Update app.py

app.py

@@ -505,11 +505,24 @@ async def interact_screenshot(request: SessionRequest, auth=Depends(verify_token
     driver = session['driver']
 
     try:
+        # Pegar tamanho real do viewport
+        viewport = driver.execute_script("""
+            return {
+                width: window.innerWidth,
+                height: window.innerHeight,
+                scrollX: window.scrollX,
+                scrollY: window.scrollY,
+                pageWidth: document.documentElement.scrollWidth,
+                pageHeight: document.documentElement.scrollHeight
+            };
+        """)
+
         screenshot = driver.get_screenshot_as_png()
         screenshot_b64 = base64.b64encode(screenshot).decode('utf-8')
 
         return {
             "screenshot": screenshot_b64,
+            "viewport": viewport,
             "url": driver.current_url,
             "title": driver.title
         }
@@ -517,35 +530,87 @@ async def interact_screenshot(request: SessionRequest, auth=Depends(verify_token
         logger.error(f"[INTERACT-SCREENSHOT] Erro: {e}")
         raise HTTPException(status_code=500, detail=str(e))
 
+
 @app.post("/interact/click")
 async def interact_click(request: InteractClickRequest, auth=Depends(verify_token)):
-    """Clica em coordenada específica na página"""
+    """Clica em coordenada específica na página usando ActionChains"""
     session = session_manager.get_session(request.session_id)
     driver = session['driver']
 
     try:
-        actions = ActionChains(driver)
-        actions.move_by_offset(0, 0).perform()
-        actions.reset_actions()
-
-        # Usar JavaScript para clicar na posição exata
-        driver.execute_script(f"""
-            var element = document.elementFromPoint({request.x}, {request.y});
-            if (element) {{
-                element.click();
-                element.focus();
-            }}
-        """)
+        x = request.x
+        y = request.y
 
-        time.sleep(0.5)
+        # Método 1: ActionChains com offset desde o body
+        try:
+            body = driver.find_element(By.TAG_NAME, 'body')
+            actions = ActionChains(driver)
+            actions.move_to_element_with_offset(body, x - body.size['width'] // 2, y - body.size['height'] // 2)
+            actions.click()
+            actions.perform()
+        except Exception:
+            # Método 2: JavaScript com elementFromPoint + click + focus
+            driver.execute_script("""
+                var x = arguments[0];
+                var y = arguments[1];
+                
+                // Scroll para a posição se necessário
+                var vpHeight = window.innerHeight;
+                if (y > vpHeight) {
+                    window.scrollTo(0, y - vpHeight / 2);
+                    y = vpHeight / 2;
+                }
+                
+                var element = document.elementFromPoint(x, y);
+                if (element) {
+                    // Simular eventos de mouse reais
+                    var events = ['mousedown', 'mouseup', 'click'];
+                    events.forEach(function(eventType) {
+                        var event = new MouseEvent(eventType, {
+                            view: window,
+                            bubbles: true,
+                            cancelable: true,
+                            clientX: x,
+                            clientY: y
+                        });
+                        element.dispatchEvent(event);
+                    });
+                    
+                    // Focar no elemento
+                    if (element.focus) element.focus();
+                    
+                    // Se for input/textarea, colocar cursor no final
+                    if (element.tagName === 'INPUT' || element.tagName === 'TEXTAREA') {
+                        element.selectionStart = element.selectionEnd = element.value.length;
+                    }
+                }
+            """, x, y)
+
+        time.sleep(0.8)
 
         screenshot = driver.get_screenshot_as_png()
         screenshot_b64 = base64.b64encode(screenshot).decode('utf-8')
 
+        # Verificar qual elemento está focado
+        focused = driver.execute_script("""
+            var el = document.activeElement;
+            if (el) {
+                return {
+                    tag: el.tagName,
+                    type: el.type || '',
+                    id: el.id || '',
+                    name: el.name || '',
+                    isInput: (el.tagName === 'INPUT' || el.tagName === 'TEXTAREA' || el.isContentEditable)
+                };
+            }
+            return null;
+        """)
+
         return {
             "status": "clicked",
             "x": request.x,
             "y": request.y,
+            "focused_element": focused,
             "screenshot": screenshot_b64,
             "url": driver.current_url,
             "title": driver.title
@@ -554,22 +619,47 @@ async def interact_click(request: InteractClickRequest, auth=Depends(verify_toke
         logger.error(f"[INTERACT-CLICK] Erro: {e}")
         raise HTTPException(status_code=500, detail=str(e))
 
+
 @app.post("/interact/type")
 async def interact_type(request: InteractTypeRequest, auth=Depends(verify_token)):
-    """Digita texto no elemento focado"""
+    """Digita texto no elemento focado usando ActionChains"""
     session = session_manager.get_session(request.session_id)
     driver = session['driver']
 
     try:
-        # Digitar no elemento ativo
-        driver.execute_script("""
-            var el = document.activeElement;
-            if (el && (el.tagName === 'INPUT' || el.tagName === 'TEXTAREA' || el.isContentEditable)) {
-                el.value = el.value + arguments[0];
-                el.dispatchEvent(new Event('input', { bubbles: true }));
-                el.dispatchEvent(new Event('change', { bubbles: true }));
-            }
-        """, request.text)
+        text = request.text
+
+        # Método 1: ActionChains (simula digitação real tecla por tecla)
+        try:
+            actions = ActionChains(driver)
+            actions.send_keys(text)
+            actions.perform()
+        except Exception:
+            # Método 2: JavaScript direto
+            driver.execute_script("""
+                var text = arguments[0];
+                var el = document.activeElement;
+                if (el && (el.tagName === 'INPUT' || el.tagName === 'TEXTAREA' || el.isContentEditable)) {
+                    if (el.isContentEditable) {
+                        el.textContent += text;
+                    } else {
+                        var start = el.selectionStart || el.value.length;
+                        var end = el.selectionEnd || el.value.length;
+                        var before = el.value.substring(0, start);
+                        var after = el.value.substring(end);
+                        el.value = before + text + after;
+                        el.selectionStart = el.selectionEnd = start + text.length;
+                    }
+                    el.dispatchEvent(new Event('input', { bubbles: true }));
+                    el.dispatchEvent(new Event('change', { bubbles: true }));
+                    // KeyUp para frameworks reativos
+                    text.split('').forEach(function(char) {
+                        el.dispatchEvent(new KeyboardEvent('keydown', { key: char, bubbles: true }));
+                        el.dispatchEvent(new KeyboardEvent('keypress', { key: char, bubbles: true }));
+                        el.dispatchEvent(new KeyboardEvent('keyup', { key: char, bubbles: true }));
+                    });
+                }
+            """, text)
 
         time.sleep(0.3)
 
@@ -578,7 +668,7 @@ async def interact_type(request: InteractTypeRequest, auth=Depends(verify_token)
 
         return {
             "status": "typed",
-            "text": request.text,
+            "text": text,
             "screenshot": screenshot_b64,
             "url": driver.current_url,
             "title": driver.title
@@ -587,9 +677,10 @@ async def interact_type(request: InteractTypeRequest, auth=Depends(verify_token)
         logger.error(f"[INTERACT-TYPE] Erro: {e}")
         raise HTTPException(status_code=500, detail=str(e))
 
+
 @app.post("/interact/key")
 async def interact_key(request: InteractKeyRequest, auth=Depends(verify_token)):
-    """Pressiona tecla especial"""
+    """Pressiona tecla especial usando ActionChains"""
     session = session_manager.get_session(request.session_id)
     driver = session['driver']
 
@@ -612,9 +703,10 @@ async def interact_key(request: InteractKeyRequest, auth=Depends(verify_token)):
             raise HTTPException(status_code=400, detail=f"Tecla não suportada: {request.key}")
 
         actions = ActionChains(driver)
-        actions.send_keys(key).perform()
+        actions.send_keys(key)
+        actions.perform()
 
-        time.sleep(0.5)
+        time.sleep(0.8)
 
         screenshot = driver.get_screenshot_as_png()
         screenshot_b64 = base64.b64encode(screenshot).decode('utf-8')
@@ -632,6 +724,7 @@ async def interact_key(request: InteractKeyRequest, auth=Depends(verify_token)):
         logger.error(f"[INTERACT-KEY] Erro: {e}")
         raise HTTPException(status_code=500, detail=str(e))
 
+
 @app.post("/interact/scroll")
 async def interact_scroll(request: SessionRequest, auth=Depends(verify_token)):
     """Rola a página para baixo"""
@@ -655,6 +748,7 @@ async def interact_scroll(request: SessionRequest, auth=Depends(verify_token)):
         logger.error(f"[INTERACT-SCROLL] Erro: {e}")
         raise HTTPException(status_code=500, detail=str(e))
 
+
 @app.post("/interact/clear-and-type")
 async def interact_clear_and_type(request: InteractTypeRequest, auth=Depends(verify_token)):
     """Limpa o campo focado e digita novo texto"""
@@ -662,16 +756,18 @@ async def interact_clear_and_type(request: InteractTypeRequest, auth=Depends(ver
     driver = session['driver']
 
     try:
-        driver.execute_script("""
-            var el = document.activeElement;
-            if (el && (el.tagName === 'INPUT' || el.tagName === 'TEXTAREA')) {
-                el.value = '';
-                el.dispatchEvent(new Event('input', { bubbles: true }));
-                el.value = arguments[0];
-                el.dispatchEvent(new Event('input', { bubbles: true }));
-                el.dispatchEvent(new Event('change', { bubbles: true }));
-            }
-        """, request.text)
+        # Limpar com Ctrl+A + Delete
+        actions = ActionChains(driver)
+        actions.key_down(Keys.CONTROL).send_keys('a').key_up(Keys.CONTROL)
+        actions.send_keys(Keys.DELETE)
+        actions.perform()
+
+        time.sleep(0.2)
+
+        # Digitar novo texto
+        actions = ActionChains(driver)
+        actions.send_keys(request.text)
+        actions.perform()
 
         time.sleep(0.3)
 
@@ -690,6 +786,7 @@ async def interact_clear_and_type(request: InteractTypeRequest, auth=Depends(ver
         raise HTTPException(status_code=500, detail=str(e))
 
 
+
 # ============================================================
 # ROTAS DE BUSCA NO SITE
 # ============================================================