Since this is my first post here after years of using DEVONthink, I’d like to start by thanking the entire team behind DEVONthink and DTTG. DEVONthink 4 is a great release and, in my view, a fully worthy successor to DEVONthink 3. I say this as a daily user - usually keeping DEVONthink just one Cmd-Tab away.
One of the features I value most is DEVONthink’s scripting support. However, since upgrading to DEVONthink 4.2, I’ve noticed a significant regression in AppleScript performance. More specifically, AppleScripts now appear to take considerably longer to execute than in earlier versions (including 4.1.1 and prior releases).
For example, iterating over approximately 6,500 records (about 99% PDF or EML) now takes nearly 5 minutes in 4.2, whereas the same script, run against the same database and data set in 4.1.1, completes in about 8 seconds. Based on my observations, the slowdown seems to occur when accessing record properties such as the name or record type.
I’d like to ask whether this is a known issue - either within DEVONthink itself or in the surrounding ecosystem (I’m running macOS Tahoe 26.2) - or whether it’s more likely caused by something on my end.
I’d appreciate any feedback or insight on this. Thanks a lot!
@BLUEFROG sure, no problem. I’ve attached a small script where the issue is isolated along with corresponding execution times on 4.2 and 4.1.1 and the database structure/outline. Please note that execution times varies to the times mentioned above but ratio is still the same.
@cgrunenberg this is the way the properties are accessed. I do not use / access all properties, maybe a handfull (like name, record type, tags).
I’ll have to test this on Tahoe too but at least on Sequoia the performance of both versions is identical (as expected). How exactly did you run the script? Via the Script Editor.app? The Scripts menu of DEVONthink? Or the global scripts menu extra?
On Tahoe I could reproduce the issue. But a sample of DEVONthink while executing the script didn’t reveal any real bottlenecks, DEVONthink used only 10-15% of one core. Closing all windows didn’t make a difference too.
Nothing in DEVONthink’s scripting support depends on the used macOS version. Seems to be one of these annoying edge cases where the usage of the latest SDK causes unexpected side effects.
Indeed, maybe one of these edge cases, but good to hear that it is also reproducable on your side - and to know it is somehow related to Tahoe.
I usually work with scripts from DEVONthink’s Scripts Menu - launched through Keyboard shortcuts (which works fantastic btw). The provided script was launched directly from Script Editor.
Regarding your observation on CPU time, other windows or apps - same experience on my side.
If you have any news on this matter, please let me know - will do the same. I’ll stick with 4.1.1 for now and see what happens with next updates. Fingers crossed.
The rewritten TCC (Transparency, Consent, and Control) framework of macOS 26 seems to slow down Apple Event handling a lot in case of apps using the macOS 26 SDK. Some apps experience AppleScript timeouts or failures, others (like DEVONthink) poor performance.
The worst thing about macOS 26 is not Liquid Glass but the overall poor performance and unreliability If Apple would at least for one year stop adding features and instead fix the most pressing bugs and bottlenecks.
If Apple would at least for one year stop adding features and instead fix the most pressing bugs and bottlenecks.