I don't understand - what's the point of the parameter if 0 is the only valid value?

So this comes from that assert in LLVM. AFAIK the idea is quite simple: single enum attribute cannot be parameterized and fully defines semantics of it; integer attribute can be parameterized and its semantics without parameter is not fully defined.

Not really sure how inkwell got away with it unless people built LLVM without asserts

But it's provided by the C api:

Anyway, if there are no other valid values we should remove it from the rust side entirely as a safety measure and add comments about it

But it's provided by the C api:

inkwell/src/context.rs

Line 351 in f88ccb2

unsafe { Attribute::new(LLVMCreateEnumAttribute(self.0, kind_id, val)) }

As I can see, that API didn't change since 2016 and assert was added way later.

Anyway, if there are no other valid values we should remove it from the rust side entirely as a safety measure and add comments about it

I don't think we can easily remove it as there's no C API to build IntAttr, TypeAttr etc.
The worse part is there might be bad naming that was carried around: for example, there is LLVMGetLastEnumAttributeKind function that one might expect would return LastEnumAttr that is generated from tablegen file, but in reality it returns EndAttrKinds that includes EnumAttr, IntAttr, TypeAttr and ConstantRangeAttr.

That said, we may need to keep it and try to add some convenience C functions to either split LLVMCreateEnumAttribute into several functions such as LLVMCreateEnumAttribute, LLVMCreateIntAttribute etc or simply return nullptr if value cannot be used with attr kind or have LLVMGet[First|Last]EnumAttributeKind C Functions to be able to split inkwell's API into create_enum_attr(kind) and create_int_attr(kind, val)