from __future__ import annotations from loguru import logger from playwright.async_api import BrowserContext as ASyncContext, async_playwright, Page from hcaptcha_challenger import ModelHub, install from hcaptcha_challenger.agents import AgentT, Malenia async def route_continuation(route, request, host, sitekey): # 检查请求的URL，只拦截特定网站的请求 if request.url == f"https://{host}/": print("start to solve") # 修改DNS解析结果 await route.fulfill(status=200, body=""" hCAPTCHA 演示

""".replace("%%%%%%%%%%%", sitekey)) else: # 对于其他网站的请求，不做修改 await route.continue_() def patch_modelhub(modelhub: ModelHub): """ 1. Patching clip_candidates allows you to handle all image classification tasks in self-supervised mode. 2. You need to inject hints for all categories that appear in a batch of images 3. The ObjectsYaml in the GitHub repository are updated regularly, but if you find something new, you can imitate the following and patch some hints. 4. Note that this should be a regularly changing table. If after a while certain labels no longer appear, you should not fill them in clip_candidates 5. Please note that you only need a moderate number of candidates prompts, too many prompts will increase the computational complexity :param modelhub: :return: """ modelhub.clip_candidates.update( { "the largest animal in real life": [ "parrot", "bee", "ladybug", "frog", "crab", "bat", "butterfly", "dragonfly", "giraffe", "tiger", "owl", "duck" ] } ) def prelude(page: Page) -> AgentT: # 1. You need to deploy sub-thread tasks and actively run `install(upgrade=True)` every 20 minutes # 2. You need to make sure to run `install(upgrade=True, clip=True)` before each instantiation install(upgrade=True, clip=True) modelhub = ModelHub.from_github_repo() modelhub.parse_objects() # Make arbitrary pre-modifications to modelhub, which is very useful for CLIP models patch_modelhub(modelhub) agent = AgentT.from_page( # page, the control handle of the Playwright Page page=page, # modelhub, Register modelhub externally, and the agent can patch custom configurations modelhub=modelhub, # clip, Enable CLIP zero-shot image classification method clip=True, ) return agent async def hit_challenge(context: ASyncContext, host, sitekey, times: int = 8): await context.route('**/*', lambda route, request: route_continuation(route, request, host, sitekey)) page = await context.new_page() agent = prelude(page) await page.goto(f"https://{host}") logger.info("startup sitelink", url=f"https://{host}") await agent.handle_checkbox() for pth in range(1, times): # Handle challenge result = await agent.execute() if not agent.qr: return # Post-processing match result: case agent.status.CHALLENGE_BACKCALL | agent.status.CHALLENGE_RETRY: logger.warning(f"retry", pth=pth, ash=agent.ash) await page.wait_for_timeout(500) fl = page.frame_locator(agent.HOOK_CHALLENGE) await fl.locator("//div[@class='refresh button']").click() case agent.status.CHALLENGE_SUCCESS: logger.success(f"task done", pth=pth, ash=agent.ash) rqdata = agent.cr.__dict__ await context.close() return rqdata["generated_pass_UUID"] async def bytedance(host, sitekey): async with async_playwright() as p: browser = await p.firefox.launch(headless=True) context = await browser.new_context( locale="en-US" ) await Malenia.apply_stealth(context) token = await hit_challenge(context, host, sitekey) return token