Architectural And Engineering Aspects Of Integrating The Novita Ai Api Into A Web Application For Image Generation

Abstract

This paper investigates architectural and software engineering approaches to integrating the Novita AI image generation service into a web-based application that supports scalable and asynchronous request processing. The proposed solution is based on a clear separation of responsibilities between frontend and backend components, where the backend handles validation, orchestration of long-running tasks, interaction with external APIs, and data persistence. Communication is implemented using REST for synchronous operations and WebSocket-based notifications for asynchronous updates, while resource-intensive tasks are executed through a queue-based mechanism to prevent interface blocking. Integration with the Novita AI service is encapsulated within a dedicated service layer, ensuring modularity, maintainability, and extensibility. An experimental evaluation was conducted to measure execution times of key operations, including image generation, transformation, upscaling, and custom model training. The results confirm the effectiveness of asynchronous processing and demonstrate that custom models improve output consistency at the cost of a minor increase in generation time. The novelty of this study lies in the architectural justification of asynchronous API integration for AI-based image generation with support for custom model training, validated through experimental performance evaluation.