coordinate/plugins/tls_plugin2.py

import time
import json
import random
import re
import os
import uuid
import shutil
from datetime import datetime
from typing import List, Dict, Optional, Any, Callable
from urllib.parse import urljoin, urlparse, urlencode

# DrissionPage 核心
from DrissionPage import ChromiumPage, ChromiumOptions

from vs_plg import IVSPlg
from vs_types import VSPlgConfig, AppointmentType, VSQueryResult, VSBookResult, AvailabilityStatus, TimeSlot, DateAvailability, NotFoundError, PermissionDeniedError, RateLimiteddError, SessionExpiredOrInvalidError, BizLogicError 
from utils.cloudflare_bypass_for_scraping import CloudflareBypasser
from toolkit.proxy_tunnel import ProxyTunnel


class BrowserResponse:
    """模拟 requests.Response"""
    def __init__(self, result_dict):
        result_dict = result_dict or {}
        self.status_code = result_dict.get('status', 0)
        self.text = result_dict.get('body', '')
        self.headers = result_dict.get('headers', {})
        self.url = result_dict.get('url', '')
        self._json = None

    def json(self):
        if self._json is None:
            if not self.text:
                return {}
            try:
                self._json = json.loads(self.text)
            except:
                self._json = {}
        return self._json

class TlsPlugin2(IVSPlg):
    """
    TLSContact 签证预约插件 (DrissionPage 版)
    """

    def __init__(self, group_id: str):
        self.group_id = group_id
        self.config: Optional[VSPlgConfig] = None
        self.free_config: Dict[str, Any] = {}
        self.is_healthy = True
        self.logger = None
        
        # 浏览器实例
        self.page: Optional[ChromiumPage] = None
        
        self.travel_group: Optional[Dict] = None
        
        # --- [核心修改] 并发隔离与资源管理 ---
        # 生成唯一实例 ID
        self.instance_id = uuid.uuid4().hex[:8]
        self.root_workspace = os.path.abspath(os.path.join("temp_browser_data", f"{self.group_id}_{self.instance_id}"))
        # 定义子目录：代理插件目录 & 浏览器用户数据目录
        self.user_data_path = os.path.join(self.root_workspace, "user_data")
    
        # 确保根目录存在 (子目录由具体逻辑创建)
        if not os.path.exists(self.root_workspace):
            os.makedirs(self.root_workspace)
            
        # 持有隧道实例
        self.tunnel = None
        
        self.session_create_time: float = 0

    def get_group_id(self) -> str:
        return self.group_id
    
    def set_log(self, logger: Callable[[str], None]) -> None:
        self.logger = logger
    
    def _log(self, message):
        if self.logger:
            self.logger(f'[TlsPlugin] [{self.group_id}] {message}')
        else:
            print(f'[TlsPlugin] [{self.group_id}] {message}')

    def set_config(self, config: VSPlgConfig):
        self.config = config
        self.free_config = config.free_config or {}

    def health_check(self) -> bool:
        if not self.is_healthy:
            return False
        if self.page is None:
            return False
        try:
            if not self.page.run_js("return 1;"):
                return False
        except:
            return False
            
        if self.config.session_max_life > 0:
            current_time = time.time()
            elapsed_time = current_time - self.session_create_time
            if elapsed_time > self.config.session_max_life * 60:
                self._log(f"Session expired.")
                return False
        return True
    
    def _save_screenshot(self, name_prefix):
        try:
            timestamp = int(time.time())
            filename = f"{self.instance_id}_{name_prefix}_{timestamp}.jpg"
            save_path = os.path.join("data", filename)
            os.makedirs("data", exist_ok=True)
            
            # [修改] 改为 full_page=False，防止页面结构异常导致截图失败
            # 这样能截取到浏览器当前可视区域，最适合调试“卡住”的情况
            self.page.get_screenshot(path=save_path, full_page=False)
            
            self._log(f"Screenshot saved to {save_path}")
        except Exception as e:
            self._log(f"Failed to save screenshot: {e}")

    def create_session(self):
        """
        全浏览器会话创建：过盾 -> JS注入登录 -> 原生跳转
        """
        self._log(f"Initializing Session (ID: {self.instance_id})...")
        co = ChromiumOptions()
        # -------------------------------------------------------------
        # [核心修复] 解决 'not enough values to unpack'
        # -------------------------------------------------------------
        # 1. 不要用 co.auto_port()，因为它依赖解析 stdout，会被 DBus 报错干扰
        # 2. 我们手动随机生成一个端口
        import random
        import socket
        
        def get_free_port():
            with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
                s.bind(('', 0))
                return s.getsockname()[1]
        
        debug_port = get_free_port()
        self._log(f"Assigned Debug Port: {debug_port}")
        
        # 3. 强制指定端口，DrissionPage 就会直接连接，不再解析日志
        co.set_local_port(debug_port)
        
        # --- [关键配置] 设置独立的用户数据目录 ---
        # 这样每个实例的 Cache, Cookies, LocalStorage 都是完全隔离的
        # 同时也防止了多进程争抢同一个 Default 文件夹导致的崩溃
        co.set_user_data_path(self.user_data_path)
        
        # --- 1. 指定浏览器路径 (适配 Docker) ---
        chrome_path = os.getenv("CHROME_BIN")
        if chrome_path and os.path.exists(chrome_path):
            co.set_paths(browser_path=chrome_path)
        
        # --- [核心修改] 代理配置 ---
        if self.config.proxy and self.config.proxy.ip:
            p = self.config.proxy
            
            if p.username and p.password:
                self._log(f"Starting Proxy Tunnel for {p.ip}...")
                
                # 1. 启动本地隧道
                self.tunnel = ProxyTunnel(p.ip, p.port, p.username, p.password)
                local_proxy = self.tunnel.start()
                
                self._log(f"Tunnel started at {local_proxy}")
                
                # 2. Chrome 连接本地免密端口
                # 必须使用 --proxy-server 强制指定，绝对稳健
                co.set_argument(f'--proxy-server={local_proxy}')
                
            else:
                # 无密码代理，直接用
                proxy_str = f"{p.scheme}://{p.ip}:{p.port}"
                co.set_argument(f'--proxy-server={proxy_str}')
        else:
            self._log("[WARN] No proxy configured!")

        co.headless(False)
        co.set_argument('--no-sandbox')
        co.set_argument('--disable-gpu')
        # Docker 默认 /dev/shm 只有 64MB，Chromium 很容易爆内存崩溃
        co.set_argument('--disable-dev-shm-usage')
        co.set_argument('--window-size=1920,1080')
        co.set_argument('--disable-blink-features=AutomationControlled')

        try:
            self.page = ChromiumPage(co)
            
            apt_config = self.free_config.get('apt_config', {})
            if not apt_config:
                raise NotFoundError("apt_config config missing")

            login_url = "https://visas-fr.tlscontact.com/en-us/login"
            params = {
                "issuerId": apt_config["code"],
                "country": apt_config["country"],
                "vac": apt_config["code"],
                "redirect": f"/en-us/country/{apt_config['country']}/vac/{apt_config['code']}"
            }
            full_login_url = f"{login_url}?{urlencode(params)}"
            
            self._log(f"Navigating: {full_login_url}")
            self.page.get(full_login_url)
            
            # --- Cloudflare 过盾 ---
            cf = CloudflareBypasser(self.page, log=self.config.debug)
            if not cf.bypass(max_retry=15):
                raise BizLogicError("Cloudflare bypass timeout")

            # --- 登录页面检查 ---
            if not self.page.ele('#email-input-field'):
                self._log("Reloading Login Page...")
                self.page.get(full_login_url)
                if not self.page.wait.ele_displayed('#email-input-field', timeout=15):
                    self._save_screenshot("login_load_fail")
                    raise BizLogicError("Login form not loaded")

            # --- JS 注入登录 ---
            g_token = ""
            if self.page.ele('.g-recaptcha') or self.page.ele('xpath://iframe[contains(@src, "recaptcha")]'):
                self._log("Solving ReCaptcha...")
                rc_params = {
                    "type": "ReCaptchaV2TaskProxyLess", "page": self.page.url,
                    "siteKey": "6LcDpXcfAAAAAM7wOEsF_38DNsL20tTvPTKxpyn0", 
                    "apiToken": self.free_config.get("capsolver_key", "")
                }
                g_token = self._solve_recaptcha(rc_params)

            username = self.config.account.username
            password = self.config.account.password
            
            # 使用 JS 直接操作 DOM 并 click，让浏览器处理 302
            js_login = f"""
            var u = document.getElementById('email-input-field');
            if(u) {{ u.value = "{username}"; u.dispatchEvent(new Event('input', {{bubbles:true}})); }}
            
            var p = document.getElementById('password-input-field');
            if(p) {{ p.value = "{password}"; p.dispatchEvent(new Event('input', {{bubbles:true}})); }}
            
            var g = document.getElementById('g-recaptcha-response');
            if(g) {{ g.value = "{g_token}"; }}
            
            var btn = document.getElementById('btn-login');
            if(btn) {{ btn.click(); return true; }} else {{ return false; }}
            """
            
            self._log("Submitting Login via JS...")
            if not self.page.run_js(js_login): raise BizLogicError("Login button missing")

            # --- 等待跳转 ---
            self._log("Waiting for redirect...")
            self.page.wait.url_change('login-actions', exclude=True, timeout=45)
            
            # 检查是否失败
            if "login-actions" in self.page.url or "auth" in self.page.url:
                err = "Unknown Login Error"
                if "Invalid username" in self.page.html: err = "Invalid Credentials"
                self._save_screenshot("login_submit_fail")
                raise BizLogicError(f"Login Failed: {err}")

            # --- 提取 Dashboard 信息 ---
            self._log("Waiting for dashboard...")
            self.page.wait.load_start()
            time.sleep(5)
            
            html = self.page.html
            self._check_page_is_session_expired_or_invalid("My travel group", html)
            groups = self._parse_travel_groups(html)
            
            target_city = apt_config['city'].lower()
            for g in groups:
                if g['location'].lower() == target_city:
                    self.travel_group = g
                    break
            
            if not self.travel_group:
                self._save_screenshot("group_not_found")
                raise NotFoundError(f"Group not found for {target_city}")
            
            self.session_create_time = time.time()
            self._log(f"Session Ready. Group: {self.travel_group['group_number']}")

        except Exception as e:
            self._log(f"Session Create Error: {e}")
            self.cleanup()
            raise e

    def query(self, apt_type: AppointmentType) -> VSQueryResult:
        res = VSQueryResult()
        res.success = False
        apt_config = self.free_config.get('apt_config', {})
        group_num = self.travel_group['group_number']
        interest_month = self.free_config.get("interest_month", time.strftime("%m-%Y"))
        
        url = f'https://visas-fr.tlscontact.com/en-us/{group_num}/workflow/appointment-booking'
        params = {
            'location': apt_config["code"],
            'month': interest_month,
        }
        
        # DrissionPage 自动处理 Cloudflare，直接 fetch 即可
        try:
            resp = self._perform_request("GET", url, params=params, retry_count=1)
        except Exception as e:
            self._log(f"Query request failed: {e}")
            raise e

        self._check_page_is_session_expired_or_invalid('Book your appointment', resp.text)

        # 解析 Slots
        all_slots = self._parse_appointment_slots(resp.text)

        target_labels = self.free_config.get("target_labels", ["", "pta"])
        # 根据配置过滤
        available = [s for s in all_slots if s.get("label") in target_labels]
        
        if available:
            res.success = True
            earliest_date = available[0]["date"]
            earliest_dt = datetime.strptime(earliest_date, "%Y-%m-%d")
            res.availability_status = AvailabilityStatus.Available
            res.earliest_date = earliest_dt
            date_map: dict[datetime, list[TimeSlot]] = {}
            for s in available:
                date_str = s["date"]
                dt = datetime.strptime(date_str, "%Y-%m-%d")
                date_map.setdefault(dt, []).append(
                    TimeSlot(time=s["time"], label=str(s.get("label", "")))
                )
            res.availability = [DateAvailability(date=d, times=slots) for d, slots in date_map.items()]
        else:
            res.success = False
            res.availability_status = AvailabilityStatus.NoneAvailable
        return res

    def book(self, slot_info: VSQueryResult, user_inputs: Dict = None) -> VSBookResult:
        res = VSBookResult()
        res.success = False
        
        apt_config = self.free_config.get('apt_config', {})
        group_num = self.travel_group['group_number']
        
        available_dates = [da.date for da in slot_info.availability]
        exp_start = user_inputs.get('expected_start_date', '')
        exp_end = user_inputs.get('expected_end_date', '')
        support_pta = user_inputs.get('support_pta', True)

        target_labels = ['']
        if support_pta:
            target_labels.append('pta')
        available_dates_str = [
            da.date.strftime("%Y-%m-%d")
            for da in slot_info.availability
        ]
        valid_dates = self._filter_dates(available_dates_str, exp_start, exp_end)
        if not valid_dates:
            raise NotFoundError(message="No dates match user constraints")
        
        selected_date = None
        selected_time = None
        selected_label = None
        
        for d in valid_dates:
            for da in slot_info.availability:
                if da.date == d:
                    for t in da.times:
                        if t.label in target_labels:
                            selected_date = d
                            selected_time = t
                            selected_label = t.label
                            break
            if selected_date: break
            
        if not selected_date:
             raise NotFoundError(message="No suitable slot found")

        # 2. 解决 ReCaptcha V3 (Action: book)
        page_url = f'https://visas-fr.tlscontact.com/en-us/{group_num}/workflow/appointment-booking?location={apt_config["code"]}&month={selected_date[:7]}'
        
        api_token = self.free_config.get("capsolver_key", "")
        rc_params = {
            "type": "ReCaptchaV3Task",
            "page": page_url,
            "action": "book", 
            "siteKey": "6LcTpXcfAAAAAM3VojNhyV-F1z92ADJIvcSZ39Y9",
            "apiToken": api_token,
            "proxy": self._get_proxy_url() # ProxyLess
        }
        g_token = self._solve_recaptcha(rc_params)

        # 3. 构造 Next.js Payload
        # 注意：在 JS 中构造 FormData 比在 Python 中拼 Multipart 更容易且不易出错
        ACTION_ID = "60d0616946df1fc4e7c094ca6a7a04f134d0be3d53"
        url = f'https://visas-fr.tlscontact.com/en-us/{group_num}/workflow/appointment-booking'
        
        # State Tree 字符串
        router_state = '%5B%22%22%2C%7B%22children%22%3A%5B%5B%22lang%22%2C%22en-us%22%2C%22d%22%5D%2C%7B%22children%22%3A%5B%5B%22groupId%22%2C%22'+str(group_num)+'%22%2C%22d%22%5D%2C%7B%22children%22%3A%5B%22workflow%22%2C%7B%22children%22%3A%5B%22appointment-booking%22%2C%7B%22children%22%3A%5B%22__PAGE__%22%2C%7B%7D%2Cnull%2Cnull%5D%7D%2Cnull%2Cnull%2Ctrue%5D%7D%2Cnull%2Cnull%5D%7D%2Cnull%2Cnull%5D%7D%2Cnull%2Cnull%2Ctrue%5D%7D%2Cnull%2Cnull%5D'

        # 构造 JS 代码执行 fetch
        # 使用 FormData 对象来处理 multipart
        js_script = f"""
        const url = "{url}";
        const formData = new FormData();
        
        formData.append('1_formGroupId', '{group_num}');
        formData.append('1_lang', 'en-us');
        formData.append('1_process', 'APPOINTMENT');
        formData.append('1_location', '{apt_config["code"]}');
        formData.append('1_date', '{selected_date}');
        formData.append('1_time', '{selected_time.time}');
        formData.append('1_appointmentLabel', '{selected_label}');
        formData.append('1_captcha_token', '{g_token}');
        formData.append('0', '[{{"status":"IDLE"}},"$K1"]');
        
        const headers = {{
            'Next-Action': '{ACTION_ID}',
            'Next-Router-State-Tree': decodeURIComponent('{router_state}'),
            'Accept': 'text/x-component'
        }};
        
        return fetch(url, {{
            method: 'POST',
            headers: headers,
            body: formData
        }}).then(async response => {{
            const text = await response.text();
            const headers = {{}};
            response.headers.forEach((value, key) => headers[key] = value);
            return {{
                status: response.status,
                body: text,
                headers: headers,
                url: response.url
            }};
        }}).catch(err => {{
            return {{ status: 0, body: err.toString(), headers: {{}}, url: url }};
        }});
        """
        
        self._log("Submitting booking request via JS Fetch...")
        res_dict = self.page.run_js(js_script)
        resp = BrowserResponse(res_dict)

        # 4. 结果判定
        # Next.js Server Action 重定向通常是 303，但 fetch 可能会自动跟随
        # 如果 fetch 跟随了，url 会变；如果没跟随（Redirect mode: manual），status 是 303
        
        if resp.status_code == 303 or (resp.status_code == 200 and "appointment-confirmation" in resp.url):
             self._log(f"Booking Success! URL: {resp.url}")
             res.success = True
             res.book_date = selected_date
             res.book_time = selected_time
             return res

        if resp.status_code == 200:
            if "APPOINTMENT_LIMIT_REACHED" in resp.text:
                 self._log("Failed: Appointment Limit Reached")
            elif "Invalid captcha" in resp.text:
                 self._log("Failed: Invalid Captcha")
            else:
                 self._log(f"Booking Failed (Unknown 200): {resp.text[:200]}")
        else:
            self._log(f"Booking Failed. Status: {resp.status_code}")

        return res

    # --- 辅助方法 ---
    
    def _get_proxy_url(self):
        # 构造代理
        proxy_url = ""
        if self.config.proxy.ip:
            s = self.config.proxy
            if s.username:
                proxy_url = f"{s.scheme}://{s.username}:{s.password}@{s.ip}:{s.port}"
            else:
                proxy_url = f"{s.scheme}://{s.ip}:{s.port}"
        return proxy_url

    def _perform_request(self, method, url, headers=None, data=None, json_data=None, params=None, retry_count=0):
        """
        在浏览器上下文中注入 JS 执行 Fetch
        """
        if not self.page:
            raise BizLogicError("Browser not initialized")

        if params:
            from urllib.parse import urlencode
            if '?' in url:
                url += '&' + urlencode(params)
            else:
                url += '?' + urlencode(params)

        fetch_options = {
            "method": method.upper(),
            "headers": headers or {},
            "credentials": "include"
        }

        # Body 处理
        if json_data:
            fetch_options['body'] = json.dumps(json_data)
            fetch_options['headers']['Content-Type'] = 'application/json'
        elif data:
             if isinstance(data, dict):
                from urllib.parse import urlencode
                fetch_options['body'] = urlencode(data)
                fetch_options['headers']['Content-Type'] = 'application/x-www-form-urlencoded'
             else:
                 fetch_options['body'] = data

        js_script = f"""
        const url = "{url}";
        const options = {json.dumps(fetch_options)};
        
        return fetch(url, options)
            .then(async response => {{
                const text = await response.text();
                const headers = {{}};
                response.headers.forEach((value, key) => headers[key] = value);
                
                return {{
                    status: response.status,
                    body: text,
                    headers: headers,
                    url: response.url
                }};
            }})
            .catch(error => {{
                return {{
                    status: 0,
                    body: error.toString(),
                    headers: {{}},
                    url: url
                }};
            }});
        """
        
        res_dict = self.page.run_js(js_script, timeout=30)
        resp = BrowserResponse(res_dict)
        
        if resp.status_code == 200:
            return resp
        elif resp.status_code == 401:
            self.is_healthy = False
            raise SessionExpiredOrInvalidError()
        elif resp.status_code == 403:
            # [关键修改] 遇到 403 Forbidden，尝试绕盾并重试
            # 最多重试 2 次
            if retry_count < 2:
                self._log(f"HTTP 403 Detected. Cloudflare session expired? Attempting refresh (Try {retry_count+1}/2)...")
                
                # 尝试刷新盾
                if self._refresh_firewall_session():
                    self._log("Firewall session refreshed. Retrying request...")
                    # 递归重试
                    return self._perform_request(method, url, headers, data, json_data, params, retry_count+1)
                else:
                    self._log("Failed to refresh firewall session.")
            
            # 如果重试失败，抛出异常
            raise PermissionDeniedError(f"HTTP 403: {resp.text[:100]}")
        elif resp.status_code == 429:
            self.is_healthy = False
            raise RateLimiteddError()
        else:
             # 如果是 0，可能是 fetch 报错
            if resp.status_code == 0:
                 raise BizLogicError(f"Network Error: {resp.text}")
            # TLS 业务错误
            raise BizLogicError(message=f"HTTP Error {resp.status_code}: {resp.text[:100]}")
        

    def _refresh_firewall_session(self) -> bool:
        """
        主动刷新页面以触发 Cloudflare 挑战并尝试通过
        """
        try:
            # 1. 刷新当前页面 (通常 Dashboard 页)
            # 这会强制浏览器重新进行 HTTP 请求，从而触发 Cloudflare 拦截页
            self._log("Refreshing page to trigger Cloudflare...")
            self.page.refresh()
            
            # 2. 调用 CloudflareBypasser
            cf = CloudflareBypasser(self.page, log=self.config.debug)
            
            # 3. 尝试过盾 (尝试次数稍多一点，因为此时可能网络不稳定)
            success = cf.bypass(max_retry=10)
            
            if success:
                # 再次确认页面是否正常加载 (非 403 页面)
                title = self.page.title.lower()
                if "access denied" in title:
                    return False
                
                # 等待 DOM 稍微稳定
                time.sleep(2)
                return True
            
            return False
        except Exception as e:
            self._log(f"Error during firewall refresh: {e}")
            return False

    def _solve_recaptcha(self, params) -> str:
        """调用 VSCloudApi 解决 ReCaptcha"""
        key = params.get("apiToken")
        if not key: raise NotFoundError("Api-token required")
        
        submit_url = "https://api.capsolver.com/createTask"
        task = {
            "type": params.get("type"),
            "websiteURL": params.get("page"),
            "websiteKey": params.get("siteKey"),
        }
        if params.get("action"):
            task["pageAction"] = params.get("action")
            
        if params.get("proxy"):
            p = urlparse(params.get("proxy"))
            task["proxyType"] = p.scheme
            task["proxyAddress"] = p.hostname
            task["proxyPort"] = p.port
            if p.username:
                task["proxyLogin"] = p.username
                task["proxyPassword"] = p.password
            
        # 注意：使用 DrissionPage 后，通常是 ProxyLess 模式
        # 除非你想让 Capsolver 也用同样的代理（通常不需要，除非风控极严）
        
        payload = {"clientKey": key, "task": task}
        import requests as req # 局部引用，避免混淆
        r = req.post(submit_url, json=payload, timeout=20)
        if r.status_code != 200:
            raise BizLogicError(message="Failed to submit capsolver task")
        
        task_id = r.json().get("taskId")
        for _ in range(20):
            r = req.post("https://api.capsolver.com/getTaskResult", json={"clientKey": key, "taskId": task_id}, timeout=20)
            if r.status_code == 200:
                d = r.json()
                if d.get("status") == "ready":
                    return d["solution"]["gRecaptchaResponse"]
            time.sleep(3)
        raise BizLogicError(message="Capsolver task timeout")

    def _parse_travel_groups(self, html: str) -> List[Dict]:
        groups = []
        js_pattern = r'\\"travelGroups\\":\s*(\[.*?\]),\\"availableCountriesToCreateGroups'
        js_match = re.search(js_pattern, html, re.DOTALL)
        if js_match:
            json_str = js_match.group(1).replace(r'\"', '"')
            data = json.loads(json_str)
            for g in data:
                groups.append({
                    'group_name': g.get('groupName'),
                    'group_number': g.get('formGroupId'),
                    'location': g.get('vacName')
                })
        else:
            self._log('Parsed travel group page, but not found travelGroups')
        return groups

    def _parse_appointment_slots(self, html: str) -> List[Dict]:
        slots = []
        pattern = r'"availableAppointments\\":\s*(\[.*\]),\\"showFlexiAppointment'
        match = re.search(pattern, html, re.DOTALL)
        
        if match:
            json_str = match.group(1).replace(r'\"', '"')
            data = json.loads(json_str)
            for day in data:
                d_str = day.get('day')
                for s in day.get('slots', []):
                    labels = s.get('labels', [])
                    lbl = ""
                    # 简化逻辑：TLS label 列表
                    if 'pta' in labels: lbl = 'pta'
                    elif 'ptaw' in labels: lbl = 'ptaw'
                    elif '' in labels or not labels: lbl = ''
                    
                    slots.append({
                        'date': d_str,
                        'time': s.get('time'),
                        'label': lbl
                    })
        return slots
  
    def _check_page_is_session_expired_or_invalid(self, keyword, html: str) -> bool:
        if not html:
            self.is_healthy = False
            raise SessionExpiredOrInvalidError()
        
        # 将 html 转小写检查
        html_lower = html.lower()
        if keyword.lower() not in html_lower: 
            if 'redirected automatically' in html_lower:
                self.is_healthy = False
                raise SessionExpiredOrInvalidError("Redirected automatically")
            if 'login' in html_lower and 'password' in html_lower:
                self.is_healthy = False
                raise SessionExpiredOrInvalidError("Redirected to login")
            if 'session expired' in html_lower:
                self.is_healthy = False
                raise SessionExpiredOrInvalidError("Session expired")
            
    def _filter_dates(self, dates: List[str], start_str: str, end_str: str) -> List[str]:
        if not start_str or not end_str:
            return dates
        valid_dates = []
        s_date = datetime.strptime(start_str[:10], "%Y-%m-%d")
        e_date = datetime.strptime(end_str[:10], "%Y-%m-%d")
        for date_str in dates:
            curr_date = datetime.strptime(date_str, "%Y-%m-%d")
            if s_date <= curr_date <= e_date:
                valid_dates.append(date_str)
        random.shuffle(valid_dates)
        return valid_dates
    
    # --- 资源清理核心方法 ---
    def cleanup(self):
        """
        销毁浏览器并彻底删除临时文件
        """
        # 1. 关闭浏览器
        if self.page:
            try:
                self.page.quit() # 这会关闭 Chrome 进程
            except Exception:
                pass # 忽略已关闭的错误
            self.page = None
        
        # 2. 删除文件
        # 注意：Chrome 关闭后可能需要几百毫秒释放文件锁，稍微等待
        if os.path.exists(self.root_workspace):
            for _ in range(3):
                try:
                    time.sleep(0.2)
                    shutil.rmtree(self.root_workspace, ignore_errors=True)
                    break
                except Exception as e:
                    # 如果删除失败(通常是Windows文件占用)，重试
                    self._log(f"Cleanup retry: {e}")
                    time.sleep(0.5)
            
            # 如果依然存在，打印警告（虽然 ignore_errors=True 会掩盖报错，但可以 check exists）
            if os.path.exists(self.root_workspace):
                 self._log(f"[WARN] Failed to fully remove workspace: {self.root_workspace}")
        # 3. [新增] 关闭代理隧道
        if self.tunnel:
            try: self.tunnel.stop()
            except: pass
            self.tunnel = None
        
    def __del__(self):
        """
        析构函数：当对象被垃圾回收时自动调用
        """
        self.cleanup()