Do you mean which distribution to select? Or, how do I derive numbers to fill in for the parameters of my prior distribution? For the latter one the article below might be of help:
Article Where Do Priors Come From? Applying Guidelines to Construct ...
A search through literature might obtain values that can be used as prior information. Preferably people would give their estimations (point and interval), which harldy happens. If I fit i.e., y~B0+B1X1, I try to find figures, numbers or descriptive information from literature from which I can extract information on the parameters. Otherwise I also use personal data on the topic that floats somewhere around in my dropbox or github.