internal-web-reader 0.2.3__tar.gz → 1.0.1__tar.gz

{internal_web_reader-0.2.3 → internal_web_reader-1.0.1}/PKG-INFO

@@ -1,7 +1,7 @@
 Metadata-Version: 2.4
 Name: internal-web-reader
-Version: 0.2.3
-Summary: MCP Server - let Claude Code read internal web pages via browser cookies
+Version: 1.0.1
+Summary: MCP Server - read internal web pages through your own Chrome browser
 Author: zhanghaoran18006
 License-Expression: MIT
 Keywords: mcp,claude,internal,wiki,confluence,gitlab
@@ -12,13 +12,16 @@ Classifier: Topic :: Software Development :: Libraries
 Requires-Python: >=3.10
 Description-Content-Type: text/markdown
 Requires-Dist: mcp[cli]<2.0.0,>=1.2.0
-Requires-Dist: playwright>=1.40.0
+Requires-Dist: httpx>=0.27.0
+Requires-Dist: websocket-client>=1.7.0
 Requires-Dist: beautifulsoup4>=4.12.0
 Requires-Dist: html2text>=2024.2.26
 
 # Internal Web Reader
 
-让 Claude Code 读内网网页。弹个浏览器窗口，登录一次就行。
+让 Claude Code 用你自己的 Chrome 浏览器读内网网页。
+
+原理：通过 Chrome DevTools Protocol (CDP) 连接到你正在运行的 Chrome，复用已有的登录 session。不需要重新登录。
 
 ## 安装
 
@@ -34,10 +37,10 @@ claude mcp add -s user internal-web-reader -- internal-web-reader
 
 ## 使用
 
-在 Claude Code 里给链接：
+重启 Claude Code，然后给链接：
 
 ```
-> 看看 https://doc.autohome.com.cn/docapi/page/share/share_1Hx8joJJIfo
+> 用 read_web_page 读 https://wiki.yourcompany.com/pages/xxx
 ```
 
-第一次访问需要登录的网站，会弹出 Chrome 窗口。登录一下，之后就不用再登了。
+首次使用时，如果 Chrome 没有启用调试端口，工具会自动重启 Chrome（你的标签页会自动恢复）。之后就直接用了。

internal_web_reader-1.0.1/README.md

@@ -0,0 +1,27 @@
+# Internal Web Reader
+
+让 Claude Code 用你自己的 Chrome 浏览器读内网网页。
+
+原理：通过 Chrome DevTools Protocol (CDP) 连接到你正在运行的 Chrome，复用已有的登录 session。不需要重新登录。
+
+## 安装
+
+```
+pip install internal-web-reader
+```
+
+## 配置 Claude Code
+
+```
+claude mcp add -s user internal-web-reader -- internal-web-reader
+```
+
+## 使用
+
+重启 Claude Code，然后给链接：
+
+```
+> 用 read_web_page 读 https://wiki.yourcompany.com/pages/xxx
+```
+
+首次使用时，如果 Chrome 没有启用调试端口，工具会自动重启 Chrome（你的标签页会自动恢复）。之后就直接用了。

{internal_web_reader-0.2.3 → internal_web_reader-1.0.1}/pyproject.toml

@@ -4,8 +4,8 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "internal-web-reader"
-version = "0.2.3"
-description = "MCP Server - let Claude Code read internal web pages via browser cookies"
+version = "1.0.1"
+description = "MCP Server - read internal web pages through your own Chrome browser"
 readme = "README.md"
 requires-python = ">=3.10"
 license = "MIT"
@@ -19,7 +19,8 @@ classifiers = [
 ]
 dependencies = [
     "mcp[cli]>=1.2.0,<2.0.0",
-    "playwright>=1.40.0",
+    "httpx>=0.27.0",
+    "websocket-client>=1.7.0",
     "beautifulsoup4>=4.12.0",
     "html2text>=2024.2.26",
 ]

internal_web_reader-1.0.1/src/internal_web_reader/__main__.py

@@ -0,0 +1,53 @@
+"""MCP Server — 通过 CDP 连接用户自己的 Chrome 来读网页。"""
+
+import logging
+
+from mcp.server.fastmcp import FastMCP
+
+from .reader import read_page
+
+logging.basicConfig(
+    level=logging.INFO,
+    format="[web-reader] %(message)s",
+    handlers=[logging.StreamHandler()],
+)
+
+mcp = FastMCP("internal-web-reader")
+
+
+@mcp.tool()
+def read_web_page(url: str) -> str:
+    """[READ WEB PAGES] Read any URL using the user's own Chrome browser.
+    Reuses your existing Chrome login sessions — no re-login needed.
+    If Chrome is not running with debug mode, it will auto-restart Chrome
+    (your tabs will be restored). Use this for ALL web pages, especially
+    internal/intranet sites (GitLab, Wiki, docs requiring VPN/SSO).
+    """
+    page = read_page(url)
+
+    lines = [f"# {page.title}", f"> {page.url}", "", page.content]
+    if 0 < len(page.links) <= 30:
+        lines += ["", "---", "## Links"]
+        for lk in page.links:
+            lines.append(f"- [{lk['text']}]({lk['href']})")
+    return "\n".join(lines)
+
+
+@mcp.tool()
+def list_web_links(url: str) -> str:
+    """List all links on a web page."""
+    page = read_page(url)
+    if not page.links:
+        return f"No links found on {url}."
+    lines = [f"# {page.title}", f"> {len(page.links)} links", ""]
+    for lk in page.links:
+        lines.append(f"- [{lk['text']}]({lk['href']})")
+    return "\n".join(lines)
+
+
+def main():
+    mcp.run()
+
+
+if __name__ == "__main__":
+    main()

internal_web_reader-1.0.1/src/internal_web_reader/reader.py

@@ -0,0 +1,334 @@
+"""通过 CDP 连接用户自己的 Chrome，复用已有登录 session。
+
+流程:
+  1. 检测 Chrome 是否在运行
+  2. 如果没启用调试端口 → 自动重启 Chrome 加上 --remote-debugging-port=0
+  3. 从 DevToolsActivePort 文件读取端口
+  4. 通过 CDP 协议连接，打开 URL，拿页面内容
+"""
+
+import json
+import logging
+import os
+import subprocess
+import sys
+import time
+from dataclasses import dataclass, field
+from urllib.parse import urljoin
+
+import httpx
+from bs4 import BeautifulSoup
+import html2text
+
+log = logging.getLogger("internal_web_reader")
+
+_H2T = html2text.HTML2Text()
+_H2T.body_width = 0
+_H2T.ignore_links = False
+_H2T.ignore_images = False
+_H2T.protect_links = True
+_H2T.wrap_links = False
+_H2T.unicode_snob = True
+
+
+@dataclass
+class PageResult:
+    url: str
+    title: str
+    content: str
+    links: list[dict] = field(default_factory=list)
+
+
+def _stderr(msg: str):
+    sys.stderr.write(f"\n[web-reader] {msg}\n")
+    sys.stderr.flush()
+
+
+# ── Chrome detection & launch ────────────────────────────
+
+def _chrome_exe() -> str:
+    """Find Chrome executable path."""
+    candidates = [
+        os.path.expandvars(r"%PROGRAMFILES%\Google\Chrome\Application\chrome.exe"),
+        os.path.expandvars(r"%PROGRAMFILES(X86)%\Google\Chrome\Application\chrome.exe"),
+        os.path.expandvars(r"%LOCALAPPDATA%\Google\Chrome\Application\chrome.exe"),
+    ]
+    for p in candidates:
+        if os.path.isfile(p):
+            return p
+    # Fallback: hope it's on PATH
+    return "chrome"
+
+
+def _user_data_dir() -> str:
+    return os.path.join(
+        os.environ.get("LOCALAPPDATA", ""),
+        "Google", "Chrome", "User Data",
+    )
+
+
+def _chrome_running() -> bool:
+    if sys.platform != "win32":
+        return "chrome" in subprocess.getoutput("pgrep -la chrome 2>/dev/null")
+    result = subprocess.run(
+        ["tasklist", "/FI", "IMAGENAME eq chrome.exe", "/NH"],
+        capture_output=True, text=True, timeout=5,
+    )
+    return "chrome.exe" in result.stdout
+
+
+def _get_debug_port() -> int | None:
+    """Read DevToolsActivePort file to get the debugging port."""
+    port_file = os.path.join(_user_data_dir(), "DevToolsActivePort")
+    if not os.path.isfile(port_file):
+        return None
+    try:
+        with open(port_file) as f:
+            lines = f.read().strip().split("\n")
+        return int(lines[0])
+    except Exception:
+        return None
+
+
+def _restart_chrome_with_debug():
+    """Kill Chrome and relaunch with --remote-debugging-port=0."""
+    _stderr("正在重启 Chrome 以启用调试端口（你的标签页会自动恢复）...")
+
+    # Kill Chrome
+    if sys.platform == "win32":
+        subprocess.run(["taskkill", "/F", "/IM", "chrome.exe"],
+                       capture_output=True, timeout=10)
+    else:
+        subprocess.run(["pkill", "-f", "chrome"], capture_output=True, timeout=10)
+
+    time.sleep(2)
+
+    # Relaunch with debugging port
+    chrome = _chrome_exe()
+    subprocess.Popen(
+        [chrome, "--remote-debugging-port=0", "--restore-last-session"],
+        creationflags=subprocess.CREATE_NO_WINDOW if sys.platform == "win32" else 0,
+    )
+
+    # Wait for DevToolsActivePort to appear
+    _stderr("等待 Chrome 启动...")
+    for _ in range(30):
+        time.sleep(1)
+        port = _get_debug_port()
+        if port:
+            _stderr(f"Chrome 调试端口就绪: {port}")
+            return port
+
+    raise RuntimeError("Chrome 启动超时，无法获取调试端口")
+
+
+def _ensure_debug_port() -> int:
+    """Ensure Chrome has a debugging port available."""
+    port = _get_debug_port()
+    if port:
+        return port
+
+    if _chrome_running():
+        _stderr("Chrome 正在运行但未启用调试端口，需要重启...")
+        return _restart_chrome_with_debug()
+    else:
+        _stderr("Chrome 未运行，正在启动...")
+        chrome = _chrome_exe()
+        subprocess.Popen(
+            [chrome, "--remote-debugging-port=0"],
+            creationflags=subprocess.CREATE_NO_WINDOW if sys.platform == "win32" else 0,
+        )
+        for _ in range(30):
+            time.sleep(1)
+            port = _get_debug_port()
+            if port:
+                _stderr(f"Chrome 调试端口就绪: {port}")
+                return port
+        raise RuntimeError("Chrome 启动超时")
+
+
+# ── CDP communication ────────────────────────────────────
+
+def _cdp_request(port: int, method: str, params: dict | None = None, session_id: str | None = None) -> dict:
+    """Send a CDP command via HTTP (simple target-level commands)."""
+    # For simple commands we can use the HTTP endpoint
+    url = f"http://127.0.0.1:{port}/json"
+    resp = httpx.get(url, timeout=5)
+    return resp.json()
+
+
+def _cdp_ws_command(port: int, target_ws_url: str, method: str, params: dict = None, msg_id: int = 1) -> dict:
+    """Send a CDP command via WebSocket and get the response."""
+    import websocket
+
+    ws = websocket.create_connection(target_ws_url, timeout=30)
+    cmd = {"id": msg_id, "method": method}
+    if params:
+        cmd["params"] = params
+    ws.send(json.dumps(cmd))
+
+    while True:
+        raw = ws.recv()
+        data = json.loads(raw)
+        if data.get("id") == msg_id:
+            ws.close()
+            if "error" in data:
+                raise RuntimeError(f"CDP error: {data['error']}")
+            return data.get("result", {})
+
+
+def _cdp_navigate_and_get(port: int, url: str, timeout: float = 60) -> tuple[str, str, str]:
+    """
+    Navigate to URL in a new CDP target, wait for load, get HTML + title.
+    Returns (final_url, title, html).
+    """
+    import websocket
+
+    # Create a new tab/target
+    resp = httpx.get(f"http://127.0.0.1:{port}/json/new?{url}", timeout=10)
+    target = resp.json()
+    ws_url = target["webSocketDebuggerUrl"]
+    tab_id = target["id"]
+
+    try:
+        ws = websocket.create_connection(ws_url, timeout=timeout)
+
+        # Enable page events
+        ws.send(json.dumps({"id": 1, "method": "Page.enable"}))
+        _recv_response(ws, 1)
+
+        # Navigate
+        ws.send(json.dumps({"id": 2, "method": "Page.navigate", "params": {"url": url}}))
+        nav_result = _recv_response(ws, 2)
+
+        # Wait for page load + SPA rendering
+        deadline = time.time() + timeout
+        load_fired = False
+        while time.time() < deadline:
+            try:
+                ws.settimeout(2)
+                raw = ws.recv()
+                data = json.loads(raw)
+                if data.get("method") == "Page.loadEventFired":
+                    load_fired = True
+                    break
+            except Exception:
+                continue
+
+        # Wait for SPA/dynamic content to render
+        if load_fired:
+            time.sleep(3)
+            # Check if page has meaningful content, if not wait more
+            for attempt in range(5):
+                ws.send(json.dumps({"id": 20 + attempt, "method": "Runtime.evaluate",
+                                    "params": {"expression": "document.body.innerText.length"}}))
+                try:
+                    check = _recv_response(ws, 20 + attempt)
+                    text_len = check.get("result", {}).get("value", 0)
+                    if text_len > 200:
+                        break
+                except Exception:
+                    pass
+                time.sleep(2)
+
+        # Get page info
+        ws.settimeout(10)
+        ws.send(json.dumps({"id": 10, "method": "Runtime.evaluate",
+                            "params": {"expression": "document.title"}}))
+        title_resp = _recv_response(ws, 10)
+        title = title_resp.get("result", {}).get("value", "") or url
+
+        ws.send(json.dumps({"id": 11, "method": "Runtime.evaluate",
+                            "params": {"expression": "document.URL"}}))
+        url_resp = _recv_response(ws, 11)
+        final_url = url_resp.get("result", {}).get("value", url)
+
+        ws.send(json.dumps({"id": 12, "method": "Runtime.evaluate",
+                            "params": {"expression": "document.documentElement.outerHTML",
+                                       "returnByValue": True}}))
+        html_resp = _recv_response(ws, 12)
+        html = html_resp.get("result", {}).get("value", "")
+
+        ws.close()
+        return final_url, title, html
+
+    finally:
+        # Close the tab
+        try:
+            httpx.get(f"http://127.0.0.1:{port}/json/close/{tab_id}", timeout=5)
+        except Exception:
+            pass
+
+
+def _recv_response(ws, expected_id: int) -> dict:
+    """Read WebSocket messages until we get the response for our command ID."""
+    deadline = time.time() + 30
+    while time.time() < deadline:
+        try:
+            ws.settimeout(5)
+            raw = ws.recv()
+            data = json.loads(raw)
+            if data.get("id") == expected_id:
+                if "error" in data:
+                    raise RuntimeError(f"CDP: {data['error']}")
+                return data.get("result", {})
+        except Exception as e:
+            if "timed out" in str(e).lower():
+                continue
+            raise
+    raise TimeoutError(f"CDP response timeout for id={expected_id}")
+
+
+# ── Main API ─────────────────────────────────────────────
+
+def read_page(url: str, timeout: float = 60) -> PageResult:
+    """Read a web page using the user's running Chrome via CDP."""
+    port = _ensure_debug_port()
+    _stderr(f"正在通过 Chrome 读取: {url}")
+
+    final_url, title, html = _cdp_navigate_and_get(port, url, timeout)
+
+    if not html:
+        return PageResult(url=final_url, title=title, content="[empty page]")
+
+    soup = BeautifulSoup(html, "html.parser")
+
+    for tag in soup.find_all(["script", "style", "noscript", "iframe", "svg"]):
+        tag.decompose()
+
+    main = None
+    for sel in ("article", "main", '[role="main"]', ".content", ".article-content",
+                ".markdown-body", "#content", "#main"):
+        el = soup.select_one(sel)
+        if el and len(el.get_text(strip=True)) > 100:
+            main = el
+            break
+    target = main or soup.body or soup
+
+    links = _extract_links(target, final_url)
+    markdown = _H2T.handle(str(target)).strip()
+
+    if len(markdown) > 200_000:
+        markdown = markdown[:200_000] + "\n\n... [truncated]"
+
+    _stderr(f"读取完成: {title}")
+    return PageResult(url=final_url, title=title, content=markdown, links=links)
+
+
+def _extract_links(soup, base_url: str) -> list[dict]:
+    seen: set[str] = set()
+    links: list[dict] = []
+    for a in soup.find_all("a", href=True):
+        href = a["href"].strip()
+        if not href or href.startswith("#") or href.startswith("javascript:") or href.startswith("mailto:"):
+            continue
+        try:
+            absolute = urljoin(base_url, href)
+        except Exception:
+            continue
+        if absolute in seen:
+            continue
+        seen.add(absolute)
+        text = a.get_text(strip=True)[:100] or href
+        links.append({"text": text, "href": absolute})
+    return links

{internal_web_reader-0.2.3 → internal_web_reader-1.0.1}/src/internal_web_reader.egg-info/PKG-INFO

@@ -1,7 +1,7 @@
 Metadata-Version: 2.4
 Name: internal-web-reader
-Version: 0.2.3
-Summary: MCP Server - let Claude Code read internal web pages via browser cookies
+Version: 1.0.1
+Summary: MCP Server - read internal web pages through your own Chrome browser
 Author: zhanghaoran18006
 License-Expression: MIT
 Keywords: mcp,claude,internal,wiki,confluence,gitlab
@@ -12,13 +12,16 @@ Classifier: Topic :: Software Development :: Libraries
 Requires-Python: >=3.10
 Description-Content-Type: text/markdown
 Requires-Dist: mcp[cli]<2.0.0,>=1.2.0
-Requires-Dist: playwright>=1.40.0
+Requires-Dist: httpx>=0.27.0
+Requires-Dist: websocket-client>=1.7.0
 Requires-Dist: beautifulsoup4>=4.12.0
 Requires-Dist: html2text>=2024.2.26
 
 # Internal Web Reader
 
-让 Claude Code 读内网网页。弹个浏览器窗口，登录一次就行。
+让 Claude Code 用你自己的 Chrome 浏览器读内网网页。
+
+原理：通过 Chrome DevTools Protocol (CDP) 连接到你正在运行的 Chrome，复用已有的登录 session。不需要重新登录。
 
 ## 安装
 
@@ -34,10 +37,10 @@ claude mcp add -s user internal-web-reader -- internal-web-reader
 
 ## 使用
 
-在 Claude Code 里给链接：
+重启 Claude Code，然后给链接：
 
 ```
-> 看看 https://doc.autohome.com.cn/docapi/page/share/share_1Hx8joJJIfo
+> 用 read_web_page 读 https://wiki.yourcompany.com/pages/xxx
 ```
 
-第一次访问需要登录的网站，会弹出 Chrome 窗口。登录一下，之后就不用再登了。
+首次使用时，如果 Chrome 没有启用调试端口，工具会自动重启 Chrome（你的标签页会自动恢复）。之后就直接用了。

{internal_web_reader-0.2.3 → internal_web_reader-1.0.1}/src/internal_web_reader.egg-info/requires.txt

@@ -1,4 +1,5 @@
 mcp[cli]<2.0.0,>=1.2.0
-playwright>=1.40.0
+httpx>=0.27.0
+websocket-client>=1.7.0
 beautifulsoup4>=4.12.0
 html2text>=2024.2.26

internal_web_reader-0.2.3/README.md

@@ -1,25 +0,0 @@
-# Internal Web Reader
-
-让 Claude Code 读内网网页。弹个浏览器窗口，登录一次就行。
-
-## 安装
-
-```
-pip install internal-web-reader
-```
-
-## 配置 Claude Code
-
-```
-claude mcp add -s user internal-web-reader -- internal-web-reader
-```
-
-## 使用
-
-在 Claude Code 里给链接：
-
-```
-> 看看 https://doc.autohome.com.cn/docapi/page/share/share_1Hx8joJJIfo
-```
-
-第一次访问需要登录的网站，会弹出 Chrome 窗口。登录一下，之后就不用再登了。

internal_web_reader-0.2.3/src/internal_web_reader/__main__.py

@@ -1,54 +0,0 @@
-"""MCP Server 入口。"""
-
-import atexit
-import logging
-
-from mcp.server.fastmcp import FastMCP
-
-from .reader import read_page, close_browser
-
-logging.basicConfig(
-    level=logging.INFO,
-    format="[web-reader] %(message)s",
-    handlers=[logging.StreamHandler()],
-)
-
-mcp = FastMCP("internal-web-reader")
-
-
-@mcp.tool()
-async def read_web_page(url: str) -> str:
-    """[READ WEB PAGES] Open any URL in a local Chrome browser and return Markdown content.
-    USE THIS TOOL for all web page reading, especially internal/intranet sites
-    (GitLab, Wiki, docs requiring VPN/SSO). Do NOT use the built-in Fetch tool for web pages.
-    First visit to a login-required site will pop up a browser window for you to log in.
-    Session is saved automatically for future visits.
-    """
-    page = await read_page(url)
-
-    lines = [f"# {page.title}", f"> {page.url}", "", page.content]
-    if 0 < len(page.links) <= 30:
-        lines += ["", "---", "## Links"]
-        for lk in page.links:
-            lines.append(f"- [{lk['text']}]({lk['href']})")
-    return "\n".join(lines)
-
-
-@mcp.tool()
-async def list_web_links(url: str) -> str:
-    """List all links on a web page for navigation."""
-    page = await read_page(url)
-    if not page.links:
-        return f"No links found on {url}."
-    lines = [f"# {page.title} - Links", f"> {len(page.links)} links", ""]
-    for lk in page.links:
-        lines.append(f"- [{lk['text']}]({lk['href']})")
-    return "\n".join(lines)
-
-
-def main():
-    mcp.run()
-
-
-if __name__ == "__main__":
-    main()

internal_web_reader-0.2.3/src/internal_web_reader/reader.py

@@ -1,182 +0,0 @@
-"""用 Playwright (async) 启动浏览器读网页。登录一次，session 自动保存。"""
-
-import asyncio
-import logging
-import os
-import sys
-from dataclasses import dataclass, field
-from urllib.parse import urljoin
-
-from bs4 import BeautifulSoup
-import html2text
-
-log = logging.getLogger("internal_web_reader")
-
-_H2T = html2text.HTML2Text()
-_H2T.body_width = 0
-_H2T.ignore_links = False
-_H2T.ignore_images = False
-_H2T.protect_links = True
-_H2T.wrap_links = False
-_H2T.unicode_snob = True
-
-_USER_DATA_DIR = os.path.join(
-    os.environ.get("APPDATA", os.path.expanduser("~")),
-    "internal-web-reader",
-    "browser-profile",
-)
-
-
-@dataclass
-class PageResult:
-    url: str
-    title: str
-    content: str
-    links: list[dict] = field(default_factory=list)
-
-
-_context = None
-_pw = None
-
-
-def _stderr(msg: str):
-    sys.stderr.write(f"\n{'='*60}\n{msg}\n{'='*60}\n")
-    sys.stderr.flush()
-
-
-async def _ensure_browser():
-    global _context, _pw
-    if _context is not None:
-        return _context
-
-    from playwright.async_api import async_playwright
-
-    _pw = await async_playwright().start()
-    os.makedirs(_USER_DATA_DIR, exist_ok=True)
-
-    _context = await _pw.chromium.launch_persistent_context(
-        _USER_DATA_DIR,
-        headless=False,
-        channel="chrome",
-        viewport={"width": 1280, "height": 900},
-        locale="zh-CN",
-        args=["--disable-blink-features=AutomationControlled"],
-    )
-    return _context
-
-
-def _is_login_url(url: str) -> bool:
-    url_lower = url.lower()
-    keywords = ["/login", "/signin", "/sso", "/cas", "/oauth", "/auth", "passport"]
-    return any(k in url_lower for k in keywords)
-
-
-async def _wait_for_login(page, timeout_s: int = 300):
-    _stderr(
-        "  需要登录！请切换到弹出的浏览器窗口完成登录。\n"
-        "  登录成功后会自动继续，不用回来操作终端。"
-    )
-
-    await page.evaluate("""() => {
-        const b = document.createElement('div');
-        b.id = '__iwr__';
-        b.style.cssText = 'position:fixed;top:0;left:0;right:0;z-index:999999;'
-            + 'background:#ff6b35;color:white;padding:16px;text-align:center;'
-            + 'font-size:18px;font-weight:bold;font-family:sans-serif;'
-            + 'box-shadow:0 4px 12px rgba(0,0,0,0.3);';
-        b.textContent = 'Internal Web Reader: Please log in here. Auto-continuing after login...';
-        document.body.prepend(b);
-    }""")
-
-    elapsed = 0
-    while elapsed < timeout_s:
-        await asyncio.sleep(2)
-        elapsed += 2
-
-        current_url = page.url.lower()
-        has_pw = await page.query_selector('input[type="password"]')
-
-        if not _is_login_url(current_url) and not has_pw:
-            await page.evaluate("() => { const e = document.getElementById('__iwr__'); if(e) e.remove(); }")
-            _stderr("  登录成功！正在读取页面...")
-            await page.wait_for_load_state("domcontentloaded", timeout=30_000)
-            return True
-
-    _stderr("  登录等待超时，尝试继续...")
-    return False
-
-
-async def read_page(url: str, timeout: float = 60) -> PageResult:
-    ctx = await _ensure_browser()
-    page = ctx.pages[0] if ctx.pages else await ctx.new_page()
-
-    try:
-        await page.bring_to_front()
-    except Exception:
-        pass
-
-    await page.goto(url, wait_until="domcontentloaded", timeout=timeout * 1000)
-
-    # 登录检测
-    if _is_login_url(page.url) or await page.query_selector('input[type="password"]'):
-        await _wait_for_login(page)
-
-    final_url = page.url
-    html = await page.content()
-    title = await page.title() or url
-
-    soup = BeautifulSoup(html, "html.parser")
-
-    for tag in soup.find_all(["script", "style", "noscript", "iframe", "svg", "nav", "footer", "aside"]):
-        tag.decompose()
-
-    main = None
-    for sel in ("article", "main", '[role="main"]', ".content", ".article-content", ".markdown-body", "#content", "#main"):
-        el = soup.select_one(sel)
-        if el and len(el.get_text(strip=True)) > 100:
-            main = el
-            break
-    target = main or soup.body or soup
-
-    links = _extract_links(target, final_url)
-    markdown = _H2T.handle(str(target)).strip()
-
-    if len(markdown) > 200_000:
-        markdown = markdown[:200_000] + "\n\n... [truncated]"
-
-    return PageResult(url=final_url, title=title, content=markdown, links=links)
-
-
-async def close_browser():
-    global _context, _pw
-    if _context:
-        try:
-            await _context.close()
-        except Exception:
-            pass
-    if _pw:
-        try:
-            await _pw.stop()
-        except Exception:
-            pass
-    _context = None
-    _pw = None
-
-
-def _extract_links(soup, base_url: str) -> list[dict]:
-    seen: set[str] = set()
-    links: list[dict] = []
-    for a in soup.find_all("a", href=True):
-        href = a["href"].strip()
-        if not href or href.startswith("#") or href.startswith("javascript:") or href.startswith("mailto:"):
-            continue
-        try:
-            absolute = urljoin(base_url, href)
-        except Exception:
-            continue
-        if absolute in seen:
-            continue
-        seen.add(absolute)
-        text = a.get_text(strip=True)[:100] or href
-        links.append({"text": text, "href": absolute})
-    return links