At the moment there are just two commands available in the tool, but there are already some drawbacks from the initial design around commands. Mainly:

• There is no decent way to unit test the output being printed to stdout/stderr.
• The init command handles the output within the wrapper function, while the server command will produce message and status for wrapper consumer. This is a bit inconsistent.
• Both init and server commands have its own logic behind returning status and error messages.
• Testing exit codes is from different commands is cumbersome and lacks consistency.

The following should be done:

• All command implementations (from the commands module) should accept the standard output and error streams as function parameters. This would help in two ways:
  • Commands can take care of all output, especially since commands "have better idea" of what the output means.
  • It will be easier to unit-test the command output - since it would be as simple as providing an instance of io.StringIO.
• In turn, all output handling should be removed from the cli module, and the wrappers and main would only take care of passing-in the sys.stdout and sys.stderr.
• All commands should return an exit code. This would make it simpler to unit-test exit codes.
• Tests which are manipulating the sys.argv should be restructured/rewritten to adjust for the new layout. This would probably mean moving a lot of tests into test_commands.py.
• Tests should be updated to ensure that they do not accidentally use non-temporary dedicated directory when generating artifacts. There might be a couple of those that are misbehaving.