Playwright Server MCP

import asyncio import uuid import base64 import os import json import logging from typing import Any, Optional, Dict, List, Union import mcp.types as types from playwright.async_api import async_playwright, TimeoutError as PlaywrightTimeoutError from playwright_server.tools.base import ToolHandler, Property # 获取日志记录器 logger = logging.getLogger('playwright_server.tools.action') class ActionToolHandler(ToolHandler): name = "playwright_action" description = "根据给定的locator 和action 执行操作" inputSchema = [ Property(name="by", typ="string", description="选择元素方式，遵循playwright的选择器原则，可选项: get_by_text,get_by_placeholder,get_by_label,get_by_role,get_by_alt_text,get_by_title,get_by_test_id,get_by_css,locator", required=True), Property(name="by_value", typ="string", description="选择元素的值", required=True), Property(name="action", typ="string", description="操作类型，可选项: fill,click,dblclick,hover,tap,check,uncheck,select_option,press,type,focus,blur,drag_to,screenshot,get_text,count,is_visible,is_enabled,is_hidden,is_disabled,is_checked,get_attribute,evaluate,wait_for,press_sequentially,clear,scroll_into_view,set_input_files,select_text,dispatch_event,get_inner_text,get_inner_html,get_content,all_inner_texts,all_text_contents", required=True), Property(name="value", typ="string", description="操作的值", required=False), Property(name="force", typ="boolean", description="是否强制执行动作，无视actionability检查", required=False), Property(name="delay", typ="number", description="按键之间的延迟(毫秒)", required=False), Property(name="position", typ="string", description="相对于元素的点击位置，格式为: 'x,y'", required=False), Property(name="timeout", typ="number", description="操作超时时间(毫秒)", required=False), Property(name="selector", typ="string", description="拖拽目标元素的选择器，用于drag_to操作", required=False), Property(name="attribute_name", typ="string", description="要获取的属性名称，用于get_attribute操作", required=False), Property(name="script", typ="string", description="要在元素上执行的JavaScript代码，用于evaluate操作", required=False), Property(name="filename", typ="string", description="截图保存的文件名，用于screenshot操作", required=False), Property(name="no_wait_after", typ="boolean", description="是否不等待动作完成后的导航", required=False), Property(name="strict", typ="boolean", description="是否严格匹配元素（如果找到多个元素是否报错）", required=False), Property(name="state", typ="string", description="等待的元素状态，用于wait_for操作，可选: 'attached', 'detached', 'visible', 'hidden'", required=False), Property(name="modifiers", typ="array", description="修饰键，用于键盘操作，可选: 'Alt', 'Control', 'Meta', 'Shift'", required=False), Property(name="button", typ="string", description="鼠标按钮，用于点击操作，可选: 'left', 'right', 'middle'", required=False), Property(name="options", typ="string", description="选择框的选项，用于select_option操作，JSON格式的字符串，可以包含 value, label, index", required=False), Property(name="files", typ="string", description="要上传的文件路径，用于set_input_files操作，多个文件用逗号分隔", required=False), Property(name="event_name", typ="string", description="要分发的事件名称，用于dispatch_event操作", required=False), Property(name="event_data", typ="string", description="事件数据，JSON格式的字符串，用于dispatch_event操作", required=False), Property(name="exact", typ="boolean", description="是否精确匹配文本", required=False), Property(name="has_text", typ="string", description="元素必须包含的文本", required=False) ] async def handle(self, name: str, arguments: dict | None) -> list[types.TextContent | types.ImageContent | types.EmbeddedResource]: logger.info(f"开始执行操作: by={arguments.get('by')}, by_value={arguments.get('by_value')}, action={arguments.get('action')}") try: page = self.get_page() logger.debug(f"当前页面URL: {page.url}") by = arguments.get("by") by_value = arguments.get("by_value") action = arguments.get("action") value = arguments.get("value") force = arguments.get("force", False) delay = arguments.get("delay") position = arguments.get("position") timeout = arguments.get("timeout") selector = arguments.get("selector") attribute_name = arguments.get("attribute_name") script = arguments.get("script") filename = arguments.get("filename", "screenshot.png") no_wait_after = arguments.get("no_wait_after", False) strict = arguments.get("strict", False) state = arguments.get("state") modifiers = arguments.get("modifiers", []) button = arguments.get("button", "left") options_str = arguments.get("options") files_str = arguments.get("files") event_name = arguments.get("event_name") event_data_str = arguments.get("event_data") exact = arguments.get("exact", False) has_text = arguments.get("has_text") logger.debug(f"操作参数: force={force}, delay={delay}, timeout={timeout}, strict={strict}, exact={exact}") # 解析复杂参数 options = {} if options_str: try: logger.debug(f"解析options字符串: {options_str}") options = json.loads(options_str) except json.JSONDecodeError as e: logger.error(f"options参数格式错误: {str(e)}") return [types.TextContent(type="text", text="options参数格式错误，必须是有效的JSON字符串")] files = [] if files_str: logger.debug(f"解析files字符串: {files_str}") files = [f.strip() for f in files_str.split(",")] event_data = {} if event_data_str: try: logger.debug(f"解析event_data字符串: {event_data_str}") event_data = json.loads(event_data_str) except json.JSONDecodeError as e: logger.error(f"event_data参数格式错误: {str(e)}") return [types.TextContent(type="text", text="event_data参数格式错误，必须是有效的JSON字符串")] # 构建选项字典 action_options: Dict[str, Any] = {} if force is not None: action_options["force"] = force if delay is not None: action_options["delay"] = delay if timeout is not None: action_options["timeout"] = timeout if no_wait_after is not None: action_options["no_wait_after"] = no_wait_after if strict is not None: action_options["strict"] = strict if modifiers: action_options["modifiers"] = modifiers if button: action_options["button"] = button if exact is not None: action_options["exact"] = exact logger.debug(f"构建的action_options: {action_options}") # 处理位置参数 if position: try: logger.debug(f"解析位置参数: {position}") x, y = map(float, position.split(',')) action_options["position"] = {"x": x, "y": y} except ValueError as e: logger.error(f"位置格式错误: {str(e)}") return [types.TextContent(type="text", text="位置格式错误，应为'x,y'")] # 根据by参数获取locator locator = None locator_options = {"exact": exact} if exact is not None else {} try: logger.info(f"开始定位元素: by={by}, by_value={by_value}") if by == "get_by_text": locator = page.get_by_text(by_value, **locator_options) elif by == "get_by_placeholder": locator = page.get_by_placeholder(by_value, **locator_options) elif by == "get_by_label": locator = page.get_by_label(by_value, **locator_options) elif by == "get_by_role": locator = page.get_by_role(by_value, name=has_text, **locator_options) elif by == "get_by_alt_text": locator = page.get_by_alt_text(by_value, **locator_options) elif by == "get_by_title": locator = page.get_by_title(by_value, **locator_options) elif by == "get_by_test_id": locator = page.get_by_test_id(by_value) elif by == "get_by_css": locator = page.locator(by_value, has_text=has_text) elif by == "locator": locator = page.locator(by_value, has_text=has_text) else: logger.error(f"不支持的by类型: {by}") return [types.TextContent(type="text", text=f"不支持的by类型: {by}")] logger.debug("元素定位成功") except Exception as e: logger.error(f"定位元素失败: {str(e)}", exc_info=True) return [types.TextContent(type="text", text=f"定位元素失败: {str(e)}")] # 根据action执行不同操作 result = None try: logger.info(f"开始执行操作: action={action}") if action == "fill": logger.debug(f"填充文本: {value}") await locator.fill(value, **action_options) result = "填充成功" elif action == "click": logger.debug("点击元素") await locator.click(**action_options) result = "点击成功" elif action == "dblclick": logger.debug("双击元素") await locator.dblclick(**action_options) result = "双击成功" elif action == "hover": logger.debug("悬停在元素上") await locator.hover(**action_options) result = "悬停成功" elif action == "tap": logger.debug("轻触元素") await locator.tap(**action_options) result = "轻触成功" elif action == "check": logger.debug("选中复选框") await locator.check(**action_options) result = "选中成功" elif action == "uncheck": logger.debug("取消选中复选框") await locator.uncheck(**action_options) result = "取消选中成功" elif action == "select_option": if options: logger.debug(f"选择选项(使用options): {options}") await locator.select_option(**options, **action_options) else: logger.debug(f"选择选项(使用value): {value}") await locator.select_option(value=value, **action_options) result = "选择选项成功" elif action == "press": logger.debug(f"按下按键: {value}") await locator.press(value, **action_options) result = "按键成功" elif action == "type": logger.debug(f"输入文本: {value}") await locator.type(value, **action_options) result = "输入文本成功" elif action == "focus": logger.debug("聚焦元素") await locator.focus(**action_options) result = "聚焦成功" elif action == "blur": logger.debug("失焦元素") await locator.blur(**action_options) result = "失焦成功" elif action == "drag_to": logger.debug(f"拖拽元素到: {selector}") target = page.locator(selector) await locator.drag_to(target, **action_options) result = "拖拽成功" elif action == "screenshot": logger.debug(f"元素截图，保存到: {filename}") screenshot_path = f"{filename}" await locator.screenshot(path=screenshot_path, **action_options) with open(screenshot_path, "rb") as image_file: encoded_string = base64.b64encode(image_file.read()).decode("utf-8") os.remove(screenshot_path) logger.info(f"截图完成，文件大小: {len(encoded_string)} 字节") return [types.ImageContent(type="image", data=encoded_string, mimeType="image/png")] elif action == "get_text": logger.debug("获取元素文本") text = await locator.text_content(**action_options) logger.debug(f"获取到的文本: {text}") result = text elif action == "count": logger.debug("计算匹配元素数量") count = await locator.count() logger.debug(f"匹配的元素数量: {count}") result = str(count) elif action == "is_visible": logger.debug("检查元素是否可见") is_visible = await locator.is_visible(**action_options) logger.debug(f"元素可见性: {is_visible}") result = str(is_visible) elif action == "is_enabled": logger.debug("检查元素是否启用") is_enabled = await locator.is_enabled(**action_options) logger.debug(f"元素启用状态: {is_enabled}") result = str(is_enabled) elif action == "is_hidden": logger.debug("检查元素是否隐藏") is_hidden = await locator.is_hidden(**action_options) logger.debug(f"元素隐藏状态: {is_hidden}") result = str(is_hidden) elif action == "is_disabled": logger.debug("检查元素是否禁用") is_disabled = await locator.is_disabled(**action_options) logger.debug(f"元素禁用状态: {is_disabled}") result = str(is_disabled) elif action == "is_checked": logger.debug("检查元素是否选中") is_checked = await locator.is_checked(**action_options) logger.debug(f"元素选中状态: {is_checked}") result = str(is_checked) elif action == "get_attribute": logger.debug(f"获取元素属性: {attribute_name}") attribute = await locator.get_attribute(attribute_name, **action_options) logger.debug(f"属性值: {attribute}") result = attribute elif action == "evaluate": logger.debug(f"在元素上执行脚本: {script}") evaluated = await locator.evaluate(script, **action_options) logger.debug(f"脚本执行结果: {evaluated}") result = str(evaluated) elif action == "wait_for": if not state: state = "visible" logger.debug(f"等待元素状态: {state}") await locator.wait_for(state=state, **action_options) result = f"等待元素状态 '{state}' 成功" elif action == "press_sequentially": logger.debug(f"序列按键输入: {value}") await locator.press_sequentially(value, **action_options) result = "序列按键输入成功" elif action == "clear": logger.debug("清除元素内容") await locator.clear(**action_options) result = "清除内容成功" elif action == "scroll_into_view": logger.debug("将元素滚动到视图") await locator.scroll_into_view_if_needed(**action_options) result = "滚动到元素成功" elif action == "set_input_files": if files: logger.debug(f"设置文件输入: {files}") await locator.set_input_files(files, **action_options) result = "设置文件输入成功" else: logger.error("没有指定要上传的文件") return [types.TextContent(type="text", text="没有指定要上传的文件")] elif action == "select_text": logger.debug("选择元素中的文本") await locator.select_text(**action_options) result = "选择文本成功" elif action == "dispatch_event": if not event_name: logger.error("没有指定事件名称") return [types.TextContent(type="text", text="没有指定事件名称")] logger.debug(f"分发事件: {event_name}, 数据: {event_data}") await locator.dispatch_event(event_name, event_data, **action_options) result = f"分发事件 '{event_name}' 成功" elif action == "get_inner_text": logger.debug("获取元素内部文本") inner_text = await locator.inner_text(**action_options) logger.debug(f"内部文本: {inner_text}") result = inner_text elif action == "get_inner_html": logger.debug("获取元素内部HTML") inner_html = await locator.inner_html(**action_options) logger.debug(f"内部HTML长度: {len(inner_html)}") result = inner_html elif action == "get_content": logger.debug("获取元素内容") content = await locator.text_content(**action_options) logger.debug(f"内容: {content}") result = content elif action == "all_inner_texts": logger.debug("获取所有内部文本") all_inner_texts = await locator.all_inner_texts() logger.debug(f"内部文本数量: {len(all_inner_texts)}") result = str(all_inner_texts) elif action == "all_text_contents": logger.debug("获取所有文本内容") all_text_contents = await locator.all_text_contents() logger.debug(f"文本内容数量: {len(all_text_contents)}") result = str(all_text_contents) else: logger.error(f"不支持的action类型: {action}") return [types.TextContent(type="text", text=f"不支持的action类型: {action}")] logger.info(f"操作执行成功: {action}") return [types.TextContent(type="text", text=f"{result}")] except PlaywrightTimeoutError as e: logger.error(f"操作超时: {str(e)}", exc_info=True) return [types.TextContent(type="text", text=f"操作超时: {str(e)}")] except Exception as e: logger.error(f"操作执行失败: {str(e)}", exc_info=True) return [types.TextContent(type="text", text=f"操作执行失败: {str(e)}")] except Exception as e: logger.error(f"操作处理器全局异常: {str(e)}", exc_info=True) return [types.TextContent(type="text", text=f"操作处理器发生错误: {str(e)}")]