Developing ChatGPT Plugins


What languages are used to develop ChatGBT plugin ?and what IDE ? can I use swift and Xcode ? and how and what format its packaged and distributed to online store or something ?

Kindest Regards

Plugin is basically a web service with a bit of metadata packaging,
Apple is all client side stuff…
People use Python, NodeJS… I use Wolfram Language.

1 Like

Aha, thanks allot. So they are web services that uses OpenAI API’s ? I mean if the features are already there in ChatGPT why we can’t just use it directly from it ?

Kindest Regards

1 Like

Imagine ChatGPT as a ready build app by Open AI. It’s pretty much a demonstration of the capabilities.

Hope that helps.

You can also give likes to helpful posts and mark topics as solved. This helps the other community members.

1 Like

aha so you mean the API’s have more capabilities ? is there model training feature for example ?

Kindest Regards

The API can do whatever you want. I strongly recommend to take a look at all the different types of examples in the documentation.

It’s packed with so many examples and ideas or will definitely answer your question.

1 Like

not quite.
When you develop an genAI application then your code calls the LLM - in this case using OpenAI APIs.
When you develop a plugin then the LLM - in this case OpenAI GPT4 - calls your API.

LLMs are language models. They are good for a lot of things… but there are areas where you just need some precise computation. Then the right thing to do it to write it and expose the functionality as a plugin.

Read this from Stephen Wolfram

1 Like

Sure I will do but just wanted to have a general idea, thanks allot.

Thanks allot I’ll go through it carefully. But I have a question no one answered what about trained models ? is it something that OpenAI don’t offer or other competitors only use it for marketing and what OpenAI offer is enough ?

Oh, I forgot about that. You can fine-tune Open AI models but fine-tuning is often not what people need. In short, most problems can be solved with proper prompting. And you fine-tune when this option is exhausted.

I suggest this video to get you started:

By the way, I don’t disagree with this statement

…there are areas where you just need some precise computation. Then the right thing to do it to write it and expose the functionality as a plugin.

My perspective is that you can leverage the API to get the results of such computations via functions and use the results for inference.
Since I remember that you are still exploring the differences between ChatGPT on the web and the API this is should be clarified.

1 Like

Yes now things are more clarified , actually the confusing came to me from Google. by the way ChatGPT says that Bard is developed by OpenAI, don’t know if its a mistake or true Bard is a licensed copy from API, I think its same case as Pilot from Microsoft. Thanks allot

Bard came out after the knowledge cut-off for ChatGPT’s dataset. Any comments it makes about “Bard” being an LLM model will be hallucination.