Zooming is available e.g. via the toolbar of the help viewer.
I’ve recently made the upgrade to Sequoia. Stayed with DEVONthink 3 to make sure everything was in working order and all the mission critical stuff is where it should be. I do complete wipes of the system drive and reinstall all software instead of just updating, so it takes a little longer but I learn all the things I have installed over the last few years and get to prune and tidy up.
This 4.1 feature release makes me think it would be possible to upgrade straight to DT 4 without doing the “run it as a parallel test in a different user account” thing. I have a complete CCC back up of the pre-Sequoia and the current system drive, plus an Arq cloud back up so I think I am in a pretty good place.
Side note: I recently figured out how to get CCC to properly close (not kill) Devonthink before running the clone of the system drive and then re-open after it runs. One less thing to worry about or have to remember to do.
Just wanted to congratulate the team on what seems a wild release. It’s incredible that under V3, AI was some keyword-detecting yet deterministic mechanism, and that on V4 AI has become an auto-programming layer for extending DT through tensor models for which ‘randomness’ is a parameter. Eastgate and Tinderbox also had a similar epiphany and went from being critical of AI (which is very justified in places) to seeing it as a similarly first-tier agent to interact with its environment, though Claude’s MCP - I hope these systems work their way through into Ollama (and run within the ram limits I have!).
I’m clearly going to have to buy a new Mac soon: I’m getting tired of not being able to update to 4.x and use all the new goodies.
Neat. Could you elaborate on that?
I got the two shell scripts here
near the bottom of the page but the “quit” one used killall which was too brutal. With some help from stack exchange and CCC support I was able to change the quit script to
#!/bin/sh
app="DEVONthink 3"
### No edits required below this line ###
osascript -e "quit app \"$app\""
Add them to the advanced settings “quit” for preflight, “open” for postflight and it runs them before and after the clone process.
There’s some security warnings on that page to use a utility in CCC to secure the Scripts folder that you should run after you’ve copied the scripts into the app specific script folder.
FYI the script for the open one is
#!/bin/sh
app_name="DEVONthink 3"
### No edits required below this line ###
sudo -u `stat -f %Su /dev/console` open -a "$app_name"
Thanks for the kudos but also be aware: access to external AI has not replaced DEVONthink’s own internal AI. It is still a very powerful core component and outshines external AI in many ways. It also continues to improve, even to this day.
And if you haven’t read it: Getting Started > AI Explained in of the built-in Help and manual.
What’s the sudo needed for?
Anyone got the “OpenAI (compatible)” to run with an Azure OpenAI Instance?
Can’t figure out how to enter URL to my instance to make it work…..
Exactly. This here works just fine as a shell command
osascript -l AppleScript -e 'tell Application id "DNtp" to activate'
And to stop a bit more gracefully
osascript -l AppleScript -e 'tell Application id "DNtp" to if it is running then quit saving "yes"'
See…
The intricacies of this command are beyond my knowledge. I will defer to Mike Bombich of Bombich Software and the Carbon Copy Cloner support website.
I tried it with this script:
#!/bin/sh
### app_name="DEVONthink 3"
### No edits required below this line ###
osascript -l AppleScript -e 'tell Application id "DNtp" to activate'
and while that last line worked fine in the terminal, when I ran it in CCC I got this:
The backup task completed successfully, but CCC encountered errors while performing ancillary tasks unrelated to copying files.
The post flight shell script exited with a non-zero exit status.
Verify that you can run the script manually without errors, and that it has a shell interpreter line at the top of the script (e.g. “#!/bin/sh”). Add “exit 0” to the end of your postflight script to suppress error reporting.
I’ll stick with the vendor supplied script for now as it works for me.
EDIT
Got an answer on the sudo question from Mike Bombich:
The sudo invocation is actually the opposite of what most people would think. Rather than using it to execute another utility with elevated privileges, we need to use it in this case to execute another utility with reduced privileges (and more importantly, in a separate user security context). CCC runs its pre and postflight scripts as the root user, so typically when you want to interact with an application running in the logged-in-user’s security context, we need to call sudo with the
-uargument to run the command as a non-root user.
Unfortunately my settings are broken with »Europa«. Opening DEVONthink and clicking on the settings menu item – app crashes. Tried several times …
This is not the place to report crashes.
And please be patient as your support ticket just came in 11 minutes ago.
Pricing Model Last year I decided to go all in on Devonthink, but I held back because I worried about the sustainability of apps in the Mac space that are 20+ years old and that have key developers that might be considering retiring at some point. Then I saw the pricing structure of 4 and was relieved, as it was what I wanted - a revenue model that looked sustainable and flexible but not a subscription. It made me think the app has a future but would not leave me stranded.
Editor I am a PhD student and a pastor, so I write a lot, and I liked the changes to the editor. I now just write entirely in Devonthink in markdown. I was doing a lot of this before, but the changes just made it a bit better.
AI - I don’t use a ton of the chat AI in Devonthink now. I mainly use NotebookLM and Gemini. That said, I like the model-agnostic and future-forward approach of Devonthink. I am experimenting with the on device models to see what they might be useful for with automation.
PKM - I have moved from my Obsidian notes into Devonthink. I am very happy about it. I have found that search is so powerful I don’t spend much time organizing like I did in Obsidian. I think the AI-tagging will be good as well. I also like the updated graph view.
DTTG I am also in the beta for DTTG, and I like the look a lot. Big small thing for me is the ability to make the markdown templates. I am not sure if this is just me getting better at shortcuts or not, but I finally was able to create a shortcut that sends files to a “log” group and changes the title in a way to match my log (i am a Star Trek fan and like to have personal log of voice files and written entries). Small thing but made me very happy. I probably use DTTG over 60% of the time so the editor and templates are very important for me.

