diff --git a/src/services/browser_captcha.py b/src/services/browser_captcha.py
index 69535da..922d521 100644
--- a/src/services/browser_captcha.py
+++ b/src/services/browser_captcha.py
@@ -199,9 +199,21 @@ def validate_browser_proxy_url(proxy_url: str) -> tuple[bool, str]:
return True, None
class TokenBrowser:
- """简化版浏览器：每次获取 token 时启动新浏览器，用完即关
-
- 每次都是新的随机 UA，避免长时间运行导致的各种问题
+ """持久化浏览器：复用浏览器进程和上下文，保持 cookie 连续性
+
+ 架构说明：
+ - 浏览器进程常驻，避免重复启动开销
+ - BrowserContext 每 50 次请求后轮换，获得新 UA/指纹
+ - 惰性初始化：首次调用 get_token() 时才启动浏览器
+ - 异常时自动重建浏览器进程
+
+ 与临时浏览器架构的区别：
+ - 旧架构：每次获取 token 都启动新浏览器（慢）
+ - 新架构：复用浏览器进程，仅在必要时轮换 context（快）
+
+ 线程安全：
+ - 并发控制由服务层的 semaphore 管理
+ - 多个 TokenBrowser 实例之间独立运行
"""
# UA 池
@@ -349,79 +361,160 @@ def __init__(self, token_id: int, user_data_dir: str, db=None):
self.token_id = token_id
self.user_data_dir = user_data_dir
self.db = db
- self._semaphore = asyncio.Semaphore(1) # 同时只能有一个任务
+ # 注意：移除了 self._semaphore，避免与 BrowserCaptchaService._token_semaphore 形成双重锁定
+ # 并发控制由服务层统一管理
self._solve_count = 0
self._error_count = 0
-
- async def _create_browser(self) -> tuple:
- """创建新浏览器实例（新 UA），返回 (playwright, browser, context)"""
- import random
-
- random_ua = random.choice(self.UA_LIST)
- base_w, base_h = random.choice(self.RESOLUTIONS)
- width, height = base_w, base_h - random.randint(0, 80)
- viewport = {"width": width, "height": height}
-
- playwright = await async_playwright().start()
- Path(self.user_data_dir).mkdir(parents=True, exist_ok=True)
-
- # 代理配置
- proxy_option = None
+ # 持久化浏览器状态
+ self._playwright = None
+ self._browser = None
+ self._context = None
+ self._request_count = 0
+ self._max_requests_per_context = 50 # 每 50 次请求轮换 context
+
+ async def _get_proxy_option(self) -> Optional[Dict[str, str]]:
+ """获取代理配置"""
try:
if self.db:
captcha_config = await self.db.get_captcha_config()
- raw_url = captcha_config.browser_proxy_enabled and captcha_config.browser_proxy_url
- if raw_url:
- proxy_option = parse_proxy_url(raw_url.strip())
- if proxy_option:
- debug_logger.log_info(f"[BrowserCaptcha] Token-{self.token_id} 使用代理: {proxy_option['server']}")
- except: pass
-
- try:
- browser = await playwright.chromium.launch(
- headless=False,
- proxy=proxy_option,
- args=[
- '--disable-blink-features=AutomationControlled',
- '--no-sandbox',
- '--disable-dev-shm-usage',
- '--disable-setuid-sandbox',
- '--no-first-run',
- '--no-zygote',
- f'--window-size={width},{height}',
- '--disable-infobars',
- '--hide-scrollbars',
- ]
- )
- context = await browser.new_context(
- user_agent=random_ua,
- viewport=viewport,
- )
- return playwright, browser, context
+ if captcha_config: # 检查 None
+ raw_url = (captcha_config.browser_proxy_enabled and
+ captcha_config.browser_proxy_url)
+ if raw_url:
+ proxy_option = parse_proxy_url(raw_url.strip())
+ if proxy_option:
+ debug_logger.log_info(f"[BrowserCaptcha] Token-{self.token_id} 使用代理: {proxy_option['server']}")
+ return proxy_option
except Exception as e:
- debug_logger.log_error(f"[BrowserCaptcha] Token-{self.token_id} 启动浏览器失败: {type(e).__name__}: {str(e)[:200]}")
- # 确保清理已创建的对象
+ debug_logger.log_warning(f"[BrowserCaptcha] Token-{self.token_id} 获取代理配置失败: {type(e).__name__}: {str(e)[:100]}")
+ return None
+
+ async def _ensure_browser(self):
+ """确保浏览器进程已启动（惰性初始化）"""
+ if self._browser and self._playwright:
+ return
+ # 清理旧的残留
+ await self._close_browser_only()
+
+ self._playwright = await async_playwright().start()
+ Path(self.user_data_dir).mkdir(parents=True, exist_ok=True)
+
+ proxy_option = await self._get_proxy_option()
+ base_w, base_h = random.choice(self.RESOLUTIONS)
+ width, height = base_w, base_h - random.randint(0, 80)
+
+ self._browser = await self._playwright.chromium.launch(
+ headless=False,
+ proxy=proxy_option,
+ args=[
+ '--disable-blink-features=AutomationControlled',
+ '--no-sandbox',
+ '--disable-dev-shm-usage',
+ '--disable-setuid-sandbox',
+ '--no-first-run',
+ '--no-zygote',
+ f'--window-size={width},{height}',
+ '--disable-infobars',
+ '--hide-scrollbars',
+ ]
+ )
+ debug_logger.log_info(f"[BrowserCaptcha] Token-{self.token_id} 浏览器进程已启动")
+
+ async def _get_or_create_context(self):
+ """获取或创建浏览器上下文，到期后轮换"""
+ await self._ensure_browser()
+
+ # 检查是否需要轮换 context
+ if self._context and self._request_count < self._max_requests_per_context:
+ self._request_count += 1
+ return self._context
+
+ # 关闭旧 context
+ if self._context:
try:
- if playwright:
- await playwright.stop()
- except: pass
- raise
-
- async def _close_browser(self, playwright, browser, context):
- """关闭浏览器实例"""
+ await self._context.close()
+ except:
+ pass
+ debug_logger.log_info(f"[BrowserCaptcha] Token-{self.token_id} 轮换 context（已用 {self._request_count} 次）")
+
+ random_ua = random.choice(self.UA_LIST)
+ base_w, base_h = random.choice(self.RESOLUTIONS)
+ width, height = base_w, base_h - random.randint(0, 80)
+ viewport = {"width": width, "height": height}
+
+ self._context = await self._browser.new_context(
+ user_agent=random_ua,
+ viewport=viewport,
+ )
+ self._request_count = 1
+ return self._context
+
+ async def _close_browser_only(self):
+ """关闭浏览器进程和 context，供内部使用"""
try:
- if context:
- await context.close()
- except: pass
+ if self._context:
+ await self._context.close()
+ except Exception as e:
+ debug_logger.log_warning(f"[BrowserCaptcha] Token-{self.token_id} 关闭 context 失败: {type(e).__name__}: {str(e)[:100]}")
+ self._context = None
try:
- if browser:
- await browser.close()
- except: pass
+ if self._browser:
+ await self._browser.close()
+ except Exception as e:
+ debug_logger.log_warning(f"[BrowserCaptcha] Token-{self.token_id} 关闭 browser 失败: {type(e).__name__}: {str(e)[:100]}")
+ self._browser = None
try:
- if playwright:
- await playwright.stop()
- except: pass
+ if self._playwright:
+ await self._playwright.stop()
+ except Exception as e:
+ debug_logger.log_warning(f"[BrowserCaptcha] Token-{self.token_id} 停止 playwright 失败: {type(e).__name__}: {str(e)[:100]}")
+ self._playwright = None
+ self._request_count = 0
+
+ async def close(self):
+ """公开的关闭方法"""
+ await self._close_browser_only()
+ async def _simulate_human_behavior(self, page):
+ """模拟人类浏览行为，提升 reCAPTCHA v3 评分"""
+ try:
+ viewport = page.viewport_size or {"width": 1920, "height": 1080}
+ w, h = viewport["width"], viewport["height"]
+
+ # 初始延迟 - 模拟用户看到页面后的反应时间
+ await asyncio.sleep(random.uniform(0.5, 1.5))
+
+ # 随机鼠标移动 2-4 次
+ for _ in range(random.randint(2, 4)):
+ x = random.randint(100, w - 100)
+ y = random.randint(100, h - 100)
+ await page.mouse.move(x, y, steps=random.randint(3, 8))
+ await asyncio.sleep(random.uniform(0.1, 0.4))
+
+ # 随机滚动
+ await page.mouse.wheel(0, random.randint(50, 200))
+ await asyncio.sleep(random.uniform(0.2, 0.5))
+
+ # 尝试点击 textarea（模拟用户准备输入）
+ try:
+ textarea = page.locator("textarea").first
+ if await textarea.is_visible(timeout=500):
+ await textarea.click()
+ await asyncio.sleep(random.uniform(0.2, 0.5))
+ except Exception as e:
+ debug_logger.log_debug(f"[BrowserCaptcha] Token-{self.token_id} 点击 textarea 失败: {type(e).__name__}")
+
+ # 最后再移一下鼠标
+ await page.mouse.move(
+ random.randint(200, w - 200),
+ random.randint(200, h - 200),
+ steps=random.randint(3, 6)
+ )
+ await asyncio.sleep(random.uniform(0.3, 0.8))
+ except Exception as e:
+ # 行为模拟失败不影响主流程，但记录日志便于调试
+ debug_logger.log_debug(f"[BrowserCaptcha] Token-{self.token_id} 行为模拟失败: {type(e).__name__}: {str(e)[:100]}")
+
async def _execute_captcha(self, context, project_id: str, website_key: str, action: str) -> Optional[str]:
"""在给定 context 中执行打码逻辑"""
page = None
@@ -433,7 +526,63 @@ async def _execute_captcha(self, context, project_id: str, website_key: str, act
async def handle_route(route):
if route.request.url.rstrip('/') == page_url.rstrip('/'):
- html = f""""""
+ html = f"""
+
+
+
+
+ Flow - Google Labs
+
+
+
+
+
+
+
+
+
My Project
+
+
+
+
+
+
+
+
+