Hi everyone,
I’m building an n8n workflow that uses two Supabase Vector Store tools connected to the same Supabase project — one for experts and another for websites.
Each table (personal_pr_experts and personal_pr_sites) has its own embeddings column and index.
Both nodes are connected to the same Supabase credentials, but whenever I query the second Vector Store, it always retrieves results from the first one (experts) — even though I explicitly specify the correct table name (personal_pr_sites).
Here’s what I’ve already tried:
-
Dropped and recreated all
match_documentsfunctions; -
Created separate RPC functions (
match_sites,match_experts); -
Rebuilt vector indexes for both tables;
-
Even reinstalled the Vector Store tools — the issue persists.
It looks like the n8n Vector Store node somehow caches or reuses the first RPC function, ignoring the table name parameter.
Question:
Is it currently possible to use multiple Supabase Vector Store tools within the same n8n project, each pointing to a different table?
Or does n8n internally link to a single RPC function per Supabase connection, regardless of table name?
Any advice or workaround (e.g., using filters, separate credentials, or namespaces) would be really appreciated.
Thanks in advance!