I wish n8n's AI sub-nodes would work with normal workflows

It’s really cool that n8n has some built-in AI tools like a data loader, embeddings, token splitter, etc.

But I find myself wanting to build a workflow the normal way, and use just one or two of these tools in that workflow.

It seems unintuitive that these AI tools are a totally separate type of node that doesn’t interact with the normal nodes. Would love to be able to use the token splitter to chunk up some text in the workflow above for example.

Am I missing something?

Hello Giovanni,

I think it would be better to directly insert data into Supabase vector store using the specifc node for it.

The vector store nodes should handle the hard work for you, integrating with OpenAI embeddings without the need for an HTTP node.

If you still need to insert rows using the standard Supabase node, you can use the in-memory vector store to handle embeddings and shape the output before inserting into database.