How-To: download files from nextcloud and embedd into pgvector

Help me Build my Workflow
1

Hi, I want to download a couple of files from nextcloud and embedd them into pgvector.

Below is the workflow.

Following problems:

(a) In the “Delete old embeddings” node, the DateTime does not seem to have the “fromHTTP” and thus produces a NaN. According to the Luxon documentation, there is such function.
(b) Even overcoming this issue (i.e. adjusting the SQL accordingly), only one item is forwarded from the “Delete old embeddings” node.
(c) I bypassed the “Delete old embeddings” node so the “Embedd” node receives three items. The “Embedd” node executes without complaint (green check mark), yet if executes only exactly once and does not write anything to the data base. Actually, even no file content is forwarded to the OpenAI embedding.

What am I doing wrong?

My N8N-Version: 1.83.2
Environment: Docker

2
3

Sorry, I could not add the code in the initial post, but I had to split the post. Doing the code in the original post constantly resulted in http-403.

4

Another small progress: I made the “Delete” node paralle to the “Embedd” node. In this way, both nodes receive three items.

Unfortunately, the files I download are of mime-type “text/markdown” and the “Embedd” node complains that it cannot process them. Is there any way how to do it? I tried the “Extract from file” but this node does not return anything - even I enable the switch “Always output data”.

5

Ok, I managed to solve the problem.

I save the result of the “List” operation in a data base table, do the clean-up of old embeddings using SQL, and then run each file individually via the embedding.

Not so nice, a few nodes more, but works very smooth.