Gpt send content of file in action

I want that Gpt send to a third party the content of a file that the user upload but always get the file ID name and not the content.

Someone knows how to do? It’s possible to do?

Thank you.

What do you mean by getting the file ID name and not the content? Aren’t you asking for the content?

Either way though, the answer is no.

Uploaded documents to GPT is the user’s data and treated as part of a conversation dataset, and the user has sole control over how their conversation data is used. That is solely between the responsibility of OpenAI and the user, not us.

You can build an app that could manage the connections between some things (look at zapier), but plugins like zapier basically just provides the cable for the user to transmit things; zapier is not really touching or handling the contents of user data (nor should they). Again, that would be a huge privacy violation, and would result in a lot of problems.

Hello Macha,
Thank you for the answer, the problem i want to solve is to resume a audio file. So first i need to pass the audio to whisper api.
We GPT send files or contents show the data that is going to send and need your approvement.
I do not see any problem with privacy meanwhile GPT show you that is going to send some data.
By the way, do you know how to solve that problem in a GPT?

Thank you!

Wait, are you using the API? ChatGPT and OpenAI’s API calls are different things and mutually exclusive. This forum category is for ChatGPT, the web UI interface.

The two cannot be integrated in the way you’re describing.

I’ll admit I’m still having some trouble fully understanding what you’re trying to say though. I would run your posts through a grammar check on a word or google doc before you post (or have ChatGPT help you translate).

sorry for my bad writing. I’ll tell you what I want to do.
I want to create a GTP that, by uploading an audio file, gives you a summary.
GTP does not have the ability to transcribe audio so I use the actions to tell it that when an audio is uploaded it calls the whisper API and obtains the transcription of the audio.
This way it is as if you inserted text and made a summary of the text.
How would you do it? I hope this time you understood me better

1 Like

No need to apologize, it’s perfectly okay. I just want to be able to help you in the way you need it. This response is much better.

Unfortunately though, I still do not believe that is possible. You would also skyrocket your API bill if you let that GPT loose in the wild. You currently cannot access nor interact with a user’s data file like that.

You would need to make your own API endpoint to handle that, then pass the transcribed audio back. Problem is, you can’t pass a data file like that via an API call. Again, this is what I was saying earlier; even if you could, the user would have to give you their data file to process, which could be a huge privacy breach because of the way in which you would have to process that kind of data.

You would need to build this app from scratch. This cannot be done currently via GPT builder.

I just stumbled upon that post and I have essentially the same question. I also thought about whisper, but for now, it is more about creating knowledge bases. So the user can work with different GPTs (different virtual employees/experts) or summarize a document and then ask knowledge GPT (my GPTs are all private) and asks it to commit the data to the knowledge base. From the text, the GPT creates metadata in the form of #something. A script over API fires up, creates embeddings and stores embeddings, text and metadata in a pinecone data base. Another endpoint does the querying and retrieval. Instead of text, I wanted to have the possibility to also take an uploaded file as a whole and commit it to the knowledge base. So I created another endpoint that accepts files - and would fire, when the user directs to “upload a file to knowledge”. The API works perfectly over swagger. Only the custom GPT refuses… Isn’t there any way to get around this? (I mean other than doing it in scripts outside chat GPT using the OpenAI API). Many thanks in advance.