Why there is no “Always output data” setting in any of the vector stores nodes?
Hi @Pedro5 Welcome to the community!
the reason the vector store nodes don’t have “Always output data” today is not that they always find something, but simply that this option hasn’t been implemented for these nodes yet. An “Always output data” toggle would still be useful here so the workflow can continue even when the vector search returns nothing
Hi @Pedro5 ,
This is a known limitation that confuses many users working with the new AI features.
The Reason:
The Vector Store nodes (like Supabase, Pinecone, etc.) belong to the AI/LangChain category of nodes. Unlike standard n8n nodes (like HTTP Request or Set) which are designed to process lists of JSON items, these nodes are built on a different underlying architecture.
They are primarily designed to pass Configuration Objects (e.g., “Here is my database connection”) to an AI Agent or Chain, rather than transforming data row-by-row. Because of this specialized “Helper” role, they do not inherit the standard settings wrapper that includes “Always Output Data”, “Retry on Fail”, or “Continue On Fail”.
Workaround:
If your workflow stops because the node returns “No output data” (e.g., after an Insert operation that provides no feedback, or a Retrieval that finds no matches):
Check your Operation: Ensure you are using the node in a mode that generates data (like “Retrieve” or “Insert Documents”) and not just defining the store.
Use a Parallel Branch: If you need the workflow to continue regardless of the Vector Store’s result, connect the previous node to both the Vector Store node and a Merge node (in “Pass Through” mode) to ensure the execution path continues.
Best regards,
Hi @Anshul_Namdev and @A_A4 ,
Ty for the quick reply, what is strange is that the previous Supabase version (1.1) node had it, so it was removed on purpose I guess. In my case I have metadata filter so it only retrieves data from the specified user_id, but when the user doesnt have data it doesnt output any data. It would be great to have the setting back but ty for the idea of the parallel branch.
Ty both
@Pedro5 The new vector store nodes use a different internal architecture and are treated more like helper/configuration nodes, so they don’t inherit generic options like Always output data. I guess this is intentional because of their intent of search and storage mechanism.
Hope this helps.
