ktx2 = textures
drc = 3d shapes (I think)
ogg = audio
All of these would normally be bundled in the game installer, but are sent down piece by piece over the network in this case.
Then there's some wasm and js for the game's business logic. The browser has WebGL APIs that enable running all of this.
I'm assuming they used a library or engine like Unity, Godot or Three.js that supports WebGL as an execution target.
The NPC at the top of the building sais it's three.js
Definitely Three.js or at least something similarly low-level. I doubt you can get this kind of performance with Godot or Unity on the web.
There is Needle.Tools for porting Unity projects to WebGL/3js