Vision Agent: Agentic Program Generation for Vision Tasks

[@IndianArmchairGeneral]

wow so cool!!! ^_^

[@caiosouza541]

Cool ! Will gonna try :)

[@cagdasucar3932]

why generate a program? should this not work like calling gpt in a loop like other agent frameworks?

[@dw61w]

like autoGPT? I think they're calling GPT in a loop based on the number of GET requests to openai. But they're saying outputting python is cheaper because you don't need to call openai api when you run the program on another video later