N8n Dockers > Send local pdf-invoices to OpenAI > Extract data to Google Sheets > Rename local files - Is this even possible?

Questions
1

I’ve spent the last few hours trying to get my automation to work, but couldn’t. I’m running the local version of n8n via Docker on Windows.

Here’s what I’m trying to do:

  1. I have a local folder on my Windows computer where I drop PDF invoices. I want n8n to send them to OpenAI.

  2. OpenAI should then read each PDF and extract data like the date, company name, amount, etc., and put it in a Google Sheet.

  3. Finally, n8n should rename the PDF files in the local folder using a specific naming format based on the extracted data.

I’m already doing all of this with the help of ChatGPT - copy-pasting data back and forth - but now I want to fully automate it using n8n.

I could write down everything I tried and all the things I ran into, but I think that would just make things unnecessarily complicated. Is there someone who could walk me through how to do this step-by-step in a fool-proof way?

Or is this kind of automation not possible with n8n local (Dockers) at all?

Thanks!

1 Like
2

Definitely possible, you could even use a vector store to load the PDF, extract what you need, and clean up right after.

It shouldn’t be too hard to set up in n8n, especially if you use the AI Agent node or even just a basic LLM chain.

Just tell the agent what to extract, feed it the PDF (or chunks via the vector store), and you should get pretty solid results.

You could even take it further by using OpenAI’s Assistants API, trained with context on what to do — helps make the output way more accurate.

3

My problem is, I can’t even make the “Read/Write Files from Disk” work. When I press “Test Step” it says:" Node executed successfully" but in the left window it says “No output data returned”.

Don’t forget, that I’m using n8n local through Dockers, which may be different from the web version.

Current situation: I have a Container with a mounted local folder, with my pdfs. I have no idea how to proceed from here. For example, what would be the very first workflow step?

4

Most people just use google drive or similar.

check this post out as discussion here about similar

5

I saw those threads, but I would like to upload them from Windows folders. The solutions with Google Drive or OneDrive require a Google Subscription. Also, in my case I’m using n8n through Dockers, which seem so work a bit different from the web version.

6

You need to connect a domain to sure n8n instance for callback URLs tbh, when settingup oauths,

could u share ure current workflow plz

n8n attach workflow
just hit paste after clicking the code node and I can see what u have etc

I would like to try out too on my local version.

7

I appreciate your help with this, but I already tried to connect Google Drive or OneDrive. It seems to require a paid subscription to work. But also if it wouldn’t, I’d like to use a local Windows folder :slight_smile:

Regarding my workflow, as mentioned, I’d need someone to guide me step by step. All the Youtube tutorials seem to use the web version. I’m using the local Dockers version. Also, I’m not even sure if I can just send pdfs through the OpenAI API. People seem to convert it to other formats first. My automation idea looks so easy at first but in reality, it’s not the case, so I’m looking for someone who can guide me (Dockers version, Local files, PDFs send to OpenAI, Renaming…)

1 Like
8

Maybe something like this

9

Thank you, I’ll take a look at it! :slight_smile:

10

Also you could use like an OCR or https://cloud.google.com/document-ai?hl=en I found yesterday which may help,

Also, you may only need to store the data first, and not connect the other ai agent, and have the ai agent call from another flow and access the vector store, btw a lot of people use superbase I’ve seen

1 Like
11

I am trying to do the same thing on community addition. It is easy to send the pdf to a file storage of choice with the name of the ordiginal attachment.

The problem I find is after running an AI node. There is no way to convert JSON back to Binary data needed and there is no rename node either. I hope someone can either create the needed node.

The only other thing i can think of is exporting the data, and doing it outside of n8n or exploring other workflow tools to see if any of them are better suited for this type of use case.

1 Like