The evolution of mobile networks to 5G has introduced a key concept such as network slicing which allows network providers to create multiple logical networks over a physical infrastructure to satisfy the growing user demand on the network. In this work, we present an open-source, SDN-driven testbed for 5G standalone network slicing that enables dynamic throughput management via both reactive and predictive controllers. Hosted on Proxmox VE, our architecture combines Open5GS for the 5G core, UERANSIM and srsRAN for RAN emulation and real-world SDR connectivity, OpenDaylight as the SDN controller, and Open vSwitch for data-plane slicing. Two adaptation strategies—a threshold-based allocator and an LSTM-based predictor trained on VoD traffic—are evaluated through throughput measurements. Our results demonstrate that both schemes enforce slice isolation and guarantee prioritized bandwidth under variable loads, with the LSTM approach offering smoother performance during traffic surges. This modular platform opens a reproducible path for rapid prototyping of slice orchestration policies and paves the way for advanced AI/ML-driven slicing.