qazen-cli 0.1.8 → 0.1.9

package/dist/lib/visionNavigator.js

@@ -114,6 +114,8 @@ export class VisionNavigator {
                 return;
             }
             this.visitedUrls.add(normalizedFinal);
+            // Mark original URL as visited too — prevents re-queuing via redirect.
+            this.visitedUrls.add(normalizedUrl);
             const screenshot = await this.takeScreenshot(url);
             const analysis = await this.analyzeScreenshot(screenshot, url, context, onEvent);
             if (!analysis)
@@ -265,13 +267,9 @@ Return ONLY the JSON object. No markdown, no explanation.`,
             return null;
         }
     }
-    async clickElement(element, pageMap, onEvent) {
-        onEvent({
-            type: "action",
-            message: `Clicking "${element.description}" (${element.visualLocation || element.elementType})`,
-        });
-        const urlBefore = this.page.url();
+    async getElementCoordinates(description, visualLocation) {
         const screenshot = await this.takeScreenshot("click-target");
+        const hint = visualLocation ? ` (${visualLocation})` : "";
         const coordResponse = await this.anthropic.messages.create({
             model: MODEL,
             max_tokens: 200,
@@ -285,7 +283,7 @@ Return ONLY the JSON object. No markdown, no explanation.`,
                         },
                         {
                             type: "text",
-                            text: `Find "${element.description}" on screen.
+                            text: `Find "${description}"${hint} on screen.
 Return ONLY JSON: {"x": number, "y": number, "found": boolean}
 x and y are pixel coordinates (image is 1280x720).
 If not found, return {"x": 0, "y": 0, "found": false}`,
@@ -295,24 +293,50 @@ If not found, return {"x": 0, "y": 0, "found": false}`,
             ],
         });
         const coordText = firstTextBlock(coordResponse.content);
-        let coords;
         try {
-            coords = JSON.parse(stripCodeFences(coordText));
+            return JSON.parse(stripCodeFences(coordText));
         }
         catch {
             return null;
         }
-        if (!coords.found)
+    }
+    async clickElement(element, pageMap, onEvent) {
+        onEvent({
+            type: "action",
+            message: `Clicking "${element.description}"`,
+        });
+        const coords = await this.getElementCoordinates(element.description, element.visualLocation);
+        if (!coords?.found)
             return null;
+        // Set up navigation listener BEFORE clicking so we catch the redirect.
+        let navigationUrl = null;
+        const navigationPromise = this.page.waitForNavigation({
+            timeout: 3000,
+            waitUntil: "domcontentloaded",
+        })
+            .then(() => {
+            navigationUrl = this.page.url();
+        })
+            .catch(() => {
+            /* no navigation — fine */
+        });
         await this.page.mouse.click(coords.x, coords.y);
         this.totalActions++;
-        pageMap.actions_taken.push(`Clicked "${element.description}" at (${coords.x}, ${coords.y})`);
-        await this.page.waitForTimeout(2000);
-        const urlAfter = this.page.url();
-        if (urlAfter !== urlBefore) {
-            const newScreenshot = await this.takeScreenshot(urlAfter);
-            this.screenshots.push(newScreenshot);
-            return urlAfter;
+        await Promise.race([navigationPromise, this.page.waitForTimeout(2000)]);
+        if (navigationUrl) {
+            const normalizedNew = this.normalizeUrl(navigationUrl);
+            // If we landed on an already-mapped page, go back and skip.
+            if (this.visitedUrls.has(normalizedNew)) {
+                onEvent({
+                    type: "vision_analysis",
+                    message: `Skipping ${normalizedNew} — already mapped`,
+                });
+                await this.page.goBack({ waitUntil: "domcontentloaded", timeout: 10000 }).catch(() => { });
+                await this.page.waitForTimeout(1000);
+                return null;
+            }
+            pageMap.actions_taken.push(`Navigated to ${navigationUrl}`);
+            return navigationUrl;
         }
         return null;
     }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "qazen-cli",
-  "version": "0.1.8",
+  "version": "0.1.9",
   "description": "QAZen CLI — capture authenticated browser sessions for enterprise SSO testing",
   "license": "MIT",
   "author": "QAZen",