fix(gemini+grok): restore ask on current UIs

clis/gemini/utils.js

@@ -299,7 +299,11 @@ function readGeminiSnapshotScript() {
         transcriptLines,
         composerHasText: composerText.length > 0,
         isGenerating,
-        structuredTurnsTrusted: turns.length > 0 || transcriptLines.length === 0,
+        // After filtering out zero-state greetings, empty turns means
+        // "no actual conversation yet" — still trust the structured diff.
+        // Previously this flipped to false on the /app root because the
+        // welcome text produced transcriptLines > 0.
+        structuredTurnsTrusted: true,
       };
     })()
   `;
@@ -417,9 +421,22 @@ function getTurnsScript() {
       ];
 
       const roots = selectors.flatMap((selector) => Array.from(document.querySelectorAll(selector)));
+      // Filter out zero-state welcome text ("Hi <user>, Where should we start?")
+      // which Gemini renders inside .visible-primary-message containers using
+      // .message-text class. These match our turn selectors but are NOT actual
+      // conversation turns, breaking the strict prefix match in diffTrustedStructuredTurns.
+      const isInRealConversation = (el) => {
+        if (el.closest('[class*="conversation-container"]')) return true;
+        if (el.closest('chat-history, [class*="chat-history"]')) return true;
+        // Exclude zero-state greeting areas
+        if (el.closest('[class*="visible-primary-message"]')) return false;
+        if (el.closest('[class*="zero-state"], [class*="empty-state"], [class*="zero_state"]')) return false;
+        return false;
+      };
       const unique = roots
         .filter((el, index, all) => all.indexOf(el) === index)
         .filter(isVisible)
+        .filter(isInRealConversation)
         .sort((left, right) => {
           if (left === right) return 0;
           const relation = left.compareDocumentPosition(right);
@@ -1917,7 +1934,12 @@ export async function waitForGeminiResponse(page, baseline, promptText, timeoutS
                 lastStructured = structuredCandidate;
                 structuredStableCount = 1;
             }
-            if (!current.isGenerating && structuredStableCount >= 2) {
+            // Gemini's send-button aria-label sticks at "Stop response" after
+            // the response has already finished streaming, making isGenerating
+            // unreliable as a completion signal. Fall back to: if the candidate
+            // text has been stable for ~8 seconds (4 polls at 2s each), treat
+            // it as done regardless of isGenerating.
+            if ((!current.isGenerating && structuredStableCount >= 2) || structuredStableCount >= 4) {
                 return structuredCandidate;
             }
             continue;

clis/grok/ask.js

@@ -63,19 +63,47 @@ async function runDefaultAsk(page, prompt, timeoutMs, newChat) {
         await page.wait(3);
     }
     const promptJson = JSON.stringify(prompt);
+    // Grok's input is a contenteditable div now, not a <textarea>. Use native
+    // CDP typing (page.nativeType) so React recognizes the input and enables Submit.
+    const composerHandle = await page.evaluate(`(() => {
+      const box = document.querySelector('div[contenteditable="true"]')
+                  || document.querySelector('textarea');
+      if (!box) return { ok: false, msg: 'no composer' };
+      box.focus();
+      // Clear any existing text
+      if (box.tagName === 'TEXTAREA') { box.value = ''; }
+      else { box.textContent = ''; box.dispatchEvent(new InputEvent('input', { bubbles: true })); }
+      return { ok: true };
+    })()`);
+    if (!composerHandle?.ok) {
+        return [{ response: '[SEND FAILED] ' + JSON.stringify(composerHandle) }];
+    }
+    if (page.nativeType) {
+        try { await page.nativeType(prompt); } catch { }
+    } else {
+        await page.evaluate(`(() => {
+      const box = document.querySelector('div[contenteditable="true"]') || document.querySelector('textarea');
+      if (!box) return;
+      if (box.tagName === 'TEXTAREA') { box.value = ${promptJson}; }
+      else { box.textContent = ${promptJson}; }
+      box.dispatchEvent(new InputEvent('input', { bubbles: true, data: ${promptJson}, inputType: 'insertText' }));
+    })()`);
+    }
+    await page.wait(1.5);
     const sendResult = await page.evaluate(`(async () => {
     try {
-      const box = document.querySelector('textarea');
-      if (!box) return { ok: false, msg: 'no textarea' };
-      box.focus(); box.value = '';
-      document.execCommand('selectAll');
-      document.execCommand('insertText', false, ${promptJson});
-      await new Promise(r => setTimeout(r, 1500));
-      const btn = document.querySelector('button[aria-label="\\u63d0\\u4ea4"]');
-      if (btn && !btn.disabled) { btn.click(); return { ok: true, msg: 'clicked' }; }
+      // English aria (current Grok UI)
+      const btnEN = document.querySelector('button[aria-label="Submit"]');
+      if (btnEN && !btnEN.disabled && btnEN.getAttribute('aria-disabled') !== 'true') {
+        btnEN.click(); return { ok: true, msg: 'clicked-en' };
+      }
+      // Chinese aria (older UI)
+      const btnCN = document.querySelector('button[aria-label="\\u63d0\\u4ea4"]');
+      if (btnCN && !btnCN.disabled) { btnCN.click(); return { ok: true, msg: 'clicked-cn' }; }
       const sub = [...document.querySelectorAll('button[type="submit"]')].find(b => !b.disabled);
       if (sub) { sub.click(); return { ok: true, msg: 'clicked-submit' }; }
-      box.dispatchEvent(new KeyboardEvent('keydown', { key: 'Enter', code: 'Enter', keyCode: 13, bubbles: true }));
+      const box = document.querySelector('div[contenteditable="true"]') || document.querySelector('textarea');
+      if (box) box.dispatchEvent(new KeyboardEvent('keydown', { key: 'Enter', code: 'Enter', keyCode: 13, bubbles: true }));
       return { ok: true, msg: 'enter' };
     } catch (e) { return { ok: false, msg: e.toString() }; }
   })()`);