DT3 auto group/auto classify?

Anybody have any other suggestions on how to get this taken seriously? The Dev did respond once to my private message but it was pretty much as I described above…only a few users want this so “tough luck” in essence. No offer of any help at all.

I would suggest emailing rather - possibly even through the built-in help inside DT3. That way you will know that they received it. I think they are facing significant loads at the moment, and might not be checking PMs, since it possibly falls outside their CMS.

We’ve actually seen this thread and the PM and replied to both (see e.g. my & Jim’s replies above and sgmiller even quoted from my reply to his PM). There’s currently no alternative and whether this feature will return depends on feedback.


If feedback is needed, then I feel it was one of DT’s distinguishing features (ie using AI to help manage your data). It seems a shame that it’s gone and I would welcome its return.


I have already noted the replies…it was not the lack of response but the nature of the response that is the issue here. I can only speak for myself and will end just repeating what I have already said which is that this is, for me, Auto Group is an indispensable part of my work flow and I continue to find the seeming lack of concern for this of us who need it to be not understandable. I guess what I expected in response to a very long-standing customer is not “well not many people seem to use it so you are out of luck” but rather “we understand how important this is to you and we will try to find a solution.” Is that too much to ask?

Well I guess it is too much to ask…this is just really disheartening.

Received a PM from one of the devs…not sure why it was not posted here but since it was private I will only summarize:

Auto Group:

  1. wasn’t widely used

So what…when yu have so many features not all will be “widely” used. Some of us still depend on it.

  1. didn’t work.

Not true…just needs to be renamed “Group Similar”

  1. we have a huge workload impacting our personal lives

This blew me away …like the rest of us are floating on pool chairs on a tropical island? This threatens my actual livelihood which is why I am so angry

Still not a single word of understanding or any help offered. …unreal.

1 Like

Have you thought of solving the problem with what exists in DEVONthink 3? Smart groups, maybe tags, or metadata or scripts – or something? It’s only software – not worth the obvious stress – and maybe some other method will work. Someone here probably has a similar “work flow” – if not, then that sort of answers the question of “who uses this” – and some of us might have alternative answers to help you.

“Have you thought of solving the problem with what exists in DEVONthink 3? Smart groups, maybe tags, or metadata or scripts – or something?

None of that can replace Auto Group.

It’s only software – not worth the obvious stress – and maybe some other method will work.

With respect, If you don’t understand the “stress” then it’s clearly not critical to you but it is to me and others. Please tell me how I am supposed to take as many as 1000 results from Devon agent and quickly weed out the duplicates and similar items without auto group.? Maybe you would be stressed too if you were looking at untold hours of additional work on an almost weekly basis.In fact, I’m not even sure I’m capable of doing it without that feature.

Someone here probably has a similar “work flow” – if not, then that sort of answers the question of “who uses this” –

No idea what this is supposed to mean. There are others who have responded here who also use the feature and don’t want to see it eliminated. None of us can see how to replace it.

“ some of us might have alternative answers to help you.”

Ok…like what?

Yesterday, as a sort of experiment and because I’m thinking about a new research project, I tried to figure out a workflow that would at least partially replace auto-grouping.

I set up a search in DEVONagent which resulted in something like 350 “hits.” After archiving those hits to my global inbox, I created a new database with some preliminary groups and attempted to manually move items into those groups. It quickly became apparent that approach was going to take entirely too much time.

Then I attempted to set up some smart groups. But (1) it was harder than I thought it would be to devise the correct formula for the smart groups and (2) the resulting groups didn’t do what was needed – i.e., create groupings that were similar in their contents and matched what I needed for the research project.

I could have accomplished relatively quickly and easily what I needed if auto-grouping was still available.

So while I’m open to suggestions about a workflow that would mimic auto-grouping, I’d still prefer to have auto-grouping back in DT3.

So if the developers are listening: Please, pretty please bring back auto-grouping. Thanks.

IMHO, if anyone are so in need of the function, they sure can invest the time in making things work - perhaps even in a better way…
I have been experimenting with some dynamic smart auto grouping and have some success (still testing…). It involve “human” learning but can be more accurate and items can be assigned to multiple related groups, too.
(0) Use a new or duplicated database for testing!
(1) Set up a list of tag groups, or convert your groups into tag groups. If your group names are long, setup a short name in aliases field.
(2) Think about the keywords you need for identifying how items should be grouped. You may start with a few keywords for each group, but can increase the level of sophistication as you learn from the results of group assignment.
(3) Use a series of smart rules. Assuming that all files are first downloaded into a common folder (e.g., inbox), set the “search in” scope to that folder.
(4) Each smart rule will assign one or multiple tag groups (as tags) to the item according to the match in keywords or any complex criteria.

You can run the rules on demand, or trigger, by newly import item. OR, a really interesting way is to run the rules with scope being set to database (or the highest relevant parent group) on hourly/daily basis - as your “human” learning is deepened, your refined criteria/keywords will be automatically reapplied to all existing items and being allocated to the more relevant groups!

The above-mentioned solution is just one possible suggestion.
I have been a died-hard Lotus Agenda fan 3 decades ago and is now try to mimic the dynamic classification ability of Agenda in DT3. Perhaps my method will work, or NOT.

PS. It is true that setting up criteria could be time consuming, but I am guessing that for specific subject/topics human learning may be more precise. Not unless DT3 is really using some super-powered deep cognitive learning algorithm! (or perhaps DT is already using schemata of connected words but it’s going to be hard to mimic specific individual cognition process of info-connection?)


That’s helpful. Thanks.

I’ll give it a try.

This system does not address what I am talking about. As I explained, I routinely do searches which generate hundreds and hundreds of documents and the topics are different each time. If I had any idea about “the keywords you need for identifying how items should be grouped”, I wouldn’t need Auto Group in the first place. That is not going to happen until I can first whittle down the size of the document set by setting rid of duplicates and similar items and by quickly creating a rough structure. Only then would I have any idea how to start creating and assigning groups.

That is (or was) the beauty of DT which is that it accommodates the different needs of different users and what I do need is a “quick and dirty” way to pre-process hundreds of documents which is precisely what Auto Group now does. I highly doubt that any cobbled together workflow is going to be able to substitute for that.

Anyway, I have a solution for all of this which could easily solve the problems as the Devs have presented them. See my next reply down below.

Anyway, what they are doing makes absolutely no sense. This issue as it has been presented is that Auto Group “doesn’t work as expected”. The answer to that is not to eliminate it but to instead change the expectations. That could be simply done by renaming the feature “Group Similar Items” instead of “Auto Group” and explain that it does not “automatically” parse items into correct groups but rather groups similar items together through textual analysis, helping to find duplicates and/or create a preliminary structure. Surely that would eliminate the confusion that some users have with the feature.

Problem solved with no angry customers.


Thanks ProfJerry, you are describing exactly my situation here.
After reading @korm’s commentary above (DT3 auto group/auto classify?), I started to reproduce the feature somehow with existing tools.
With similar “success” as you: none.
But with the previous “Auto-Group” feature, it would have been pretty simple.
So, my feedback here is: Please, bring back a similar feature to Auto-Group and name it as you wish.
Many Thanks!

1 Like

Changing the name might seem superfluous but the devs keep saying it was causing user issues because “it didn’t work as expected.” Seems like a no brainer than to reframe the feature so as to avoid such issues.

1 Like

what about working with a second user account on your machine just for the sake of autogroup integration? on you 2nd account you have DTOP installed with all the gory auto/classify features Once youve gathered and sorted out your search results you close your DT database under your 2nd account, you simply switch to the actual DP on your main account (if necessary) and keep on working…

Shouldnt loose much time nor any feature set

just my 2c

Of course if I’m forced to I can do that, but it’s wonky and any wonkiness leaves open the possibility for errors.

As I have thought about this, it’s just so unnecessary, rename the feature something in accord with what it really does and then there are no problems. What is so hard about that?

Setting aside any design considerations, it might be non-trivial to re-introduce or re-activate the code alongside other functional changes.

I’d also like to see this back, but it might not be that easy. Even assuming it’s acceptable to the devs

From what I know about programming, that is not the issue nor have the devs claimed that it is. They have focused almost entirely on it “not working as expected “ which is my I am pushing so hard for the name change.

But glad to have another vote for bringing it back.