Use GPT-4o not 4o-mini as the provider in n8n, since it fully supports function calling needed to trigger custom tools automatically..