I’d like to have some tags. Not sure if I need them going forward, but I want to try some things. And I think I remember a smart rule condition “tagCount”, but I cannot find it anymore?
Can I set up a smart rule that checks whether a document has any tags? Like, say, I want to check if a document has less than two tags so I can ask the AI agent to suggest some more.
Also, can I bring the AI to suggest and later return only so many tags that in sum I have no more than five on any document?
Those are two different queries.
Item is Tagged = yes.
Item has two tags? No. This would require scripting or potentially using a Chat - Continue if smart action.
While I don’t think this is the best use of AI, this is possible though the tags are at the LLM’s discretion…
Here’s an example with an Apply Script action…
Thanks, that will help. Yeah, I agree that this is right now not completely helpful. Since DT also searches in the text, an AI agent generating tags from the text is offering no additional info.
My thinking is I want the AI to only suggest existing tags, and then check those before completing the task (and having the tags added). I will try to set this up today and see how it goes.
Another question: as far as I can see, within DT I can only use AI in ask mode, not in Agent or Edit, right? I would like my agent to edit documents in some cases when it asks me if it’s OK to do so. And I would like my agent to be able to read all my markdown notes (Zettel) in a certain group (my Zettelkasten). How can I give it access to that group, so it uses my notes as its context?
How many tags do you have? And if the AI should only suggest existing tags – how do these tags come into being in the first place?
Aside: Please open a new thread for a new question. Burying an unrelated question in a thread makes it (and possible answers) difficult to find for others.
My thinking is I want the AI to only suggest existing tags, and then check those before completing the task (and having the tags added). I will try to set this up today and see how it goes.
I would recommend you look at the AI > Tags inspector for suggested tags, powered by DEVONthink’s AI.
Another question: as far as I can see, within DT I can only use AI in ask mode, not in Agent or Edit, right?
Yes, that’s essentially correct.
I would like my agent to edit documents in some cases when it asks me if it’s OK to do so.
You’d need to enable the AI > Chat > Allow property & content changes.
And I would like my agent to be able to read all my markdown notes (Zettel) in a certain group (my Zettelkasten). How can I give it access to that group, so it uses my notes as its context?
You’re not giving it access in the way you’re likely imagining it. If AI > Chat > Search: Database is enabled, AI can query a selection or the current location, i.e., what is selected in the Navigate sidebar. So if you had that group selected in the Navigate sidebar, AI would see that as “your database”.
See Getting Started > AI in Practice > Controlling AI’s Access.
Thanks for these insights. AI>Tags is what I wanted. No need for the smart rule then.
I see that I have about 3.500 tags. In order for AI to help me tagging (like, for instance, Frederico Vittici does halfway down this article ➜ From the Creators of Shortcuts, Sky Extends AI Integration and Automation to Your Entire Mac - MacStories), I should simply reduce my sprawling tags.
My plan would be to read them out, put them in a spreadsheet and list a new tag for all the old ones (which in some cases would be just the old one). Then run a script to retag all documents so afterwards I’ll only have between 20-50 general tags. They would be a bit different than before, of course, more general and more likely to describe my sorting / file management ideas rather than the documents contents.
Does anyone have suggestions how to achieve this? Does anyone ever did that before?
You have it all mapped out already. Create a mapping from the old to the new tags and run a script to modify all the tags accordingly while you have a coffee. Or ten.
What suggestions are you after?
