Dall-E seed parameter on API, +1
2 Likes
Not a model API question but a XYZ ChatGPT app question so may not even be the group doing the AMA.
Specifically will the Windows desktop ChatGPT app get an API to customize it?
Would really like to be able to automate copy/paste between a Canvas and VSCode while in a project conversation either via a manual trigger (button) or automatically. (other ideas)
How about APIs for other ChatGPT apps such as
- Mac OS
- iOS
- Android
In other words one could look at these apps as a precursor to a replacement/enhanced OS and having early access to an OS API is priceless for developers. 
1 Like
We would love to have better document processing capabilities. Lots of applications that e.g. automate answering questions from PDFs currently require additional tools (other than OpenAI) to preprocess the documents. Here some examples:
-
When a question needs context of both images and text (or just images), the model often fails because the vanilla RAG performed by the OpenAI API when asking a question about a document does not add the images in its context window, and therefore questions about images only really work when the documents are small.
-
Tables are processed reasonably well when the format is obvious, but it would be awesome if there was some option to extract them using Vision (e.g. scanned tables etc…), additional costs to use this option would be acceptable (in my case).
Also, although not a big issue, it would be be better if non-assistant endpoints (e.g. completions) could also take documents as input. Personally I consider ‘assistants’ as an endpoint made for interactive chats/use, and other endpoints for automatic processing use-cases, this distinction seems to be the most straightforward, and therefore both should be able to take documents.
Besides document processing, it would be awesome if vision capabilities get more intuitive ‘object detection’, ‘segmentation’ e.g. options (I believe currently it is possible to do this by prompting it to output bboxes etc, but it is not user-friendly, I would love something similar to the idea of forced json output, but then forced bbox|segmentation/depth outputs, etc… perhaps even with a seperate model, something like Microsofts Florence but then better.) If this is already possible, it would be great to have better resources about this (currently I havent seen any examples even though its a very large use case.). At last, it would just be great if spatial intelligence improved, and models could answer questions about maps better etc.
I would really love it if some of these features got added in the future, even just a few of them! Anyway, thanks a lot for all the great features that have already been shipped 
2 Likes
My question is why are there no human support for Open AI services?
Microsoft Startup benefits for eligible claimers since May and August are yet to be credited to the right people who got in and submitted org ID. Why is Open AI not transparent with the whole situation?
1 Like
Assistant API: citations with quoted text and bounding boxes of the quoted text.
2 Likes
Can we expect visual grounding for references cited by Assistants API, or visual grounding more generally?
2 Likes
A post was split to a new topic: How can ChatGPT handle current time automatically?
Could you please update the list of API Community Libraries? There are some that are outdated and archived.
For example, in Java, the listed library was abandoned months ago by its author, and on the other hand, there are more up-to-date alternatives such as: simple-openai (I’m the author) or Lambdua/openai4j to name a few.
By the way, I followed your procedure to add my library to the list several times, but it never worked and I never received any feedback.
3 Likes
Great to see everyone here and all the questions!
For this AMA, please keep questions related to the API and associated subsystems.
Thank you!
5 Likes
Ooh, nice one. I have also submitted swift clients to that form several times. I think the form is a dead end. Perhaps it was maintained by someone no longer w the company
3 Likes
This AMA is for questions for the API team to discuss today’s launches, but don’t want to leave you hanging. I don’t work at Microsoft for Startups, but I think their criteria for free credit grants is startups with verified funding. I.e., If you’re in the program, you’re eligible for the grant, and after your grant application is approved, you’ll receive the credits. Not everyone in the program will get free credits.
7 Likes
I guess a link to the youtube event would be handy for those without it:
https://youtu.be/XKABimtOWME (Just about to start)
After this the AMA will be here on the forum.
6 Likes
Is it possible to do both supervised and preference fine tuning on a model?
Can you do preference fine tuning on an existing model fine tuned with supervision?
4 Likes
Hi there, what micro controller was used for the fawn?
Thanks,
Kyle
2 Likes
Thanks for sharing the link. But is it me, or is there no audio?
1 Like
Is caching available in the assistants api?
1 Like
What was the most difficult task you’ve worked on so far at OpenAI? What is the most fun task you’ve worked on so far?
2 Likes
- Can the Vector Stores associated with Assistants API be used by third party LLMs like OSS models/Gemini for context retrieval? Or it has to be one of the OpenAI models only?
- Is there a limit on the creation of number of Assistants? And can each Assistant along with the associated Vector Stores be associated with a unique organization, so that each organization has a separate RAG store of its own segmented by Assistant ID?
Wish, you people brought some improvements to the Assistant API as well today.
1 Like