Hi guys these days i’ve been looking for a way to create OpenAPI Schema for my GPT’s here on the forum but, as i can see, no one have clear vision on how to do this because openai disable the openapi builder that was before.
so for you ai fans i create a GPT called “OpenAPI Builder”, a tool to transform any API call to a schema that you can put in your own GPT’s.
I hope it can help everyone while we dont have anything else
have a good day!
Edit 19-12:
The GPT can search inside openapi whole documentation.
More schema examples to evade wrong answer with complex apis.
Export the results as .json or .yaml files
Transform large scale of api’s at once (In Progress)
Test the results inside the chat (In Progress)
Thanks. MyGPT is billed as a no code no brainer but as usual for non-dev’s like me, the advanced features require coding experience. So I believe your GPT will. help. Thanks for making it. I’ll report back when i have a practical use for it
I just want to say THANK YOU! I know nothing about schemas and I got one created in 6 steps using your custom GPT!! Here they are:
(1) Asked your custom GPT what information it needed.
(2) Ignored everything it said and just copied the Semantic Scholar webpage talking about their graph API into your custom GPT.
(3) It gave me a schema to use and I copied that into my own custom GPT.
(4) Copied the error codes back into your GPT, got a revision.
(5) Saved the revision in my GPT.
(6) Tested… and success!
btw another GPT that may help you is “API Alchemist”. It can create the whole API flow from idea to deployment and use. (It can create the whole code, openapi schema and deployment steps for anything you want to add to your GPT’s)
First of all thanks for making this. Having said that here are some issues you could tackle (with function calls probably).
there are url parameters that are being added in the url value when it should be added in the path, since the domain call wont register it in the "servers: url: " part.
When you point out the error and it fixes it, in the next iteration it completely forgot and prints it wrong again.
Definitely it’s useful in getting the scaffolding of the schema up, but many times the devil’s in the details and so it fails.
I’m working on an Asana schema, which has a well documented and easy to work with API, but getting it to work with OpenAI is a huge pain. They seem to require some fields in the URL and it just continually just ignores the values in the schema and just puts some random numbers in there.
It’s really frustrating. It really drives me away from using their builder when coding it works flawlessly but how they want you do it doesn’t work as you would expect it to.