n8n - how to emulate ChatGPT's 'Upload a file' and interrogate it

Hi guys

I'm either struggling with something dead-simple or near-impossible; I can't seem to figure out how to post a file or an image to model (Gemini or OpenAi or whatever) and interrogate it against a vector store.

It's almost exactly the functionality found in ChatGPT's Projects area; you can add a bunch of files which gets turned into the project's vector store and then you can interrogate this vector store via chat alongside file uploads.

In a perfect world there would be an agent that has access to a vector store and then pulls a file from Airtable and queries it. I'm not sure if this is possible though; if there other flows that make it happen I'm all ears..

11 comments