vision-corelisted

# vision-core Scene-agnostic eyes for a text-only agent. Read one shared vision config (`VISION_API_KEY` / `VISION_BASE_URL` / `VISION_MODEL` / `LANG`) — no extra credentials. Pick the tool by the question you are answering: | Question | Tool | |---|---| | "What does this image show / say?" | `glance` | | "Where is X?" — a thing you can name | `ground` | | "Where are all the Xs?" — every instance of a kind | `detect` | | "What is its exact shape, size, offset?" | `trace` | | "Cut this box out as its own image file" | `crop` | | "Which colours dominate a region, and which palette value fits it?" | `palette` | | "Where do two images differ?" | `pixel-diff` | | "OCR this long screenshot / scrolling page / chat history" | `long-ocr` | | "Extract the icon/logo foreground as transparent PNG" | `extract-fg` | | "Turn this HTML file into a screenshot" | `html-shot` | `glance` answers what something is; `ground` and `detect` answer where. You give `ground` a description of a particular thing; you give `detect` a kind and it enumerates the instances. Both give real coordinates on a 0-1000 grid scaled to your image — accurate enough to crop, click, and compare. When a number has to be exact (sizes, offsets, shapes), `trace` derives it from the actual pixels. ## Use the provided tools before hand-rolled pixels - cut a box out of an image → `crop`, not Pillow - sample a region's palette → `palette` - compare two images → `pixel-diff` - vectorize to SVG → `trace` - locate / inventory