I auth’d ChatGPT with google and now it can read a private sheet and analyze it with some python. All I want to do is port this scenario to the API.
I have searched for an answer in the forums and elsewhere; all of them have been negative, but they are also at least a year old or older. So I was looking for something more up to date as the world changes so quickly.
ChatGPT told me I could, and even gave me code. But then I asked it if it was hallucinating and it said yes. 
thank you!
Write the retrieval as a function doing all the auth flow locally.
Then pass a version of the parsed sheet to the LLM?
tbh this is no different to using any API.
If you want to get fancy you could expand the granularity of the retrieval to give the LLM more discretion.
So that’s using the Actions or function calling approach, yes? I guess I was trying to avoid being the middle man for the data flow; it’s nice in ChatGPT when that data acquisition step is done in the background. I was hoping for a similar level of abstraction, but it may not be possible.
I hear you but often if you are the middle man there can be advantages like being better able to control running costs. The effort up front is admittedly greater.