Hello,
I am trying to use Python to connect to a custom GPT model in the GPT store. I am looking at the Assistants API, but it isn’t clear to me that it can connect to custom GPTs rather than just the ones that OpenAI provides.
Can I use the Assistants API to connect to a custom model and query it?
Thanks
There is no “connecting” as the GPT is exclusive to ChatGPT Plus and ChatGPT users.
As programmer, you would have to replicate the behavior the instructions evoke, with the subset of features that Assistants has, such as no browsing or URL clicking, no dalle, no internet connection to APIs by actions…but the ability to choose whatever model you want and call your own code’s tools.
I’m not clear or really understanding what you’re saying here. When you say "replicate the behavior the instructions evoke… ", What instructions? What I want to do is call a specifically trianed GPT model from the store and query it. Can I use Assistants to do this?
The answer is basically a big no. GPT is a feature just for ChatGPT.
On the API, you also have to pay for API services separately by usage, so you can’t just “put my GPT on my website” without considering the budget.
There is no “training”. There is only a text box in a GPT with the instructions the AI receives. A fancy “custom instructions”.
I’ll send you over to this long overview.
Indeed. This would need to be beyond, at least , an authentication wall (if not a paywall).
I guess the nearest you get to training is that you can add docs to a custom knowledge tab in your GPT. If you disable web browsing I think this will then restrict your GPT to answering from the documents in it’s knowledge base.
But as to accessing that GPT I’m at a loss, nearest I can gather is as OP suggests using an Assistant and then using the assistant API.