Upload 3 files

Dockerfile

@@ -1,4 +1,4 @@
-FROM python:3.10-slim-bullseye as builder
+FROM python:3.11-slim-bullseye as builder
 
 # Build dummy packages to skip installing them and their dependencies
 RUN apt-get update \
@@ -12,14 +12,13 @@ RUN apt-get update \
  && equivs-build adwaita-icon-theme \
  && mv adwaita-icon-theme_*.deb /adwaita-icon-theme.deb
 
-FROM python:3.10-slim-bullseye
+FROM python:3.11-slim-bullseye
 COPY --from=builder /*.deb /
 WORKDIR /app
-# RUN echo "deb http://deb.debian.org/debian/ unstable main contrib non-free" >> /etc/apt/sources.list
+RUN echo "deb http://deb.debian.org/debian/ unstable main contrib non-free" >> /etc/apt/sources.list
 RUN apt update
-RUN apt upgrade -y
 RUN apt install -y python3 python3-pip libgl1-mesa-glx wget libglib2.0-dev sudo libpci-dev psmisc
-RUN pip install playwright hcaptcha_challenger requests loguru flask Flask-Limiter
+RUN pip install -U playwright hcaptcha_challenger requests loguru flask Flask-Limiter
 RUN dpkg -i /libgl1-mesa-dri.deb \
  && dpkg -i /adwaita-icon-theme.deb \
  # Install dependencies
@@ -34,14 +33,14 @@ RUN dpkg -i /libgl1-mesa-dri.deb \
  && chown -R foxer:foxer .
 
 RUN rm -rf /root/.cache
-RUN chmod 777 -R "/usr/local/lib/python3.10/"
+RUN chmod 777 -R "/usr/local/lib/python3.11/"
 RUN chmod 777 -R "/app/"
 RUN playwright install firefox --with-deps
 
 USER foxer
 RUN playwright install firefox
-COPY hcaptcha_solver.py .
-COPY token .
+COPY back.py .
+
 COPY api.py .
 EXPOSE 8081
 

api.py

@@ -1,8 +1,5 @@
-import hashlib
 import os
 import asyncio
-import uuid
-import shutil
 from flask import Flask, jsonify, request, logging as flog
 from flask_limiter.util import get_remote_address
 import hcaptcha_solver
@@ -21,12 +18,6 @@ def get_ipaddr():
 handler = flog.default_handler
 
 
-def generate_uuid():
- unique_identifier = str(uuid.uuid4())
- hashed_string = hashlib.sha256(unique_identifier.encode()).hexdigest()
- return hashed_string
-
-
 def get_token():
  default_token = "init_token"
  if os.path.exists("token"):
@@ -80,13 +71,7 @@ def solver_captcha():
  data = request.get_json(force=True, silent=True)
  if not check_request(require_data, data):
  return jsonify(msg="Unauthorized Request"), 403
- dir_path=generate_uuid()
- hcaptcha_solver.solver.install(upgrade=True)
- resp=asyncio.run(hcaptcha_solver.bytedance(data["host"], data["site_key"], dir_path))
- if os.path.exists(dir_path):
- shutil.rmtree(dir_path)
- if os.path.exists("tmp_dir"):
- shutil.rmtree("tmp_dir")
+ resp = asyncio.run(hcaptcha_solver.bytedance(data["host"], data["site_key"]))
  return resp
 
 

hcaptcha_solver.py

@@ -1,42 +1,9 @@
-from pathlib import Path
-
-import traceback
+from __future__ import annotations
 from loguru import logger
-from playwright.async_api import BrowserContext as ASyncContext, async_playwright
-import hcaptcha_challenger as solver
+from playwright.async_api import BrowserContext as ASyncContext, async_playwright, Page
+from hcaptcha_challenger import ModelHub, install
 from hcaptcha_challenger.agents import AgentT, Malenia
 
-# Init local-side of the ModelHub
-solver.install(upgrade=True, clip=True)
-
-# Save dataset to current working directory
-tmp_dir = Path(__file__).parent.joinpath("tmp_dir")
-
-
-@logger.catch
-async def hit_challenge(context: ASyncContext, host, sitekey, user_data_dir, times: int = 8):
- await context.route('**/*', lambda route, request: route_continuation(route, request, host, sitekey))
- page = context.pages[0]
- agent = AgentT.from_page(page=page, tmp_dir=tmp_dir, self_supervised=True)
- await page.goto(f"https://{host}")
-
- await agent.handle_checkbox()
-
- for pth in range(1, times):
- result = await agent()
- print(f">> {pth} - Challenge Result: {result}")
- match result:
- case agent.status.CHALLENGE_BACKCALL:
- 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:
- rqdata = agent.cr.__dict__
- await context.close()
- return rqdata["generated_pass_UUID"]
- case default:
- print(result)
-
 
 async def route_continuation(route, request, host, sitekey):
  # 检查请求的URL，只拦截特定网站的请求
@@ -105,21 +72,103 @@ async def route_continuation(route, request, host, sitekey):
  await route.continue_()
 
 
-async def bytedance(host, sitekey, user_data_dirs):
- print(user_data_dirs)
- # playwright install firefox --with-deps
- try:
- async with async_playwright() as p:
- context = await p.firefox.launch_persistent_context(
- user_data_dir=Path(__file__).parent.joinpath(user_data_dirs),
- headless=True,
- locale="en-US"
- )
- await Malenia.apply_stealth(context)
- token = await hit_challenge(context, host, sitekey, Path(__file__).parent.joinpath(user_data_dirs))
- return token
- except Exception as e:
- await context.close()
- traceback.print_exc()
- print(e)
- return traceback.format_exc()
+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=False)
+ context = await browser.new_context(
+ locale="en-US"
+ )
+ await Malenia.apply_stealth(context)
+
+ await hit_challenge(context, host, sitekey)