Configure this n8n FAQ workflow for MandAPI in a local reviewable copy. Use Chat Trigger, Basic LLM Chain and an OpenAI Chat Model; keep the final Edit Fields mapping from text to output. Create an OpenAI-type credential named MandAPI with Base URL https://api.mandapi.com/v1, and let me enter the secret myself. Never put the key in exported JSON or use an OpenRouter credential for it. Set a real Chat Completions MODEL_ID available to my key and explicitly keep responsesApiEnabled false. Preserve the private, inactive trigger until I choose to publish it. Replace the sample FAQ only with facts I provide. Unknown questions and refund requests should show a human-review message without sending email or taking financial actions. Verify node parameters locally; do not claim import or runtime success without actually checking it. A live model request requires my authorization.