This paper investigates a modular beamforming framework for reconfigurable intelligent surface (RIS)-aided multi-user (MU) communications in the near-field regime, built upon a novel antenna architecture integrating an active multi-antenna feeder (AMAF) array with a transmissive RIS (T-RIS), referred to as AT-RIS. This decoupling enables coordinated yet independently configurable designs in the AMAF and T-RIS domains, supporting flexible strategies with diverse complexity-performance trade-offs. Several implementations are analyzed, including diagonal and non-diagonal T-RIS architectures, paired with precoding schemes based on focusing, minimum mean square error, and eigenmode decomposition. Simulation results demonstrate that while non-diagonal schemes maximize sum rate in scenarios with a limited number of User Equipments (UEs) and high angular separability, they exhibit fairness and scalability limitations as UE density increases. Conversely, diagonal T-RIS configurations, particularly the proposed focusing-based scheme with uniform feeder-side power allocation, offer robust, fair, and scalable performance with minimal channel state information. The findings emphasize the critical impact of UEs' angular separability and reveal inherent trade-offs among spectral efficiency, complexity, and fairness, positioning diagonal AT-RIS architectures as practical solutions for scalable near-field MU multiple-input single-output systems.