[bug] Error Installing high_quality_translation Validator

[OrenRachmil]

Describe the bug
I encountered multiple errors while attempting to install the high_quality_translation validator from Guardrails Hub, Below are the details of the issue.

Steps to reproduce the behavior:
guardrails hub install hub://guardrails/high_quality_translation

Expected behavior

Failed to install brainlogic-grhub-high-quality-translation
Exit code: 1
stderr: error: subprocess-exited-with-error
  
  × python setup.py egg_info did not run successfully.
  │ exit code: 1
  ╰─> [17 lines of output]
      Traceback (most recent call last):
        File "<string>", line 2, in <module>
        File "<pip-setuptools-caller>", line 34, in <module>
        File "C:\Users\RachmilO\AppData\Local\Temp\pip-install-wh6bfpm6\sentencepiece_4f30feedef6b4bcab8112afb4b33d30e\setup.py", line 126, in <module>
          subprocess.check_call([
        File "C:\Users\RachmilO\AppData\Local\Programs\Python\Python312\Lib\subprocess.py", line 408, in check_call
          retcode = call(*popenargs, **kwargs)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^
        File "C:\Users\RachmilO\AppData\Local\Programs\Python\Python312\Lib\subprocess.py", line 389, in call
          with Popen(*popenargs, **kwargs) as p:
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^
        File "C:\Users\RachmilO\AppData\Local\Programs\Python\Python312\Lib\subprocess.py", line 1026, in __init__
          self._execute_child(args, executable, preexec_fn, close_fds,
        File "C:\Users\RachmilO\AppData\Local\Programs\Python\Python312\Lib\subprocess.py", line 1538, in _execute_child
          hp, ht, pid, tid = _winapi.CreateProcess(executable, args,
                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      FileNotFoundError: [WinError 2] The system cannot find the file specified
      [end of output]

Library version:
guardrails-ai 0.6.0

Additional context
Python Version: 3.12
Windows 11

[CalebCourier]

Hi @OrenRachmil, while not obvious from the error text, I believe the root issue here is a transient dependency of this validator requires cmake to build its wheel. Installing cmake resolved the installation issue on Linux and MacOS and yield a different more details error on Windows. I'm still working through the specific code path necessary to successfully install this validator on Windows and will report back findings here.
I've also filed an issue with this validator's authors so they can update their documentation accordingly: BrainLogicHub/high_quality_translation_validator#6

[CalebCourier]

@OrenRachmil, to successfully install this validator on windows I had to setup multiple c++ build tools including cmake and g++. I think the easiest way to get all the necessary tools installed and configured correctly is to install Visual Studio Community edition along with the "Desktop development with C++" workload. If this seems excessive, it's because it is. The culprit is a package from Google called sentencepiece that unbabel-comet (a dependency of this validator) depends on. An potential alternative to setting up these build tools, is to pre-install a wheel as described here before installing the validator.

It should be noted that this issue is only present in sentencepiece 0.1.x which unbabel-coment has pinned in it's pyproject.toml. The unbabel-comet repository does have a user submitted issue to support sentencepiece 0.2.x which does not have these install requirements.

[OrenRachmil]

@CalebCourier Thank you very much for your prompt response and effort.
I will try to reinstall by following your guidelines.
Thanks again

[CalebCourier]

One last note on this issue:

It appears that the additional dependencies (C++ build tools) are only needed in versions of python 3.12 and above since there are wheels published for the python versions below this and thus the wheel does not need to be built on the client machine.

This leaves an additional option of downgrading python to 3.11 or below to avoid needing to setup the C++ tools mentioned here like cmake, g++ etc.

Since there are workarounds documented here, and the core issue is not with our library or even the validator developed by our partners, I'm going to close this issue for now.