Pipe each prompt through whichever AI you're testing, then paste the reply back.
Polling GET /readings every 2s · App posts to POST /readings
GET /readings
POST /readings