Download a model from within LM Studio. This is the most confusing part because it offers a lot of them. And I don’t remember what the first-time user experience is like. It might just suggest a model the first time you run it. If you get to a view like this (in Discover/Model Search) the top suggested result is probably fine to start.
Once you’ve installed models, then you can switch to using it in DEVONthink. In Settings > AI, you would choose LM Studio as your provider, and then your downloaded model from the model selection. It would look something like this.
Similar situation here. I put $10 on my account on December 11, and my current balance is $6.39. I also have a work account, so consider myself a moderate user of my personal balance.
I am just as puzzled as you, Sir/Ma’am. Again, I’ mgoing to offer to take full blame here. I don’t really understand the underpinnings of AI but when I purchase something with AI, I, perhaps wrongly do not automatically assume I need to buy another subscription to something. Again, maybe (probably) it’s in the fine print somewhere but I would have realized my mistaken assumption if there was (more?) disclaimers (something) telling me clearly that I need to buy another subscription to another service for AI to work. I’ve owned DT since DT2 but over a few months, I only move away from it. I was hoping the inclusion of AI would keep me in the DT community a bit longer.
I had purchased Devonthink licence back in August 2024.
When I logged into my account today, I found that I have been offered Version 4.0 upgrade, the the updates will be only till August 2025.
I find it nice that they are offering new version for older version that too purchased 10-months ago.
On a lighter note, I am still learning Devonthink and don’t think I’m using all DT3 features yet. No clue what I am going to do with all the new features in DT4
It certainly could have been clarified better but I don’t think that’s essential.
The nature of AI is that the more reputable the software vendor and the more capable the AI features, the more likely they will use an API model.
“Free” AI usually suggests either (a) The AI capabilities of the softare are fairly limited and thus won’t cost the software vendor much; or (b) The software vendor has some other way to make more money if you use the software longer i.e. advertising or selling your data.
I agree it’s annoying to not get credits with a subscription, though. It’s reasonable to think if you pay for OpenAI, you can authenticate and use that subscription with other apps. The UI never suggests you’re missing a funded API key unless you find your way to platform playground.
I don’t want to pay more just to have some credit for a third party software. I’d rather decide how much or if I want to feed the vendor. It’s optional for a reason
Indeed - the issue is more with OpenAI than with DevonTech.
What susprised me when this all started is that after paying for an OpenAI subscription I couldn’t even use their API to write my own apps - no less in 3rd party apps.
Their subscription vs API are distinctly different offerings. But this is not unique to OpenAI. Anthropic does the same. So does Google.
The DEVONthink Manual has a terrific explanation and discussion of indexing vs importing. Important to read and understand the manageable risks with indexing.
Sometimes there are discussion on the DEVONthink forum with lots of experienced people. Recommend you go their to see older posts.
Me … I import everything to gain max benefits of DEVONthink. But there are reasons for some to import but if those reasons do not apply, I would say don’t do it. See the DEVONthink Manual for more.
I’m using DEVONthink only to access archival reference material. Since it is unchanging I use indexing to save space. If I were to import, it would mean doubling the storage requirement without any advantage.
)