@proagentstore/cli 0.4.4 → 0.4.6

package/dist/browser-runner/runner.js

@@ -594,6 +594,11 @@ export class LocalRunner {
         const initial = this.browserContext.pages()[0];
         if (initial)
             this.trackPage(initial);
+        // A seeded real profile can restore MANY tabs on launch; keep a single clean
+        // workspace tab so the agent's page is unambiguous and visible (not lost among
+        // restored blank tabs — the "why is the window blank?" symptom).
+        await this.closeStalePages().catch(() => undefined);
+        await this.activePage?.bringToFront().catch(() => undefined);
         return this.browserContext;
     }
     /** Mark a page as the active one; fall back to the newest remaining page when it closes. */
@@ -623,11 +628,15 @@ export class LocalRunner {
     /** The page the brain drives and the human takes over — created on demand. */
     async getActivePage() {
         const context = await this.getBrowserContext();
-        if (this.activePage && !this.activePage.isClosed())
-            return this.activePage;
-        const page = context.pages().at(-1) ?? (await context.newPage());
-        this.activePage = page;
-        return page;
+        if (!this.activePage || this.activePage.isClosed()) {
+            this.activePage = context.pages().at(-1) ?? (await context.newPage());
+        }
+        // Keep the agent's working tab in the FOREGROUND so the user actually SEES what
+        // the agent is doing (and can act on it during a takeover). Without this, on a
+        // real Chrome profile that restores several tabs, the agent may drive a
+        // background tab while the user stares at a blank foreground tab.
+        await this.activePage.bringToFront().catch(() => undefined);
+        return this.activePage;
     }
     /** The CDP endpoint of the launched Chrome (from DevToolsActivePort in its profile). */
     async readCdpEndpoint() {
@@ -680,6 +689,17 @@ export class LocalRunner {
         const head = (cut >= 0 ? text.slice(0, cut) : text).replace(/```[a-z]*\n?|```/g, " ").replace(/\s+/g, " ").trim();
         return head.slice(0, 400);
     }
+    /** A small JPEG screenshot (base64, no data: prefix) of the page — one per action,
+     *  so the cloud can store it per step and the whole run can be replayed visually. */
+    async shot(page) {
+        try {
+            const buf = await page.screenshot({ type: "jpeg", quality: 45 });
+            return buf.toString("base64");
+        }
+        catch {
+            return undefined;
+        }
+    }
     /**
      * After a WRITE action, read the field's REAL value back via the standard
      * browser_evaluate tool (evaluates on the element by ref — no reinvented DOM
@@ -823,6 +843,7 @@ export class LocalRunner {
                 url: active.url(),
                 title: await active.title().catch(() => ""),
                 challenge: await detectHumanChallenge(active),
+                screenshot: await this.shot(active),
             };
         }
         const mcp = await this.getMcp();
@@ -846,6 +867,7 @@ export class LocalRunner {
                 title: await settled.title().catch(() => ""),
                 challenge: await detectHumanChallenge(settled),
                 feedback: dialogMsg ? `a native dialog was accepted: "${dialogMsg}"` : "a native dialog was accepted",
+                screenshot: await this.shot(settled),
             };
         }
         if (res.isError)
@@ -878,6 +900,7 @@ export class LocalRunner {
             title: await active.title().catch(() => ""),
             challenge: await detectHumanChallenge(active),
             feedback: feedback || undefined,
+            screenshot: await this.shot(active),
         };
     }
     // ── Agent-driven application lifecycle (called by the remote Workflow brain) ──

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@proagentstore/cli",
-  "version": "0.4.4",
+  "version": "0.4.6",
   "description": "CLI for creating, publishing, and running ProAgentStore agents",
   "license": "MIT",
   "type": "module",