Update app.py

app.py

@@ -1290,7 +1290,124 @@ async def type_text(request: dict, token: str = Depends(verify_token)):
         except:
             raise HTTPException(status_code=500, detail=str(e))
 
+# ===================== CLICK ELEMENT =====================
+@app.post("/click-element")
+async def click_element(request: dict, token: str = Depends(verify_token)):
+    """Click at specific coordinates on the page"""
+    try:
+        session_id = request.get("session_id")
+        x = request.get("x", 0)
+        y = request.get("y", 0)
+        img_width = request.get("img_width", 1920)
+        img_height = request.get("img_height", 1080)
 
+        session = session_manager.get_session(session_id)
+        if not session:
+            raise HTTPException(status_code=404, detail="Sessao nao encontrada")
+
+        driver = session["driver"]
+
+        # Recuperar janela ativa
+        try:
+            handles = driver.window_handles
+            if handles:
+                driver.switch_to.window(handles[-1])
+        except Exception:
+            pass
+
+        # Obter tamanho real da viewport do Selenium
+        viewport_width = driver.execute_script("return window.innerWidth;")
+        viewport_height = driver.execute_script("return window.innerHeight;")
+
+        # Escalar coordenadas da imagem de preview para a viewport real
+        actual_x = int(x * viewport_width / img_width) if img_width > 0 else int(x)
+        actual_y = int(y * viewport_height / img_height) if img_height > 0 else int(y)
+
+        # Garantir limites
+        actual_x = max(0, min(actual_x, viewport_width - 1))
+        actual_y = max(0, min(actual_y, viewport_height - 1))
+
+        logger.info(f"Click: preview({x},{y}) img({img_width}x{img_height}) -> real({actual_x},{actual_y}) viewport({viewport_width}x{viewport_height})")
+
+        # Detectar elemento antes do clique
+        element_info = driver.execute_script(f"""
+            var elem = document.elementFromPoint({actual_x}, {actual_y});
+            if (elem) {{
+                return {{
+                    tagName: elem.tagName || '',
+                    id: elem.id || '',
+                    text: (elem.textContent || '').substring(0, 100).trim(),
+                    type: elem.type || '',
+                    href: elem.href || '',
+                    className: (elem.className || '').substring(0, 100)
+                }};
+            }}
+            return null;
+        """)
+
+        # Clicar via JavaScript (mais confiavel que ActionChains)
+        driver.execute_script(f"""
+            var elem = document.elementFromPoint({actual_x}, {actual_y});
+            if (elem) {{
+                elem.scrollIntoView({{block: 'nearest'}});
+                elem.click();
+            }} else {{
+                var evt = new MouseEvent('click', {{
+                    bubbles: true,
+                    cancelable: true,
+                    clientX: {actual_x},
+                    clientY: {actual_y},
+                    view: window
+                }});
+                document.elementFromPoint({actual_x}, {actual_y})?.dispatchEvent(evt);
+            }}
+        """)
+
+        import time
+        time.sleep(2)
+
+        # Recuperar janela caso tenha mudado
+        try:
+            handles = driver.window_handles
+            if handles:
+                driver.switch_to.window(handles[-1])
+        except Exception:
+            pass
+
+        # Screenshot
+        screenshot = None
+        try:
+            screenshot = driver.get_screenshot_as_base64()
+        except Exception:
+            pass
+
+        current_url = ""
+        title = ""
+        try:
+            current_url = driver.current_url
+            title = driver.title
+        except Exception:
+            pass
+
+        session["last_url"] = current_url
+        session["last_activity"] = time.time()
+
+        return {
+            "success": True,
+            "message": f"Clique em ({actual_x}, {actual_y})",
+            "clicked": element_info,
+            "element": f"{element_info.get('tagName', '')}: {element_info.get('text', '')}" if element_info else "Nenhum elemento",
+            "url": current_url,
+            "current_url": current_url,
+            "title": title,
+            "screenshot": screenshot
+        }
+
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.error(f"Erro no clique: {e}")
+        return {"success": False, "message": str(e), "screenshot": None}
 # ============================================================
 # SEARCH SITE
 # ============================================================