Candy ai clone BY Triple Minds

Im currently working at NSFW Coders, where were developing an API for the Candy AI Clone project an AI chatbot system designed to handle both text-based and image-generation interactions. The goal is to make the chatbot respond contextually while generating AI visuals in real time. From a development perspective, Im focusing on: Designing a microservice-based backend that can handle high-volume API calls Managing GPU workloads efficiently for image generation Maintaining low latency while processing conversational context Ensuring proper data flow and caching between the text model and image generator Right now, were testing different model combinations and scaling options to keep the responses fluid even under heavy user activity. Im curious how others here are approaching AI-driven API scalability especially for real-time media generation. How do you balance between compute performance and API response time in similar systems? https://nsfwcoders.com/chatbot/candy-ai-clone/