Admittedly, I have a big database, with lots of indexed directories. But when I switch to DEVONthink while I’ve been away from it a while, I’ll have a 'busy ball’ spin at me for a second or so. I’m not sure what’s causing it - I switched off a bunch of Smart Rules and Smart Groups, thinking they might be creaking under the weight, but that’s not the case. I suspect the rules might still be running, but then not applying their actions as I’ve disabled them via ‘Disable Actions’ from the context menu - is that the case, and could that be creating wasted computation?
Is there any way to help diagnose where DT is spending it’s time and what might be causing these minor delays? (I’m still broadly using DTP4 as thought it were still DTPO2, but with a much bigger database, and under Sonoma)
My first guess would be that it’s caused by virtual memory and that DEVONthink has to be swapped in again after switching the active app. How much RAM does DEVONthink use (see Activity Monitor.app) and how large are your databases (see number of items/words in File > Database Properties)?
In DEVONthink, I’m seeing 150,000,000 words in a 28gb database, 31,000 items (9,000 indexed). In Activity Monitor, 2.61gb RAM with 1.26gb in VM compressed. In the ‘memory pressure’ graph, it says ‘Physical Memory: 48.00 GB / Memory Used: 38.29 GB / Cached Files: 10.08 GB / Swap Used: 4.35 GB’, so even with Ollama hiding in the background, I’d hope there would be enough leeway. I’m actually surprised how little RAM DEVONthink is using.
I had a 19gb Gemma4 sat there: I just flushed it out and checked the memory pressure stats again (Physical Memory: 48.00 GB / Memory Used: 38.80 GB / Cached Files: 10.19 GB / Swap Used: 4.35 GB) - not seeing much change - and quit/restarted DT - which has resumed at a use of 1.12GB and compressed of 0.
I’d not really be expecting to be using much swap on this machine, TBH - could accumulated peak consumptions have caused that 3GB swap usage? DT hasn’t given me the ball in the first few minutes, anyway, I’ll keep my eye on it (and use lighter models)
I’m also working with a lot of PDFs - could this be related to a magic memory leak in Tahoe?