@testivai/witness-webdriverio 0.4.0 → 0.4.1

package/README.md

@@ -99,10 +99,13 @@ Approve writes the temp capture over the baseline and backs the previous baselin
 
 Alongside the screenshot the adapter writes `dom.html` and `elements.json`, the element map: each element's box and a digest of its computed styles, collected by the same function the Playwright and Selenium adapters use.
 
-- When pixels differ but the DOM is structurally identical and no computed style changed, the report flags the change as "likely render noise" (anti-aliasing, font hinting, sub-pixel layout).
+- When pixels differ but the DOM is structurally identical, the computed styles were compared and match, and the diff is small and scattered, the report labels the change "likely render noise" (anti-aliasing, font hinting, sub-pixel layout): likely, not proven. When the diff is concentrated in one area or over 1% of pixels, it does not call it noise: that is a bad capture or a change the style check does not cover (size, overflow, transform, images).
 - When the DOM is identical but computed styles differ, it reports a style-only change, which is real, and names the restyled elements.
+- When the DOM is identical but the computed styles could not be compared (no element map on one side: `skipElementMap`, a blocked collector script, or a baseline captured without one), it says "styles not compared": could be render noise or a stylesheet-only change. It does not call that noise, and `noiseAutoPass` never passes it.
 - When the DOM also differs, it shows added / removed / attribute-change counts so you can decide whether the change is intentional.
 
+These are labels on a `changed` snapshot. Nothing auto-passes unless you configure a pass criterion in `.testivai/config.json` (`maxDiffPercent`, `maxDiffPixels`, `noiseAutoPass`, `shiftTolerance`).
+
 The element map also powers region attribution (which selectors a diff region covers) and page-shift detection.
 
 WebDriver's screenshot is the **viewport**, not the full page, so the element map keeps only the elements inside the captured viewport, in screenshot coordinates. Size the browser window to what you want compared, and capture at the same scroll position each run.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@testivai/witness-webdriverio",
-  "version": "0.4.0",
+  "version": "0.4.1",
   "description": "WebdriverIO adapter for TestivAI Witness — local-first visual regression with pixel + DOM comparison",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",
@@ -54,7 +54,7 @@
     }
   },
   "dependencies": {
-    "@testivai/witness": "2.1.1"
+    "@testivai/witness": "2.2.0"
   },
   "devDependencies": {
     "@types/jest": "^30.0.0",