co-log VS hnix

Another great Kowainik project to suggest is co-log: https://github.com/co-log/co-log

We use co-log it is simple and extensible, and can easily be integrated.

That being said, the fact that you're trying to define an instance of HasLog for ActionT in the first place is a bit suspect, since this co-log tutorial suggests you shouldn't need to do anything of the sort. I think you should be able to perform logging in your App and then lift the App to ActionT by using ActionT's MonadTrans instance. That is, something like this:

I can refere to prelude setup in HNix

I also can nominate HNix: https://github.com/haskell-nix/hnix. Nix is a lazy pure language, HNix is an implementation of Nix in lazy pure language (Haskell). The project was done by a famous expert in the field of compilers/interpreters, which worked on Borland C++ compiler back in a day. The project overall was done by a pretty start team.

Well, maintaining a rewrite of Nix in Haskell: https://github.com/haskell-nix/hnix.

HNix is a Nix (pure lazy language) interpreter/tooling in Haskell. So it is a field&case of pure-to-pure, lazy-to-lazy language creation. HNix is created with fancy design paradigms (design introduction): as recursion schemes (since Nix is a recursive language) & Abstracting Definitional Interpreters. Mentioning these two particularly, as they align HNix to GHC in a natural way, design leverages & projects GHC power to HNix, allowing to concentrate on language implementation.