I had it with Python but that is valid as there might be some edge cases that would have been left by you to check for.

I cannot see various codes, because except complaining, you did not bother to share the actual code which does not work for you. All I could do is to check your most recent code (which I can sometimes do), and I saw that it does not conform to input specification of the task.

I can also see that you completed the kata, what I presume means that you found your mistake. However I still do not understand where the "works perfectly in GHCi" came from, as it did not handle valid inputs at all.

I have no idea how you can see the various codes that I entered

We can just do that, after we ourselves had solved it. Mods can see it in even more detail.

why you doubt me?!

Why the heck do you doubt the tests?! How could it be approved and sit for years if the tests were wrong? This isn't even factoring that there seems to be a real issue of no random tests available (but that's beside the point).

this code running in GHCi, all correctly

We hear this every day, and at least 9/10 times it isn't "all correctly", or it doesn't conform to CW's interface. Your local environment is not the same as codewars environment.

one version with error handling

Were you asked to do error handling for this kata? No, it's only valid inputs. So, don't worry about handling it. For this kata, monads shouldn't even come into the picture.

It seems that you're overcomplicating the task and trying to involve I/O ? Just focus on function signature at hand and keep things pure.

Your code is not correct, and I honestly doubt it works in your GHCi correctly.