quick summary so they can pinpoint the problem for you instantly:
The Workflow Goal: A RAG chatbot ingestion pipeline using Drive 1 (List/Search) ➔ Loop Over Items (Batch Size: 1) ➔ Drive 2 (Download via {{ $json.id }}) ➔ Vector Store.
The Problem: The first node (Google Drive File/Folder) is set to Advanced Search with the correct query '19iFb6TqwX1x_LIl1B-jXFvkoLcTrjeXP' in parents and Return All turned ON [docs.n8n.io].
The Glitch: Despite the folder having 4 PDF files (all set to public link sharing via “Anyone with the link”), the node stubbornly returns only 1 single item (the Midlife Kitchen PDF), causing the Loop node down the line to never activate or count up.
What is the error message (if any)?
Please share your workflow
(Select the nodes on your canvas and use the keyboard shortcuts CMD+C/CTRL+C and CMD+V/CTRL+V to copy and paste the workflow.)
If you are using a Google Service Account for authentication, “Anyone with the link” is often not enough for the search operation to work.
While “Anyone with the link” allows a person to view a file if they have the direct URL, a search (which uses the Google Drive files.list API) only returns items that are explicitly “visible” to the authenticated user/service account. A Service Account cannot “see” a file in a search result just because it is public; it must be part of the file’s/folder’s permission list.
Go to your Google Drive in the browser.
Right-click the Folder (not just the files).
Click Share.
Paste the email address of your Service Account (e.g., your-account@project-id.iam.gserviceaccount.com) and give it “Viewer” access.
The search should immediately pick up all 4 files
Using “Advanced Search” with a manual 'ID' in parents query is powerful, but it can sometimes conflict with other hidden parameters or the “What to Search” setting in the n8n node.
Stop using the “Advanced Search” mode for this specific task and use the built-in Filter instead:
Search Method: Change from “Advanced Search” to “Search File/Folder Name” (you can leave the Search Query empty or use *).
Filter Section: Look for the “Folder” option under the “Filter” parameter.
Folder Selection: Set it to “By ID” and paste your ID: 19iFb6TqwX1x_LIl1B-jXFvkoLcTrjeXP.
What to Search: Ensure this is set to “Files”. This tells n8n to use the official parents parameter in the API request rather than a custom string, which is much more robust.
If the other 3 PDFs are actually Shortcuts to files located elsewhere (and not the actual files themselves), a standard search query for files inside a folder may skip them.
Check the icons of the files in Google Drive. If they have a small “arrow” on the icon, they are shortcuts.
If they are shortcuts, you need to change your query to include them or, ideally, move the actual files into that folder.
Filter Section: Look for the “Folder” option under the “Filter” parameter.
My Google Drive is set exactly as recommended here. Now I pasted my email address in the Folder Share option. I have executed the drive. Still, it is listing only 1 PDF
I’d check whether the other 3 PDFs are actually shortcuts rather than the original files. If the Service Account already has access to the folder, that could explain why the search only returns the one actual PDF. The shortcut files should have a small arrow on their icons in Google Drive. If so, try moving the actual PDFs into the folder and run the node again.
Hello ! I’ve seen i’ve been mentioned, i would be happy to help, could you post your current workflow please, I think it will be easier to understand your issue !
Maybe try moving the folder into a shared drive, it could be a possible way.
I could not find specific information about “Bellyluxe Asanas” in the document. The retrieved content focuses mainly on meal plans, recipes, and dietary guidelines for hormone balance and midlife health, but does not explicitly mention or describe Bellyluxe Asanas.
We do not have access to your instance, can you Select the nodes on your canvas and use the keyboard shortcuts CMD+C/CTRL+C and CMD+V/CTRL+V to copy and paste the workflow.
Hey @linda33, I think the loop body in that workflow is still the placeholder. Loop Over Items sends its loop output to Replace Me and straight back round, and Download file has nothing connected to its input at all, so neither the download nor the vector store insert ever runs. One document is all you’ll ever get out of it, whatever the Drive search returns.
Delete Replace Me, then wire the loop output into Download file, Download file into Simple Vector Store, and Simple Vector Store back into Loop Over Items.
I think you are trying to achieve this setup in the workflow, In your workflow above the wiring of the nodes was not correct and here below i have corrected the wiring and built 2 workflows as i think the setup would work the best in this architecture.
I have tested this workflow on my n8n instance and it works, You can check this out and try it yourself also let me know if you need help with anything else or encounter any further problems.
And i think the search query isn’t the issue. In name mode, the node sends Google name contains '*' plus the folder ID in parents, where the star is just treated as a literal character. Since your first Advanced Search had no name term and still returned only one file, Google is simply returning one file in that folder to whichever account n8n is signed in as.
I think a really good approach over here would be having the files you want land in a folder in Google Drive. You can watch that particular folder for any new files being added as a Google Drive trigger. Any new file that is added into that folder will automatically trigger and load that into the vector store.
The first thing I’d check is the output of the Google Drive node before the Loop Over Items node. If the folder really contains all 4 PDFs, the Drive node should return 4 items, so the loop itself probably isn’t the issue.
I’d also test the Drive node with a simpler setup: select the folder directly instead of Advanced Search if possible, keep Return All enabled, and run it by itself. If it still returns only the Midlife Kitchen PDF, check whether the other files are actually Google Drive files in that same folder rather than shortcuts or files owned from another Drive/shared location.
The sharing setting (“Anyone with the link”) shouldn’t determine whether the List/Search operation finds the files. The important part is whether the files are actually children of that folder and visible to the Google account connected to n8n.
If you paste the raw output from the Google Drive node, it should be possible to narrow down where the other 3 files are being filtered out.
Though the Google Drive Node recorded the PDFs, it was not able to read and my test chat failed to get a response. So I tried this slightly differently. It is working, and my test chat is sort of giving me the response, but I think I will have to train it strictly to get the kind of response I’d like to get.