Assistants API - Clean yaml/JSON Schema somewhere?

I’m looking for a schema YAML or JSON for the Beta Assistants API instead of trying to copy/paste the hodgepodge from the OpenAI reference docs. I’m trying write python code through GPT4 and get it to use the Beta Assistants API. It seems insane that given explicit instructions to use it, it cant seem to properly generate code to create, track, run etc Assistants. If i copy and paste examples, it briefly understands, but then seems to forget and go back to ChatCompetion approaches versus ‘client.beta.assistants.create’ style. Ive seen the API reference docs, but is there a repository of the python library schema for Assistants that i can copy/paste or upload to my coding GPT or just a GPT 4 chat so that it retains this knowledge. Thanks!

Take a look at the AdminGPT repo, which is a complete AI administrative assistant implemented with the Assistant API. Specifically, you will get a lot of cut-and-paste example from the getting started tutorial notebook, and if you want dive into functions, the Office365 Toolkit has many examples of pretty robust tools.

Also, happy to answer any specific questions about AdminGPT or the Assistant API here.

1 Like

openai-openapi/openapi.yaml at master · openai/openai-openapi ( What you asked fo and soooo much more.

1 Like

Yep… have been trying to keep pasting in the assistant part etc and eventually the context window flushes it and its back to square one . Can we all sign a petition or something to get the next model to actually know its own API :slight_smile:

thanks for source…will work with this. It brings up another project…trying to get AI to better mange outlook contacts. Saw that’s the one area they didnt have in the o365 tools :frowning:

I’m glad the code helped you move forward.

In terms of working on the Outlook Contacts, I think it would be a good next step at AdminGPT. Would you like to partner on it? The first step would be to get familiarized with the AddressBook API in the O365 Python repo. If you would like to dig into it, and help me understand how its functionality maps to what you would like the OpenAI Assistant to do, I’d be happy to help.