For example, I have my own custom model that responds in the way I configured it, such as answering only in Tajik and only in JSON format. And let’s say I want to use this model in my code by asking it questions and receiving answers in this specific format. So, the question is, how can I use my custom GPT in my code? Are there any solutions?
Welcome to the community!
You’re probably looking for assistants: https://platform.openai.com/docs/assistants/overview
While CustomGPTs can call APIs, they can’t be called through an API. Assistants are specifically made for that use-case.
what exactly is the difference between the assistants and the custom gpt? I keep trying to find something specific regarding it but alas, i have not.
Oof, yeah. It can get confusing.
CustomGPTs:
they’re on chat.openai.com
cost
you and every user pays a monthly subscription (unless they’re free now?), and for that you can chat with them as much as the rate limit allows.
access
everyone has to go through chat.openai.com and use that UI. Bots can’t use it.
notable limitations
with a custom GPT you can’t really control what model the user will be using.
Assistants
they’re on platform.openai.com
cost
the owner of the assistant pays per token (basically per word) going in and coming out of the assistant. Users don’t pay anything, unless they have an arrangement with the owner, ofc.
access
assistants don’t really have a UI. you have to use code to interact with them (the owner can play around with a simplistic demo UI on platform.openai.com/playground, but normal users don’t have access to that)
that means that as a developer you can use bots to interact with assistants (or assistants with each other), or have a program that embeds an assistant into your website, for example.
notable limitations
assistants don’t come with web browsing capabilities. you’d have to build t that yourself.
hope that helps 