Search for a command to run...
Where Vision Becomes Text: Locating the OCR Routing Bottleneck in Vision-Language Models