Create interact_element.py

tools/interact_element.py

@@ -0,0 +1,66 @@
+from smolagents.tools import Tool
+from helium import write, press, click, Text, Link, S
+from selenium.webdriver.common.keys import Keys
+
+def interact_element(driver, selector=None, text=None, action="click", input_text=None, key=None):
+    """
+    Interact with a web element (click, fill text, press keys).
+    
+    Args:
+        driver: Selenium WebDriver instance
+        selector (str): CSS selector to target the element (optional)
+        text (str): Text to locate the element (optional, for Helium)
+        action (str): Action to perform ('click', 'fill', 'press') (default: 'click')
+        input_text (str): Text to input if action='fill' (optional)
+        key (str): Key to press if action='press' (e.g., 'ENTER') (optional)
+    
+    Returns:
+        str: Result of the interaction
+    """
+    try:
+        if selector:
+            element = driver.find_element(By.CSS_SELECTOR, selector)
+        elif text:
+            if action == "click" and Link(text).exists():
+                element = Link(text).web_element
+            else:
+                element = Text(text).web_element
+        else:
+            return "Must provide selector or text"
+
+        if action == "click":
+            click(element)
+            return f"Clicked element with {selector or text}"
+        elif action == "fill":
+            if not input_text:
+                return "input_text required for fill action"
+            write(input_text, into=element)
+            return f"Filled {input_text} into element with {selector or text}"
+        elif action == "press":
+            if not key:
+                return "key required for press action"
+            key_map = {"ENTER": Keys.ENTER, "TAB": Keys.TAB, "RETURN": Keys.RETURN}
+            if key.upper() in key_map:
+                element.send_keys(key_map[key.upper()])
+            else:
+                press(key)
+            return f"Pressed {key} on element with {selector or text}"
+        else:
+            return f"Unknown action: {action}"
+    except Exception as e:
+        return f"Failed to interact with element {selector or text}: {str(e)}"
+
+# Register the tool
+tool = Tool(
+    name="interact_element",
+    description="Interacts with a web element (click, fill text, press keys).",
+    inputs={
+        "selector": {"type": "str", "default": null, "description": "CSS selector to target element"},
+        "text": {"type": "str", "default": null, "description": "Text to locate element"},
+        "action": {"type": "str", "default": "click", "description": "Action: 'click', 'fill', or 'press'"},
+        "input_text": {"type": "str", "default": null, "description": "Text to input for fill action"},
+        "key": {"type": "str", "default": null, "description": "Key to press (e.g., 'ENTER')"}
+    },
+    output_type="str",
+    function=interact_element
+)