The rapid expansion of retail networks has increased the demand for real-time video monitoring across thousands of cameras, not only to ensure security but also to enable intelligent AI-driven applications such as VIP customer identification, intrusion detection, density analysis, and behavioral analytics. Traditional systems face major challenges in scalability, high latency, and limited deployment flexibility. To address these issues, this paper proposes a hybrid distributed architecture based on Microservices, deployable across both Cloud and Edge Computing. The architecture allows each AI application to be independently scaled and integrated without disrupting existing services, supporting flexible deployment models—from Cloud-based solutions for large chains to Edge deployments for high-security or low-connectivity stores, or a combination of both. Experimental results demonstrate that the proposed system reduces hardware and operational costs, maintains low latency through automatic fault recovery, optimizes human resources, and enhances overall efficiency, while the Microservices model provides a flexible foundation for future integration of AI applications.