autokap 1.5.4 → 1.5.7

package/dist/selector-resolver.js

@@ -142,8 +142,20 @@ function findNodeByFuzzyText(tree, description) {
     return null;
 }
 function buildSelectorFromNode(node) {
+    if (isUsableSourceRefSelector(node.sourceRef)) {
+        return node.sourceRef;
+    }
     // Try to build the most specific selector from node attributes
     const ref = node.sourceRef;
+    // Extract data-ak / data-testid-style automation hooks
+    for (const attr of ['data-ak', 'data-ak-interact', 'data-testid', 'data-test', 'data-qa', 'data-cy']) {
+        const attrMatch = ref.match(new RegExp(`${attr}="([^"]+)"`));
+        if (attrMatch)
+            return `[${attr}="${escapeText(attrMatch[1])}"]`;
+        const value = node.attributes[attr];
+        if (value)
+            return `[${attr}="${escapeText(value)}"]`;
+    }
     // Extract data-testid
     const testIdMatch = ref.match(/data-testid="([^"]+)"/);
     if (testIdMatch)
@@ -159,11 +171,20 @@ function buildSelectorFromNode(node) {
     if (role) {
         return `${tag}[role="${role}"]`;
     }
-    // Fallback: use AKNode id as a CSS selector via data attribute
-    if (node.id) {
-        return `[data-ak-id="${node.id}"]`;
+    if (node.label) {
+        return `${tag}:has-text("${escapeText(node.label.slice(0, 50))}")`;
     }
-    return `${tag}:has-text("${escapeText(node.label.slice(0, 50))}")`;
+    return null;
+}
+function isUsableSourceRefSelector(sourceRef) {
+    const ref = (sourceRef ?? '').trim();
+    if (!ref || ref.startsWith('<'))
+        return false;
+    return ref === 'body'
+        || ref.startsWith('#')
+        || ref.startsWith('.')
+        || ref.startsWith('[')
+        || /^[a-z][a-z0-9-]*(?:[#.[\s>:]|$)/i.test(ref);
 }
 function tokenize(text) {
     return text

package/dist/smart-wait.js

@@ -17,6 +17,7 @@ export async function smartWaitForStability(adapter, options) {
     const maxWait = options?.maxWaitMs ?? 10000;
     const startTime = Date.now();
     const waitedFor = [];
+    const unresolvedLoaders = [];
     // 1. Wait for no visible spinners/skeletons
     const spinnerSelectors = [
         '[class*="spinner"]',
@@ -57,6 +58,14 @@ export async function smartWaitForStability(adapter, options) {
                         await sleep(300);
                     }
                 }
+                const stillVisible = await adapter.waitFor({
+                    selector,
+                    state: 'visible',
+                    timeoutMs: 100,
+                }).catch(() => false);
+                if (stillVisible) {
+                    unresolvedLoaders.push(selector);
+                }
             }
         }
         catch {
@@ -70,8 +79,8 @@ export async function smartWaitForStability(adapter, options) {
         waitedFor.push('stability-pause');
     }
     return {
-        stable: true,
-        waitedFor,
+        stable: unresolvedLoaders.length === 0,
+        waitedFor: unresolvedLoaders.length > 0 ? unresolvedLoaders : waitedFor,
         waitedMs: Date.now() - startTime,
     };
 }

package/dist/video-narration-schema.d.ts

@@ -634,6 +634,16 @@ export declare const VideoIngestPayloadSchema: z.ZodObject<{
             captureId: z.ZodOptional<z.ZodString>;
             captureName: z.ZodOptional<z.ZodString>;
             elementSelector: z.ZodOptional<z.ZodString>;
+            outscale: z.ZodOptional<z.ZodObject<{
+                padding: z.ZodOptional<z.ZodNumber>;
+                paddingTop: z.ZodOptional<z.ZodNumber>;
+                paddingRight: z.ZodOptional<z.ZodNumber>;
+                paddingBottom: z.ZodOptional<z.ZodNumber>;
+                paddingLeft: z.ZodOptional<z.ZodNumber>;
+                paddingPercent: z.ZodOptional<z.ZodNumber>;
+                clampToViewport: z.ZodOptional<z.ZodBoolean>;
+                backgroundColor: z.ZodOptional<z.ZodString>;
+            }, z.core.$strict>>;
             description: z.ZodString;
             postcondition: z.ZodObject<{
                 type: z.ZodEnum<{

package/dist/web-playwright-local.d.ts

@@ -5,7 +5,7 @@
  * This is the first (and for now only) RuntimeAdapter implementation.
  */
 import type { Browser } from './browser.js';
-import type { AKTree, VideoPageSignals } from './types.js';
+import type { AKTree, OutscaleConfig, VideoPageSignals } from './types.js';
 import type { RuntimeAdapter, ClickOptions, WaitCondition, RecordingOptions, RecordingResult, SemanticTarget } from './execution-types.js';
 import { type ResolveOptions } from './semantic-resolver.js';
 export declare class WebPlaywrightLocal implements RuntimeAdapter {
@@ -63,7 +63,7 @@ export declare class WebPlaywrightLocal implements RuntimeAdapter {
         method: string | null;
     }>;
     takeScreenshot(): Promise<Buffer>;
-    takeElementScreenshot(selector: string): Promise<Buffer>;
+    takeElementScreenshot(selector: string, outscale?: OutscaleConfig): Promise<Buffer>;
     takeCleanScreenshot(): Promise<Buffer>;
     beginRecording(options: RecordingOptions): Promise<void>;
     getElementBoundingBox(selector: string): Promise<{

package/dist/web-playwright-local.js

@@ -316,8 +316,8 @@ export class WebPlaywrightLocal {
     async takeScreenshot() {
         return this.browser.takeScreenshot();
     }
-    async takeElementScreenshot(selector) {
-        const { buffer } = await this.browser.screenshotBySelector(selector);
+    async takeElementScreenshot(selector, outscale) {
+        const { buffer } = await this.browser.screenshotBySelector(selector, outscale);
         return buffer;
     }
     async takeCleanScreenshot() {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "autokap",
-  "version": "1.5.4",
+  "version": "1.5.7",
   "description": "AI-powered CLI tool for capturing clean screenshots of websites",
   "type": "module",
   "main": "./dist/index.js",