I will look into how to make a package for MSYS2.
Regarding the other two options:
1) use only sipyco: in this case, I run the controller independently and, within the experiment script, call the sipyco client to call methods in the target, correct?
2) the idea is to make a new python environment with pip and install the packages I need, and then as a command in my device_db use a script that activates the environment and run the controller?
I use MFLI from Zurich Instrument--the API is in the zhinst
package (they also provide C API and in principle I could use that I guess, but I have already quite some code written in Python for this instrument that I would like to reuse 😃 )