That is not how I personally conceptualise DEVONthink at all. For me it is a kind of library where I store research material and where I go to find things out. It’s a research tool, not a filesystem for me. Perhaps that difference in conceptualisation is the reason why there are such differing views expressed in this thread and why there are confusions and difficulties in understanding each other here. I see DEVONthink as being all about quick capture of material, ways of sorting and categorising it if I need to, and critically, finding material again. Added to which, the fact that searches will turn up similar material which I have forgotten about is very useful to me. That is not really something I get in Finder. Backup has never been a problem for me, so I don’t really have any reason to index material instead of importing it.
DEVONthink is my document management system. All of my documents go in DT. If I need to restore a file within the DT file structure from a backup of that structure, it’s like restoring any file with a standard backup tool. It’s worked fine for me on that rare occasion I need to go back and get an older version but DT versioning has helped with that. I don’t store a ton of custom metadata, so if it’s lost it’s not a concern for me.
I explained my setup earlier in this thread (and have discussed my thoughts in this forum before, but it’s a pain tracking down comments so I’m going to do it again!), but I want to expand on it a little more.
DT (the company) doesn’t consider its app a Finder replacement, but I imagine several of us use it like that. I certainly do. Your question of “what are the trade-offs?” is easy for me to answer since I don’t think there are any. But I don’t use Terminal. I bought DT because my file storage was getting unmanageable and I needed structure. For me, a file system wasn’t bringing that (otherwise I wouldn’t have had an issue that led to me trying to find a solution). It’s not hyperbole for me to say that DT changed my life. It gave me a place to store and interact with all my files, impose a system on them and created space for me to think. Finder doesn’t really do any of the things I use DT for (amazing search, window pane to view files in place, custom naming rules, sorting by different options depending on what I’m up to, writing and linking ideas, etc.). Embracing DT also led me down various paths with other tech things that has made my life much better (markdown, shortcuts and automations, etc.).
I don’t think most of us in this forum are “evangelical”, and we recognise when DT isn’t the answer to a problem, but often it can be and mostly we’re just trying to be helpful and hoping to help others experience the same benefits we’ve found.
For my own filing, my rule isn’t “Should this go in DT?”, it’s “Why wouldn’t this go in DT?”. DT is my default with its specific databases for life admin, all my research and hobbies, etc. For me my “why not?” question is mostly answered with:
- This file type belongs to a specific app and there is zero value in adding it to DT [GIS files, music files, photos]
- This file is specifically a work file that I need regularly [this is a bit fuzzier given that I use DT for my work, but it’s basically a distinction between ‘my work’ versus ‘files which belong to my employer that I need for reference’]
- This file is huge and I don’t want it getting caught in any syncing [GIS again, some app-specific files]
That’s about it for exceptions.
One other thing is that DT gave me much more visibility of what was happening with my files. Finder doesn’t provide that. So for example, if I bring a bill into my database twice, DT will flag that and show me that I’ve filed one copy in an unusual place. If I download an interesting academic paper and I’ve actually already read it and annotated (probably happens more often than it should 
) DT will reveal that and stop me from just ending up with papers duplicating across folders. Has something weird happened to a file? DT will notice and flag the issue at some point. Finder won’t do that.
Expanding beyond the simple “backup tool” example, another filter criteria for me is, do i need to share a document? My reality today is:
- my accountant shares files via Box
- my lawyer shares via Dropbox
- my work shares via Google Drive
- my family shares via iCloud
So by necessity I have files in all of those locations. And I’ve been able to standardize all of them using file system tools (Terminal, scripting, TimeMachine, Backblaze, Gemini 2 duplicate management, CleanMyMac, Disk Utility, etc). Since i’m using those tools for all shared files, it becomes an input on what else do I / do I not use them for?
These feel like good candidates to be indexed into DT.
Another example: I will prune my folder/document hierarchy by collapsing branches (e.g. tax returns from the 2000’s) into a .dmg or .asif file. Still available to mount if I need them, but otherwise not cluttering my database/finder and index/spotlight. I wouldn’t put a .dmg or .asif in DT, because to use it, I’d have to export it, and then mount it.
In those examples, I’m managing my PKM (including a large number of documents) in DT4, but I also have a significant volume of documents outside of DT. This creates some overhead in duplicative tools/management/workflows - this is part of what I’m teasing out here from others.
Apple Photos is my closest compare. I use Photos for all of my images, and trust it to manage its package, with its own metadata management, and I accept that I don’t need to manage individual photo image files. Based on this thread I’m realizing i could envisioning DT4 as the apple-photos equivalent for non-shared file system docs. (The Photos compare is close but imperfect b/c our family does use Shared Albums)
all helpful thinking from this thread.
Probably. As long as you make sure that the Cloud services do not decide to remove them from your Mac.
Many things can be imagined. I did never even consider managing photos (or music, for that matter) in DT. EXIF support is, AFAICT, only rudimentary in DT, IPTC is probably missing completely. Organising images into the equivalent of albums is difficult in DT, there is (I think) no support for Apple’s automatic motive recognition (as poor as that may be). And the DT manual even talks about this issue, saying that they do not suggest DT as a Photos replacement.
Photos is by no means ideal – very little support for automation, silo-ing its content away in SQLite databases with non-documented structures, no clear explanation about the relationship between their metadata and standard ones as EXIF/IPTC.
I was wondering about best practices for large files
I use a separate DT database called Holding For Large Files
cross-referenced with hyperlinks in my working database
A recent example is a 5GB .mp4 video file of my granddaughter’s school play
An additional benefit was DT generated a transcript using Apple AI
For VLF I have Finder folders that are named after my several databases, and sit in my Mac SSD. The files are not indexed, but referenced, like
[_DSC9935.NEF](file:////Volumes/DIGIDOINGS/.Trashes/501/_DSC9935.NEF)
in an md or rtf.
Has been working super well.
The files are sitting in the trash?
Keen eyes Christian! 
yes this one was. I normally don’t use the Trash as a repository (well, I do, very transiently).
DT hyperlinks have more options
For example, the generated transcription has timestamps
like 01:12:26 - Good evening, everyone
using x-DEVONthink-item://B80C1FA6-DABA-4069-ADD3-336B926B1E21?time=4346.220000
wow great to know thanks!