Here is one approach. We can first reverse the input string. Then, do a global regex replacement on the following pattern:

(\w)(?=.*\1)

This will strip off any character for which we can find the same character later in the string. But, as we will be running this replacement on the reversed string, this has the actual effect of removing all duplicate letters other than their first occurrence. Finally, we reverse the remaining string again to arrive at the expected output.

var input = "abczacg";
var output = input.split("").reverse().join("").replace(/(\w)(?=.*\1)/g, "");
output = output.split("").reverse().join("");
console.log(output);

An alternative without using a regex using a Set:

[
  "abczacg",
  "aabbccdd",
  "hello world"
].forEach(s => {
  console.log([...new Set(s.split(''))].join(''))
})