A compact wireless power transfer (WPT) framework is proposed, featuring a plane-wave feeder and a transmissive 2-bit reconfigurable metasurface-based beam generator, forming a reconfigurable power router. The combined profile of the feeder and beam generator is 0.8 wavelengths, enabling intelligent wireless power transfer to multiple power-consuming targets in dynamic environments. The framework integrates a deep-learning-driven environment sensor for object detection and localization, allowing real-time power delivery. Experiments demonstrate simultaneous wireless power and information transfer, with the router capable of detecting and localizing multiple targets and selectively supplying power based on their energy needs. The framework offers low cost and compact size, making it suitable for commercialization of metasurface-based WPT routers. The system uses a stereo camera and intelligent computation unit for dynamic environment sensing and real-time power distribution. The framework also enables simultaneous wireless information and power transfer (SWIPT), enhancing wireless communication efficiency. The proposed WPT system is capable of delivering power to moving and static objects, demonstrating real-time adaptability. The system's ability to focus energy in three-dimensional space provides significant flexibility for WPT applications. The framework also shows potential for integration with wireless communication systems, enabling future advancements in 6G and indoor WPT. Despite challenges such as low power transfer efficiency and electromagnetic interference, the framework offers a promising solution for sustainable energy supply in IoT and industrial applications.A compact wireless power transfer (WPT) framework is proposed, featuring a plane-wave feeder and a transmissive 2-bit reconfigurable metasurface-based beam generator, forming a reconfigurable power router. The combined profile of the feeder and beam generator is 0.8 wavelengths, enabling intelligent wireless power transfer to multiple power-consuming targets in dynamic environments. The framework integrates a deep-learning-driven environment sensor for object detection and localization, allowing real-time power delivery. Experiments demonstrate simultaneous wireless power and information transfer, with the router capable of detecting and localizing multiple targets and selectively supplying power based on their energy needs. The framework offers low cost and compact size, making it suitable for commercialization of metasurface-based WPT routers. The system uses a stereo camera and intelligent computation unit for dynamic environment sensing and real-time power distribution. The framework also enables simultaneous wireless information and power transfer (SWIPT), enhancing wireless communication efficiency. The proposed WPT system is capable of delivering power to moving and static objects, demonstrating real-time adaptability. The system's ability to focus energy in three-dimensional space provides significant flexibility for WPT applications. The framework also shows potential for integration with wireless communication systems, enabling future advancements in 6G and indoor WPT. Despite challenges such as low power transfer efficiency and electromagnetic interference, the framework offers a promising solution for sustainable energy supply in IoT and industrial applications.