make it so we can push gpt3 prompts between dev -> stg -> production