Congratulations on the release of DEVONthink 4! The new version looks fantastic, and I’m excited to explore the enhanced features and improvements you’ve introduced.
As an avid user of DEVONthink for knowledge management and research, I greatly value the inclusion of AI features like the integration with Perplexity, Claude, and ChatGPT. These are powerful tools that bring significant efficiency to the workflow.
Looking ahead, I’d like to suggest expanding the range of supported AI APIs in future updates. In particular, emerging models such as DeepSeek are showing promising results and offer different strengths that might benefit users with diverse needs — including those focused on multilingual support, open-source preferences, or regional accessibility. I believe this would further strengthen DEVONthink 4’s position as a leading research and data management tool in the AI-enhanced productivity space.
And also, remember: DEVONthink has not become “an AI application”. Access to some external AI engines has been added as a complement to DEVONthink’s internal AI, in ways that support its core focus.
Thanks for the clarification — I understand DEVONthink’s core is not meant to be an AI application, and I appreciate the balance you’re aiming to strike between internal and external AI assistance.
That said, I just wanted to confirm whether my feature suggestion (supporting more third-party AI APIs, such as DeepSeek or user-defined endpoints) has been passed on to the development team. I fully understand if this isn’t currently on the roadmap, but it would be great to know the idea is on your radar.
Thanks again, and congratulations once more on the release of DEVONthink 4!
That said, I just wanted to confirm whether my feature suggestion (supporting more third-party AI APIs, such as DeepSeek or user-defined endpoints) has been passed on to the development team.
We are all either reading or responding here (and I am running to and fro elsewhere), so they will see what has been discussed while they sleep.
DeepSeek can be already used e.g. via Ollama, e.g. I successfully tested a version with 77b parameters on a Mac Studio. In addition, if an endpoint is e.g. compatible to the OpenAI API, then it should be possible e.g. to choose GPT4All in Settings > AI > Chat and to enter the endpoint’s URL. Of course this would work only for endpoints that don’t need authorization currently.
That makes sense — and yes, I’ve also tested DeepSeek locally via Ollama (77B on a Macbook). It’s impressive that it runs, but the performance and responsiveness are still quite limited for real-time use.
So I’d definitely appreciate future support for external APIs — especially if we could define custom headers or more. That would make it possible to plug in OpenAI-compatible services like DeepSeek’s cloud API and others.
Glad to hear this kind of flexibility might be considered — it would be a great enhancement for users who rely on larger models via API rather than local inference.
Yes, it is fully compatible with OpenAI API, here is the quote form their docs:
The DeepSeek API uses an API format compatible with OpenAI. By modifying the configuration, you can use the OpenAI SDK or softwares compatible with the OpenAI API to access the DeepSeek API.
I would very much like to be able to use their model with DT4.
I know there is a lot of controversy regarding them handling the data, but I think it’s up to me what to do with personal data.
But their price just blows every other competitor out of the water. I can throw huge amount of data into it and not worry about losing my vacation money while on reddit you can easily find articles of people getting 500 dollar bills after a couple hours of using Gemini.
I know I’m gonna buy DT4 when it’s out regardless but being able to use DeepSeek model would make it so much more exciting.