Nick: fixes interact tabs

Author: nickscamara

apps/api/src/__tests__/snips/v2/scrape-browser.test.ts

@@ -191,6 +191,63 @@ describe("Scrape browser interact replay", () => {
     scrapeTimeout,
   );
 
+  itIf(canRunReplayHappyPath)(
+    "opens a single content tab (no stray blank tab) when a session starts",
+    async () => {
+      const url = `${TEST_SUITE_WEBSITE}?testId=${crypto.randomUUID()}`;
+      let scrapeId: string | null = null;
+
+      try {
+        const scrapeResponse = await scrapeRaw(
+          {
+            url,
+            origin: "website-replay-test",
+          },
+          identity,
+        );
+
+        expect(scrapeResponse.statusCode).toBe(200);
+        expect(scrapeResponse.body.success).toBe(true);
+        expect(typeof scrapeResponse.body.scrape_id).toBe("string");
+        scrapeId = scrapeResponse.body.scrape_id as string;
+
+        // Session creation primes agent-browser and consolidates tabs, so
+        // user code must see exactly one tab: the content page.
+        const executeResponse = await interactWithReplicaRetry(
+          scrapeId,
+          {
+            language: "node",
+            timeout: 60,
+            code: `
+              console.log(JSON.stringify(page.context().pages().map(p => p.url())));
+            `,
+          },
+          identity,
+        );
+
+        expect(executeResponse.statusCode).toBe(200);
+        expect(executeResponse.body.success).toBe(true);
+
+        const lastLine =
+          executeResponse.body.stdout
+            ?.trim()
+            .split("\n")
+            .filter(Boolean)
+            .pop() ?? "[]";
+        const tabUrls = JSON.parse(lastLine) as string[];
+
+        expect(tabUrls).toHaveLength(1);
+        expect(tabUrls[0]).not.toBe("about:blank");
+        expect(tabUrls[0]).toContain(TEST_SUITE_WEBSITE);
+      } finally {
+        if (scrapeId) {
+          await scrapeStopInteractiveBrowserRaw(scrapeId, identity);
+        }
+      }
+    },
+    scrapeTimeout,
+  );
+
   itIf(!TEST_SELF_HOST)(
     "returns 400 for invalid scrape job id format",
     async () => {

apps/api/src/controllers/v2/scrape-browser.ts

@@ -703,9 +703,21 @@ async function createSessionForScrape(
       );
     }
 
-    // Ensure only one tab exists with the content page in the foreground.
-    // The replay may have created extra tabs. Find the one with content,
-    // close everything else, update the REPL's page var, and bring to front.
+    // Prime agent-browser before consolidating: its first command of a session
+    // spawns an about:blank tab, so trigger it now and let the sync below close it.
+    const primeResult = await browserServiceRequest<BrowserServiceExecResponse>(
+      "POST",
+      `/browsers/${svcResponse.sessionId}/exec`,
+      {
+        code: `agent-browser get url`,
+        language: "bash",
+        timeout: 10,
+        origin: "scrape_replay_sync",
+      },
+    ).catch(() => null);
+
+    // Keep only the content tab, repoint the REPL's page var, bring to front.
+    // agent-browser falls back to the surviving tab when its own is closed.
     await browserServiceRequest(
       "POST",
       `/browsers/${svcResponse.sessionId}/exec`,
@@ -726,19 +738,23 @@ async function createSessionForScrape(
       },
     ).catch(() => {});
 
-    // Sync agent-browser to the correct page
-    const syncResult = await browserServiceRequest<BrowserServiceExecResponse>(
-      "POST",
-      `/browsers/${svcResponse.sessionId}/exec`,
-      {
-        code: `agent-browser get url`,
-        language: "bash",
-        timeout: 10,
-        origin: "scrape_replay_sync",
-      },
-    );
+    // Verify agent-browser is on the content page after cleanup.
+    let agentUrl = (primeResult?.stdout || "").trim();
+    if (!agentUrl || agentUrl === "about:blank") {
+      const syncResult =
+        await browserServiceRequest<BrowserServiceExecResponse>(
+          "POST",
+          `/browsers/${svcResponse.sessionId}/exec`,
+          {
+            code: `agent-browser get url`,
+            language: "bash",
+            timeout: 10,
+            origin: "scrape_replay_sync",
+          },
+        );
+      agentUrl = (syncResult.stdout || "").trim();
+    }
 
-    const agentUrl = (syncResult.stdout || "").trim();
     if (!agentUrl || agentUrl === "about:blank") {
       logger.info("agent-browser on wrong page after replay, navigating", {
         agentUrl,

apps/api/src/lib/scrape-interact/browser-agent.ts

@@ -181,6 +181,30 @@ async function takeSnapshot(browserId: string): Promise<string> {
   }
 }
 
+/**
+ * Keep a single foregrounded content tab, closing agent-browser's stray
+ * about:blank tab (it safely falls back to the surviving tab).
+ */
+async function syncTabs(browserId: string): Promise<void> {
+  try {
+    await browserServiceRequest("POST", `/browsers/${browserId}/exec`, {
+      code: [
+        `const ctx = page.context();`,
+        `const pages = ctx.pages();`,
+        `if (pages.length > 1) {`,
+        `  const target = pages.find(p => { const u = p.url(); return u && u !== 'about:blank'; }) || pages[pages.length - 1];`,
+        `  for (const p of pages) { if (p !== target) await p.close().catch(() => {}); }`,
+        `  page = target;`,
+        `}`,
+        `await page.bringToFront();`,
+      ].join("\n"),
+      language: "node",
+      timeout: 5,
+      origin: "tab_sync",
+    });
+  } catch {}
+}
+
 // ---------------------------------------------------------------------------
 // Main agent — tool-calling loop via AI SDK
 // ---------------------------------------------------------------------------
@@ -213,6 +237,11 @@ export async function executePromptViaBrowserAgent(
   debugLog.add(`Prompt:  ${prompt}\n`);
   logger.info("Agent debug log", { path: debugLog.getPath() });
 
+  // Prime agent-browser and consolidate tabs first: its first command of a
+  // session spawns an about:blank tab that would otherwise get snapshotted.
+  await getCurrentUrl(browserId);
+  await syncTabs(browserId);
+
   const [initialSnapshot, initialUrl] = await Promise.all([
     takeSnapshot(browserId),
     getCurrentUrl(browserId),
@@ -258,23 +287,7 @@ export async function executePromptViaBrowserAgent(
         const output = (result.stdout || result.result || "").trim();
 
         // Ensure only one tab exists and it's in the foreground for live view
-        try {
-          await browserServiceRequest("POST", `/browsers/${browserId}/exec`, {
-            code: [
-              `const ctx = page.context();`,
-              `const pages = ctx.pages();`,
-              `if (pages.length > 1) {`,
-              `  const target = pages.find(p => { const u = p.url(); return u && u !== 'about:blank'; }) || pages[pages.length - 1];`,
-              `  for (const p of pages) { if (p !== target) await p.close().catch(() => {}); }`,
-              `  page = target;`,
-              `}`,
-              `await page.bringToFront();`,
-            ].join("\n"),
-            language: "node",
-            timeout: 5,
-            origin: "tab_sync",
-          });
-        } catch {}
+        await syncTabs(browserId);
 
         const elapsed = Date.now() - start;