Hello,
I’m interested in finding out the recommended optimal approach for testing our solutions using prompts on Google AI Studio.
Should we use one long prompt that focuses on serving a functionality with several functions, and includes a prompt-narrative to establish a constraint to the processing and the flow, similar to a class in programming?
Or should we use one dedicated prompt for each function and let our front end handle the flow?
Welcome to the forum. A place to get ideas that suit your application requirements is Google | Build with the Gemini API | Google for Developers . The barista bot is a rather extreme example of detailed prompting for multiple functions. That corresponds to the long-form prompting you mentioned.
There is no good alternative to actually testing your prompts against expected (and adversarial) situations. Some considerations are presented in Введение в быстрый дизайн | Gemini API | Google for Developers (and following sections)
Fair & clear enough. Thank you @OrangiaNebula