DevonThink behaviour for archives (.dmg, .zip, .iso)

Dear all,

I wanted to create DevonThink database that would be dedicated to classifying tons of software links and installers. Ideally, my favourite feature of DevonThink called “auto-classify” would then allow me to simply click and my indexed downloads folder would instantly become orderly, filling up my database. I understand that this behaviour works mainly for documents where DevonThink can actually read the content, but why shouldn’t this work the simplest way based solely on file names? I mean a new version of a software installer generally keeps the same filename, only changes the version number. Unfortunately, it seems that DevonThink has no classification algorithm for .zip or .dmg. In fact, DevonThink seems to treat .dmg or .zip as files with no information regarding classification and hence does not even attempt to suggest appropriate groups, where similar files already exist. It will only show similarly named files in “see also” but auto-classification does not work. Any suggestion how to make auto classification work for .zip and .dmg files? Many thanks for any suggestion,


Hello again,

from the lack of response I understand that auto classify by “content only” is a feature and nothing can be done to make it work for archives based on the filename?



That’s correct. The AI assistants ‘look’ only at the readable text content of documents, not the names of documents. DEVONthink cannot see text inside .zip or .dmg files.

Thank you for your answer. Could you recommend a way to do this via a script?

The DEVONthink scripting dictionary provides only basic “Classify” and “Compare” commands using the same AI engine as the regular interface, and so scripting would provide no different or deeper features than what Bill described.

If anyone is interested, I solved the problem with a user-made script available here:

It allows me to “classify” files into folders (i.e. groups) by typing the name of the desired folder (or indeed writing the first few letters). Shortcuts can be defined for user convenience. So in the end it’s a step further from doing this manually in Finder, even though still miles away from auto-classifying and entire folder.

Needless to say, I find it very hard to understand why such a simple feature as name-based classification is not included in DevonThink AI (even as an option for certain filetypes for example).

This might be helpful: