Troubles with RAG

I am working with Postgressql Vector store as tool for my agent, when I ask something the agent call this tool sometimes I get { “output”: “” } and other times I get the answer correctly.

How come?

Please share your workflow in a code block.

Hello Wouter,

When I ask something about payments i get { “output”: “” } but when I ask something else I get the response. What I am seeing is like somehow the azure model don’t use the tool information o something like that

I realize when I use the azure gpt-4.1 mini it is not working well with the tool. But When I use the openia model it works