How does Postgres PGVector Store Collection work?

Questions
1

This is a very simple question, but i cannot find any documentation whatsoever regarding the usage of collections. N8N doesn’t state what are the requisites, if it’s a different table, a determined column, what even changes if i use it or not.
Any help is appreciated

Screenshot 2025-08-07 at 18.39.52
Screenshot 2025-08-07 at 18.39.52576×500 23.2 KB

Describe the problem/error/question

No instruction on collection usage, i wish to partition my vector table by information domain but i’m unsure if this is a valid option.

What is the error message (if any)?

Tooltip is extremely generic and adds no new information.

Please share your workflow

Share the output returned by the last node

Information on your n8n setup

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

Hey @stranger_cacaus hope all is well. Welcome to the community.

I am hypothesizing here, but I think the concept of collections here comes from LangChain implementation, which uses said concept for logical separation of vector embeddings. Since PGVector is just a Postgres with vector extension, the way it is implemented is via two tables, where one keeps the collection metadata (name, id etc), and another one keeps the embeddings. Each record has a pointer (foreign key referencing) to the collection it belongs to. So this can be treated just like a collection for other (real) vector stores, only with a twist of relational databasiness.