Skylv Web Automation Bot — Skylv 网页 自动化 机器人

Browser 自动化 代理 功能说明

AI驱动的浏览器自动化，执行复杂网页任务。

技术选型 库 特点 适用场景 Playwright 跨浏览器、等待稳定 通用自动化 Puppeteer Chrome专用、Node原生 Chrome深度控制 Selenium 多语言、老牌稳定 多浏览器兼容 DrissionPage Python、轻量 中国网站适配 Playwright 完整实现

• 基础框架

from playwright.同步_API 导入 同步_playwright, Page, Browser, Browser上下文 from dataclasses 导入 dataclass from typing 导入 Optional 导入 json

@dataclass class Task结果: 成功: bool data: Optional[dict] = None error: Optional[str] = None screenshot: Optional[str] = None

class Browser代理: def __init__(self, headless: bool = True, slow_mo: int = 100): self.playwright = 同步_playwright().启动() self.browser: Browser = self.playwright.chromium.launch( headless=headless, slow_mo=slow_mo ) self.上下文: Browser上下文 = self.browser.new_上下文( viewport={"width": 1920, "height": 1080}, user_代理="Mozilla/5.0 (Windows NT 10.0; Win64; x64) 应用le网页Kit/537.36" ) def new_page(self) -> Page: return self.上下文.new_page() def 执行(self, task: str) -> Task结果: """执行AI指令""" page = self.new_page() try: # AI解析任务 steps = self.plan_steps(task) for step in steps: self.执行_step(page, step) return Task结果(成功=True, data={"steps_completed": len(steps)}) except 异常 as e: return Task结果(成功=False, error=str(e)) finally: page.close() def plan_steps(self, task: str) -> 列出[dict]: """AI规划步骤""" # 这里集成LLM进行任务规划 # ... pass def 执行_step(self, page: Page, step: dict): action = step["action"] if action == "goto": page.goto(step["url"], wAIt_until="domcontentloaded") elif action == "命令行工具ck": page.命令行工具ck(step["selector"]) elif action == "type": page.fill(step["selector"], step["text"]) elif action == "wAIt": page.wAIt_for_timeout(step["ms"]) elif action == "screenshot": page.screenshot(path=step["path"]) elif action == "提取": return page.查询_selector(step["selector"]).inner_text() def close(self): self.browser.close() self.playwright.停止()

• 智能元素定位

class SmartLocator: """AI智能定位器""" LOCATOR_STRATEGIES = [ "获取_by_角色", # 无障碍角色（最可靠） "获取_by_label", # 表单标签 "获取_by_placeholder", # 占位符 "获取_by_text", # 文本内容 "locator", # CSS选择器 "xpath", # XPath（最后备选） ] def find(self, page: Page, description: str) -> Locator: """根据描述智能查找元素""" strategies = [ # 按钮查找 lambda: page.获取_by_角色("button", name=description), lambda: page.获取_by_角色("link", name=description), # 输入框查找 lambda: page.获取_by_label(description), lambda: page.获取_by_placeholder(description), # 文本查找 lambda: page.获取_by_text(description, exact=True), lambda: page.获取_by_text(description), # 通配符（放在最后） lambda: page.locator(f"text={description}").first, ] for strategy in strategies: try: locator = strategy() if locator.count() > 0: return locator.first except: continue rAIse ElementNotFound(f"找不到元素: {description}") def find_many(self, page: Page, description: str) -> 列出: """批量查找元素""" # 尝试多种策略 # ... pass

• 表单自动填写

class FormFiller: """智能表单填写""" def fill_form(self, page: Page, form_data: dict): """根据字段描述自动填写""" for field_name, value in form_data.items(): try: # 尝试找对应label的输入框 locator = page.获取_by_label(field_name, exact=False) # 获取输入类型 if locator.count() == 0: # 尝试placeholder locator = page.获取_by_placeholder(field_name) if locator.count() == 0: # 尝试名称属性 locator = page.locator(f'[name="{field_name}"]') if locator.count() > 0: el = locator.first tag = el.evaluate("el => el.tagName") if tag == "SELECT": el.select_option(value) elif tag == "输入": 输入_type = el.获取_attribute("type") if 输入_type in ["检查box", "radio"]: if value: el.检查() else: