Assistant API + LangChain: I want it to use my own files from my own vector DB

I want to supply my own files to the assistant API, but I don’t want to upload them as files directly to my existing assistant - I just want the response to look at the following:

  1. The files directly uploaded to the assistant
  2. The files I have in my own database.

And cross reference those, and provide a response based on the data found in those two sources. Is that somewhat possible?

1 Like

Hi and welcome to the Developer Forum!

You can upload files directly to ChatGPT, but either way API or ChatGPT your data needs to be parsed and chunked into a format the model can either directly read as text or search, so at some point you will need to upload your data or possibly run a vector database on your own machine locally and pass the results from searches on that as context to the model in the prompt.