Vector Store Retriever with Qdrant stops working after n8n 2.3.2 update (indexing still works)

Questions
1

Describe the problem/error/question

After updating from version 1.122.4 to 2.3.2, the “Vector Store Retriever” node stops working. This node is used as the Retriever sub-node in the “Question and Answer Chain”.

The error occurs specifically in the “Vector Store Retriever” node. I am using Qdrant as the vector store. When the error occurs, I can see the following query in the Qdrant console:

127.0.0.1 "GET /collections HTTP/1.1" 200 86 "-" "qdrant-js/1.16.2" 0.000175

However, in the “Qdrant Vector Store1” node, there is no activity visible.

Before the update, everything worked fine.

Notably, the upper part of the workflow, which handles indexing new documents, still works correctly after the update.

What is the error message (if any)?

Cannot read properties of undefined (reading ‘asRetriever’)

Error details

Other info

Item Index

0

n8n version

2.3.2 (Self Hosted)

Time

1/13/2026, 10:32:32 AM

Please share your workflow

Share the output returned by the last node

Cannot read properties of undefined (reading ‘asRetriever’)

Information on your n8n setup

  • n8n version: 2.3.2
  • Database (default: SQLite): SQLite
  • n8n EXECUTIONS_PROCESS setting (default: own, main): main
  • Running n8n via (Docker, npm, n8n cloud, desktop app): npm
  • Operating system: Windows Server 2019
1 Like
2

Hi @kolbermoorer

Have you tried adding your Quadrant collection by ID insted of using the list function in Qudrant vector store node? I have in other cases had succes with hardcoding the ID.

Let me know if this helps

3

Hi,

I’m experiencing the exact same problem after the update in my workflows with the Vectore Store Retriever/Qdrant.

I’ve already tried both methods (list and by ID), also with different collections. The error persists.

I would appreciate any suggestions for a solution.

4

A quick note for anyone who might land on this page with the same problem:

My solution to work around this issue was to use an AI Agent node instead of the “Question and Answer Chain” node and connect Qdrant directly as a tool there – this eliminates the need for the “Vector Store Retriever”. Probably the better way anyway :person_shrugging: