Not sure how feasible this is or how much LLVM provides, but it seems like there isn't a lot of information available about decoded instructions. It's for example not possible to extract arguments or even easily retrieve just the instruction mnemonic.
Because it's generic over multiple architectures it probably won't ever reach the level of detail that something like iced_x86::Instruction has, but a bit more than just the bytes and the entire assembler line would be nice.