Good evening,
I am developing an automation flow in n8n, structured as an intelligent agent intended to serve a kindergarten. This agent is responsible for essential functions such as:
-
FAQ: answering general questions about the school;
-
SDR: meeting new interested parties and convincing them to schedule a visit to the school;
-
Scheduling visits;
-
Consultations on financial matters;
-
Consultations on students’ academic performance, among others.
Initially, our flow was performing its functions well. However, we recently noticed that the agent began to present “hallucinations”, failing to correctly recognize which tools (such as other agents, internal memory or database queries) it should use to respond to the various requests.
We are unsure whether the problem is related to the prompt used or whether concentrating all calls to other agents in a single orchestrator is the best approach. We would like to ask for your help to review the current configuration of the flow, identify the source of these hallucinations and ensure that the agent correctly uses the necessary tools to achieve the desired results.
Could you help us with this review?
Thank you in advance for your attention!