Machine: screen, image matching & OCR methods
| Method | Returns | Notes |
|---|---|---|
| m.screenshot(path: string) | Result[string, string] | Returns saved path. "" → auto-named in .vmlab/screenshots/ |
| m.wait_for_image(image: string, timeout_secs: int) | Result[Match, string] | Threshold 0.9, polls every 1 s |
| m.wait_for_image_opts(image, timeout_secs, threshold: float, region: List[int]) | Result[Match, string] | region = [x, y, w, h] or [] |
| m.wait_for_any(images: List[string], timeout_secs: int) | Result[Match, string] | First image to appear wins |
| m.find_image(image: string) | Result[Option[Match], string] | Single shot, no waiting |
| m.ocr() | Result[string, string] | Tesseract over the whole screen |
| m.ocr_region(region: List[int]) | Result[string, string] | Cropped [x, y, w, h] |
| m.wait_for_text(pattern: string, timeout_secs: int) | Result[Match, string] | Regex over OCR text; only m.text/m.score meaningful (coords are 0) |
Reference images resolve relative to the lab root; convention is an images/ folder beside vmlab.wcl. Matching is normalized cross-correlation on grayscale. Hits come back as a Match.
These are the Display capability, and they are on every machine. A machine that reports no framebuffer fails the call with *"machine web has no display"* — naming what this machine does not offer, not what its kind could never have. No container reports a display today; one running a display server would, and this page would need no edit.