RAG Agent: load files in supabase vector database

Questions
1

Hello :slight_smile:

I build a RAG Agent and everything works fine with demo data.

The agent gets a list with

  • Problem 1
  • Problem 2
  • Problem x

The Agent have to check a supabase vector database.

The database has two columns: “problem” and “solution” (+ “id” + “embedding” + “metadata”)

If the Agent finds the problem from the list in the database, he has to grab the solution and create a new Google Docs with the exact solution from the database (without changing it).

My big Problem: how to fill the database with my data…?

I have X PDFs in my Google Drive.

  • Title of the PDF = “problem”
  • the whole text of the PDF = “solution”

I have a “extract from file” node that gives me the needed data

[
  {
    "numpages": 1,
    "numrender": 1,
    "info": {
      "PDFFormatVersion": "1.4",
      "Language": "de",
      "EncryptFilterName": null,
      "IsLinearized": false,
      "IsAcroFormPresent": false,
      "IsXFAPresent": false,
      "IsCollectionPresent": false,
      "IsSignaturesPresent": false,
      "Title": "Problem 1",
      "Producer": "Skia/PDF m143 Google Docs Renderer"
    },
    "text": "solution to problem 1",
    "version": "5.3.31"
  }
]

The next step would be “upload to supabase” and fill the column “problem” with “Title” and the column “solution” with “Text”.

Without splitting “Text” in chunks.

But how….?

I build a custom code node

And the output is

[
  {
    "problem": "Problem 1",
    "solution": "solution to problem 1"
  }
]

But it doesn’t work with the “upload to supabase” + “default loader” node :frowning:

It can’t be that hard…?

Thanks for your help

2

If you want to see a working version you can checkout the version from Cole Medin:

He also made a video about a previous version of the flow: https://www.youtube.com/watch?v=PEI_ePNNfJQ

3

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.