hf-mcp-server/packages/e2e-python/gradio-desc-ominparser-5.32.json

{
    "tools": [
        {
            "name": "OmniParser_v2_process",
            "description": "Parses a GUI screen and returns an array of structured elements and a marked-up image showing the bounding boxes. Element array contains: Type (e.g. Icon, Text, Label), Bounding Box, Interactivity and Content",
            "inputSchema": {
                "type": "object",
                "properties": {
                    "image_input": {
                        "title": "ImageData",
                        "type": "string",
                        "format": "a http or https url to a file"
                    },
                    "box_threshold": {
                        "type": "number",
                        "description": "numeric value between 0.01 and 1.0",
                        "default": 0.05
                    },
                    "iou_threshold": {
                        "type": "number",
                        "description": "numeric value between 0.01 and 1.0",
                        "default": 0.1
                    },
                    "imgsz": {
                        "type": "number",
                        "description": "numeric value between 640 and 1920",
                        "default": 640
                    }
                }
            }
        }
    ]
}