but I'll certainly be keeping an eye on mods and seeing what others are doing

Thats excellent. I'm a little crazy, so I'm curious if it would be possible to generate levels at run time.
Is that stuff too deeply nested into the engine to get into via lua?

If I can create a mesh object from a set of points, I can probably go ham from there

So the whole game source will be available, or do I misunderstand?