feat: 新增浏览器自动化插件 (browser_auto)

基于 Playwright 的无头浏览器操控，33 个操作：
- 连接: start/stop/status
- 导航: open/back/forward/reload/title/url
- 交互: click/fill/select/check/hover/press
- 提取: text/html/attr/value/eval/query
- 截图: screenshot (全页/元素)
- 等待: wait/wait-gone/wait-url
- 高级: pdf/cookies/set-cookie/viewport/download

测试: 101 个用例全部通过

plugins/browser_auto.py

@@ -0,0 +1,549 @@
+"""
+浏览器自动化插件 - 无头浏览器操控
+基于 Playwright，支持页面导航、元素交互、截图、JS 执行
+"""
+import os
+import json
+import base64
+import time
+from typing import Optional
+
+
+class Plugin:
+    """浏览器自动化 - 网页操控、表单填写、数据抓取、截图"""
+
+    def __init__(self):
+        self.usage = """浏览器自动化工具
+
+连接:
+  start                          - 启动浏览器实例
+  stop                           - 关闭浏览器
+  status                         - 查看浏览器状态
+
+导航:
+  open <URL>                     - 打开网页
+  back                           - 后退
+  forward                        - 前进
+  reload                         - 刷新页面
+  title                          - 获取页面标题
+  url                            - 获取当前 URL
+
+交互:
+  click <选择器>                 - 点击元素
+  fill <选择器> <文本>           - 填写输入框
+  select <选择器> <值>           - 下拉选择
+  check <选择器>                 - 勾选复选框
+  uncheck <选择器>               - 取消勾选
+  hover <选择器>                 - 鼠标悬停
+  press <按键>                   - 按键 (Enter/Tab/Escape 等)
+
+提取:
+  text [选择器]                  - 获取元素文本 (默认 body)
+  html [选择器]                  - 获取元素 HTML
+  attr <选择器> <属性名>         - 获取元素属性
+  value <选择器>                 - 获取输入框值
+  eval <JS代码>                  - 执行 JavaScript
+  query <选择器>                 - 查询匹配元素列表
+
+截图:
+  screenshot [路径]              - 页面截图 (默认 screenshot.png)
+  screenshot --selector <选择器> - 元素截图
+
+等待:
+  wait <选择器> [超时秒数]       - 等待元素出现
+  wait-gone <选择器> [超时秒数]  - 等待元素消失
+  wait-url <URL片段> [超时秒数]  - 等待 URL 变化
+
+高级:
+  pdf [路径]                     - 保存页面为 PDF
+  cookies                        - 获取所有 cookies
+  set-cookie <JSON>              - 设置 cookie
+  viewport <宽> <高>             - 设置视口大小
+  download <URL> [保存路径]      - 下载文件
+"""
+        self.cli = None
+        self._browser = None
+        self._context = None
+        self._page = None
+        self._pw = None
+
+    def set_cli(self, cli):
+        self.cli = cli
+
+    def get_tool_info(self):
+        return {
+            "name": "browser_auto",
+            "description": "浏览器自动化 - 网页导航、元素交互、表单填写、数据抓取、截图、JS执行",
+            "keywords": ["浏览器", "browser", "网页", "点击", "截图", "screenshot", "爬虫",
+                         "playwright", "自动化", "表单", "抓取", "导航", "puppeteer"],
+            "usage": self.usage
+        }
+
+    def get_mcp_definition(self):
+        return {
+            "name": "browser_auto",
+            "description": "浏览器自动化工具",
+            "inputSchema": {
+                "type": "object",
+                "properties": {
+                    "operation": {"type": "string", "description": "操作类型"},
+                    "url": {"type": "string", "description": "网页 URL"},
+                    "selector": {"type": "string", "description": "CSS 选择器"},
+                    "value": {"type": "string", "description": "输入值"},
+                    "script": {"type": "string", "description": "JavaScript 代码"},
+                    "path": {"type": "string", "description": "保存路径"}
+                },
+                "required": ["operation"]
+            }
+        }
+
+    def convert_mcp_args(self, arguments):
+        op = arguments.get("operation", "")
+        url = arguments.get("url", "")
+        selector = arguments.get("selector", "")
+        return f"{op} {url} {selector}".strip()
+
+    # ── 路由 ──
+
+    def handle(self, args: str) -> str:
+        try:
+            parts = args.strip().split(maxsplit=1)
+            if not parts:
+                return "错误：请提供操作类型"
+
+            op = parts[0].lower()
+            rest = parts[1] if len(parts) > 1 else ""
+
+            handlers = {
+                "start": self._start,
+                "stop": self._stop,
+                "status": self._status,
+                "open": self._open,
+                "back": self._back,
+                "forward": self._forward,
+                "reload": self._reload,
+                "title": self._title,
+                "url": self._url,
+                "click": self._click,
+                "fill": self._fill,
+                "select": self._select,
+                "check": self._check,
+                "uncheck": self._uncheck,
+                "hover": self._hover,
+                "press": self._press,
+                "text": self._text,
+                "html": self._html,
+                "attr": self._attr,
+                "value": self._value,
+                "eval": self._eval,
+                "query": self._query,
+                "screenshot": self._screenshot,
+                "wait": self._wait,
+                "wait-gone": self._wait_gone,
+                "wait-url": self._wait_url,
+                "pdf": self._pdf,
+                "cookies": self._cookies,
+                "set-cookie": self._set_cookie,
+                "viewport": self._viewport,
+                "download": self._download,
+            }
+
+            handler = handlers.get(op)
+            if not handler:
+                return f"错误：不支持的操作 '{op}'"
+            return handler(rest)
+
+        except Exception as e:
+            return f"浏览器自动化错误: {str(e)}"
+
+    # ── 浏览器生命周期 ──
+
+    def _ensure_playwright(self):
+        """确保 Playwright 已安装并可用"""
+        try:
+            from playwright.sync_api import sync_playwright
+            return sync_playwright
+        except ImportError:
+            return None
+
+    def _start(self, args: str) -> str:
+        if self._browser and self._page:
+            return "浏览器已在运行中"
+
+        sync_playwright = self._ensure_playwright()
+        if not sync_playwright:
+            return ("❌ 未安装 Playwright。安装步骤:\n"
+                    "  pip install playwright\n"
+                    "  playwright install chromium")
+
+        try:
+            self._pw = sync_playwright().start()
+            self._browser = self._pw.chromium.launch(headless=True)
+            self._context = self._browser.new_context(
+                viewport={"width": 1280, "height": 720},
+                user_agent="Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36"
+            )
+            self._page = self._context.new_page()
+            return "✅ 浏览器已启动 (headless Chromium, 1280×720)"
+        except Exception as e:
+            return f"❌ 启动失败: {e}"
+
+    def _stop(self, args: str) -> str:
+        try:
+            if self._context:
+                self._context.close()
+            if self._browser:
+                self._browser.close()
+            if self._pw:
+                self._pw.stop()
+        except Exception:
+            pass
+        self._page = None
+        self._context = None
+        self._browser = None
+        self._pw = None
+        return "✅ 浏览器已关闭"
+
+    def _status(self, args: str) -> str:
+        if self._page and not self._page.is_closed():
+            url = self._page.url
+            title = self._page.title()
+            return f"🟢 运行中\n  URL: {url}\n  标题: {title}"
+        return "🔴 未运行"
+
+    def _require_page(self):
+        """确保页面可用，否则自动启动"""
+        if self._page and not self._page.is_closed():
+            return self._page
+        self._start("")
+        return self._page
+
+    # ── 导航 ──
+
+    def _open(self, args: str) -> str:
+        url = args.strip()
+        if not url:
+            return "错误：请提供 URL"
+        if not url.startswith(("http://", "https://", "file://")):
+            url = "https://" + url
+
+        page = self._require_page()
+        try:
+            page.goto(url, wait_until="domcontentloaded", timeout=30000)
+            title = page.title()
+            return f"✅ 已打开: {url}\n  标题: {title}"
+        except Exception as e:
+            return f"❌ 打开失败: {e}"
+
+    def _back(self, args: str) -> str:
+        page = self._require_page()
+        page.go_back()
+        return f"✅ 后退 → {page.url}"
+
+    def _forward(self, args: str) -> str:
+        page = self._require_page()
+        page.go_forward()
+        return f"✅ 前进 → {page.url}"
+
+    def _reload(self, args: str) -> str:
+        page = self._require_page()
+        page.reload()
+        return f"✅ 已刷新: {page.url}"
+
+    def _title(self, args: str) -> str:
+        page = self._require_page()
+        return page.title() or "(无标题)"
+
+    def _url(self, args: str) -> str:
+        page = self._require_page()
+        return page.url
+
+    # ── 交互 ──
+
+    def _click(self, selector: str) -> str:
+        selector = selector.strip()
+        if not selector:
+            return "错误：请提供 CSS 选择器"
+        page = self._require_page()
+        try:
+            page.click(selector, timeout=10000)
+            return f"✅ 已点击: {selector}"
+        except Exception as e:
+            return f"❌ 点击失败: {e}"
+
+    def _fill(self, args: str) -> str:
+        parts = args.strip().split(maxsplit=1)
+        if len(parts) < 2:
+            return "错误：格式: fill <选择器> <文本>"
+        selector, text = parts[0], parts[1]
+        page = self._require_page()
+        try:
+            page.fill(selector, text, timeout=10000)
+            return f"✅ 已填写: {selector}"
+        except Exception as e:
+            return f"❌ 填写失败: {e}"
+
+    def _select(self, args: str) -> str:
+        parts = args.strip().split(maxsplit=1)
+        if len(parts) < 2:
+            return "错误：格式: select <选择器> <值>"
+        selector, value = parts[0], parts[1]
+        page = self._require_page()
+        try:
+            page.select_option(selector, value, timeout=10000)
+            return f"✅ 已选择: {selector} = {value}"
+        except Exception as e:
+            return f"❌ 选择失败: {e}"
+
+    def _check(self, selector: str) -> str:
+        page = self._require_page()
+        try:
+            page.check(selector.strip(), timeout=10000)
+            return f"✅ 已勾选: {selector}"
+        except Exception as e:
+            return f"❌ 勾选失败: {e}"
+
+    def _uncheck(self, selector: str) -> str:
+        page = self._require_page()
+        try:
+            page.uncheck(selector.strip(), timeout=10000)
+            return f"✅ 已取消勾选: {selector}"
+        except Exception as e:
+            return f"❌ 取消勾选失败: {e}"
+
+    def _hover(self, selector: str) -> str:
+        page = self._require_page()
+        try:
+            page.hover(selector.strip(), timeout=10000)
+            return f"✅ 悬停: {selector}"
+        except Exception as e:
+            return f"❌ 悬停失败: {e}"
+
+    def _press(self, key: str) -> str:
+        page = self._require_page()
+        page.keyboard.press(key.strip())
+        return f"✅ 按键: {key}"
+
+    # ── 提取 ──
+
+    def _text(self, selector: str) -> str:
+        selector = selector.strip() or "body"
+        page = self._require_page()
+        try:
+            el = page.query_selector(selector)
+            if el:
+                text = el.inner_text()
+                if len(text) > 3000:
+                    text = text[:3000] + f"\n... (截断，共 {len(text)} 字符)"
+                return text
+            return f"未找到元素: {selector}"
+        except Exception as e:
+            return f"❌ 获取文本失败: {e}"
+
+    def _html(self, selector: str) -> str:
+        selector = selector.strip() or "body"
+        page = self._require_page()
+        try:
+            el = page.query_selector(selector)
+            if el:
+                html = el.inner_html()
+                if len(html) > 5000:
+                    html = html[:5000] + "\n... (截断)"
+                return html
+            return f"未找到元素: {selector}"
+        except Exception as e:
+            return f"❌ 获取 HTML 失败: {e}"
+
+    def _attr(self, args: str) -> str:
+        parts = args.strip().split(maxsplit=1)
+        if len(parts) < 2:
+            return "错误：格式: attr <选择器> <属性名>"
+        selector, attr_name = parts[0], parts[1]
+        page = self._require_page()
+        try:
+            el = page.query_selector(selector)
+            if el:
+                val = el.get_attribute(attr_name)
+                return f"{attr_name} = {val}"
+            return f"未找到元素: {selector}"
+        except Exception as e:
+            return f"❌ 获取属性失败: {e}"
+
+    def _value(self, selector: str) -> str:
+        page = self._require_page()
+        try:
+            el = page.query_selector(selector.strip())
+            if el:
+                return el.input_value()
+            return f"未找到元素: {selector}"
+        except Exception as e:
+            return f"❌ 获取值失败: {e}"
+
+    def _eval(self, script: str) -> str:
+        script = script.strip()
+        if not script:
+            return "错误：请提供 JavaScript 代码"
+        page = self._require_page()
+        try:
+            result = page.evaluate(script)
+            if isinstance(result, (dict, list)):
+                return json.dumps(result, ensure_ascii=False, indent=2)
+            return str(result)
+        except Exception as e:
+            return f"❌ JS 执行失败: {e}"
+
+    def _query(self, selector: str) -> str:
+        selector = selector.strip()
+        if not selector:
+            return "错误：请提供 CSS 选择器"
+        page = self._require_page()
+        try:
+            elements = page.query_selector_all(selector)
+            if not elements:
+                return f"未找到匹配 '{selector}' 的元素"
+
+            result = [f"找到 {len(elements)} 个元素:\n"]
+            for i, el in enumerate(elements[:20]):
+                tag = el.evaluate("el => el.tagName.toLowerCase()")
+                text = el.inner_text()[:80].replace('\n', ' ').strip()
+                result.append(f"  [{i}] <{tag}> {text}")
+            if len(elements) > 20:
+                result.append(f"  ... 还有 {len(elements) - 20} 个")
+            return '\n'.join(result)
+        except Exception as e:
+            return f"❌ 查询失败: {e}"
+
+    # ── 截图 ──
+
+    def _screenshot(self, args: str) -> str:
+        page = self._require_page()
+
+        # 解析参数
+        path = "screenshot.png"
+        selector = None
+        parts = args.strip().split()
+        i = 0
+        while i < len(parts):
+            if parts[i] == "--selector" and i + 1 < len(parts):
+                selector = parts[i + 1]
+                i += 2
+            elif not parts[i].startswith("-"):
+                path = parts[i]
+                i += 1
+            else:
+                i += 1
+
+        try:
+            if selector:
+                el = page.query_selector(selector)
+                if not el:
+                    return f"未找到元素: {selector}"
+                el.screenshot(path=path)
+            else:
+                page.screenshot(path=path, full_page=True)
+
+            size = os.path.getsize(path)
+            return f"✅ 截图已保存: {path} ({size/1024:.1f} KB)"
+        except Exception as e:
+            return f"❌ 截图失败: {e}"
+
+    # ── 等待 ──
+
+    def _wait(self, args: str) -> str:
+        parts = args.strip().split()
+        if not parts:
+            return "错误：请提供选择器"
+        selector = parts[0]
+        timeout = int(parts[1]) * 1000 if len(parts) > 1 else 30000
+        page = self._require_page()
+        try:
+            page.wait_for_selector(selector, timeout=timeout)
+            return f"✅ 元素已出现: {selector}"
+        except Exception:
+            return f"⏱️ 等待超时: {selector}"
+
+    def _wait_gone(self, args: str) -> str:
+        parts = args.strip().split()
+        if not parts:
+            return "错误：请提供选择器"
+        selector = parts[0]
+        timeout = int(parts[1]) * 1000 if len(parts) > 1 else 30000
+        page = self._require_page()
+        try:
+            page.wait_for_selector(selector, state="hidden", timeout=timeout)
+            return f"✅ 元素已消失: {selector}"
+        except Exception:
+            return f"⏱️ 等待超时: {selector}"
+
+    def _wait_url(self, args: str) -> str:
+        parts = args.strip().split()
+        if not parts:
+            return "错误：请提供 URL 片段"
+        url_part = parts[0]
+        timeout = int(parts[1]) * 1000 if len(parts) > 1 else 30000
+        page = self._require_page()
+        try:
+            page.wait_for_url(f"**{url_part}**", timeout=timeout)
+            return f"✅ URL 已变化: {page.url}"
+        except Exception:
+            return f"⏱️ 等待超时: URL 未变化"
+
+    # ── 高级 ──
+
+    def _pdf(self, args: str) -> str:
+        path = args.strip() or "page.pdf"
+        page = self._require_page()
+        try:
+            page.pdf(path=path)
+            size = os.path.getsize(path)
+            return f"✅ PDF 已保存: {path} ({size/1024:.1f} KB)"
+        except Exception as e:
+            return f"❌ PDF 生成失败: {e}"
+
+    def _cookies(self, args: str) -> str:
+        page = self._require_page()
+        cookies = page.context.cookies()
+        if not cookies:
+            return "无 cookies"
+        result = [f"Cookies ({len(cookies)} 个):\n"]
+        for c in cookies[:20]:
+            result.append(f"  {c['name']}: {c['value'][:50]}  (domain: {c.get('domain', '-')})")
+        return '\n'.join(result)
+
+    def _set_cookie(self, args: str) -> str:
+        try:
+            cookie = json.loads(args.strip())
+        except json.JSONDecodeError:
+            return "错误：cookie 必须是 JSON 格式: {\"name\": \"...\", \"value\": \"...\", \"domain\": \"...\"}"
+        page = self._require_page()
+        page.context.add_cookies([cookie])
+        return f"✅ 已设置 cookie: {cookie.get('name', '?')}"
+
+    def _viewport(self, args: str) -> str:
+        parts = args.strip().split()
+        if len(parts) < 2:
+            return "错误：格式: viewport <宽> <高>"
+        try:
+            w, h = int(parts[0]), int(parts[1])
+        except ValueError:
+            return "错误：宽高必须是整数"
+        page = self._require_page()
+        page.set_viewport_size({"width": w, "height": h})
+        return f"✅ 视口已设置: {w}×{h}"
+
+    def _download(self, args: str) -> str:
+        parts = args.strip().split()
+        if not parts:
+            return "错误：格式: download <URL> [保存路径]"
+        url = parts[0]
+        save_path = parts[1] if len(parts) > 1 else os.path.basename(url) or "download"
+        page = self._require_page()
+        try:
+            with page.expect_download() as download_info:
+                page.evaluate(f"window.open('{url}')")
+            download = download_info.value
+            download.save_as(save_path)
+            size = os.path.getsize(save_path)
+            return f"✅ 已下载: {save_path} ({size/1024:.1f} KB)"
+        except Exception as e:
+            return f"❌ 下载失败: {e}"