We present U-Cker, a web-based interactive video retrieval system developed for participation in the Video Browser Showdown (VBS) 2026. Building upon our previous version designed for the VR4B session at CBMI 2025, where usability for novice users was emphasized, the current system has been extended to address the broader requirements of VBS. U-Cker supports multimodal query formulation by combining free-text and image inputs with adjustable weights, while a built-in query assistance module supports textual inputs through spelling correction, translation, and reformulation. The retrieval engine uses CLIP embeddings and stores feature vectors of millions of keyframes in GPU memory, enabling real-time similarity computation across large-scale datasets. Compared with the previous version, the system has been adapted to handle the expanded datasets and task types of VBS2026, including AVS, KIS, and VQA tasks, as well as additional video collections beyond the V3C dataset. Preliminary evaluations indicate that U-Cker achieves efficient and accurate retrieval performance under diverse task conditions. These results demonstrate the scalability and flexibility of our system and highlight its potential for competitive performance in VBS2026.