Your analytics tool draws its charts on a canvas element, so every scraper and copy-paste attempt returns an empty box.
Omni's captures are real screenshots of rendered pixels, so a capture sees things the DOM cannot: canvas elements, charts, images, and in-page PDFs arrive exactly as displayed. That is the core idea behind capture as prompt: pixels go to vision-capable models as images.
Capture the visible area or a region around the chart, then ask a question about it, run OCR to extract the labels, or use extract-as-table if the data is tabular. Keep captures tight so text stays large after the 1,568 pixel downscale.