Spaces:
Paused
Paused
File size: 41,026 Bytes
e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 e0ab505 4279593 c8abe84 4279593 85a4a41 4279593 85a4a41 4279593 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 |
# from crawl4ai import AsyncWebCrawler
# from urllib.parse import urlparse
import aiohttp
import asyncio
# from asyncio.exceptions import TimeoutError as async_timeout
from fast_async import make_async
from bs4 import BeautifulSoup, NavigableString
# import secrets
# from datetime import datetime
# import random
import os
import re
import uuid
from typing import List, Dict, Optional #, Tuple
from io import BytesIO
import PyPDF2
from fake_useragent import FakeUserAgent
from htmlrag import clean_html, build_block_tree, EmbedHTMLPruner, BM25HTMLPruner
from transformers import AutoTokenizer, AutoConfig
import torch
import time
# class Crawler:
# def __init__(self, user_dir=None, rate_limit=1, headless=True, verbose=False):
# self.session_pool = {} # Track active sessions
# self.verbose = verbose
# self.rate_limit = rate_limit
# self.user_dir = user_dir
# self.headless = headless
# self.crawler = AsyncWebCrawler(
# context_options={"userDataDir": self.user_dir},
# headless=self.headless,
# verbose=self.verbose
# )
# # Browser context management
# self._browser_contexts = {}
# self._context_locks = {}
# async def get_browser_context(self, session_id):
# """Get or create a browser context with proper locking"""
# if session_id not in self._context_locks:
# self._context_locks[session_id] = asyncio.Lock()
# async with self._context_locks[session_id]:
# if session_id not in self._browser_contexts:
# context = await self.crawler.new_context()
# self._browser_contexts[session_id] = context
# return self._browser_contexts[session_id]
# async def cleanup_browser_context(self, session_id):
# """Safely cleanup browser context"""
# if session_id in self._context_locks:
# async with self._context_locks[session_id]:
# if session_id in self._browser_contexts:
# try:
# await asyncio.shield(
# self._browser_contexts[session_id].close()
# )
# except Exception as e:
# print(f"Error cleaning up browser context: {e}")
# finally:
# del self._browser_contexts[session_id]
# def create_session(self):
# """Create a new session with secure ID"""
# session_id = secrets.token_urlsafe(32) # Secure session ID
# self.session_pool[session_id] = {
# 'created_at': datetime.now(),
# 'last_used': datetime.now(),
# 'requests_count': 0
# }
# return session_id
# def rotate_session(self, session_id):
# """Implement session rotation logic"""
# if self.session_pool[session_id]['requests_count'] > 100:
# self.cleanup_session(session_id)
# return self.create_session()
# return session_id
# def is_dynamic_page(self, html_content: str) -> Tuple[bool, Optional[str]]:
# """Analyzes HTML content to determine if a webpage is dynamically loaded"""
# def _check_structural_indicators(soup: BeautifulSoup) -> Dict[str, int]:
# """Check structural indicators of dynamic content loading."""
# scores = {
# 'empty_containers': 0,
# 'repeated_structures': 0,
# 'api_endpoints': 0,
# 'state_management': 0
# }
# # 1. Check for empty content containers
# main_containers = soup.find_all(['main', 'div', 'section'],
# class_=lambda x: x and any(term in str(x).lower()
# for term in ['content', 'main', 'feed', 'list', 'container']))
# for container in main_containers:
# # Check if container is empty or has minimal content
# if len(container.find_all()) < 3:
# scores['empty_containers'] += 1
# # Check for repeated similar structures (common in dynamic lists)
# children = container.find_all(recursive=False)
# if children:
# first_child_class = children[0].get('class', [])
# similar_siblings = [c for c in children[1:]
# if c.get('class', []) == first_child_class]
# if len(similar_siblings) > 0:
# scores['repeated_structures'] += 1
# # 2. Check for API endpoints in scripts
# scripts = soup.find_all('script', {'src': True})
# api_patterns = ['/api/', '/graphql', '/rest/', '/v1/', '/v2/']
# for script in scripts:
# if any(pattern in script['src'] for pattern in api_patterns):
# scores['api_endpoints'] += 1
# # 3. Look for state management setup
# state_patterns = [
# r'window\.__INITIAL_STATE__',
# r'window\.__PRELOADED_STATE__',
# r'__REDUX_STATE__',
# r'__NUXT__',
# r'__NEXT_DATA__',
# r'window\.__data'
# ]
# inline_scripts = soup.find_all('script')
# for script in inline_scripts:
# if script.string:
# for pattern in state_patterns:
# if re.search(pattern, script.string):
# scores['state_management'] += 1
# return scores
# def _check_modern_framework_indicators(soup: BeautifulSoup) -> Dict[str, int]:
# """Check for indicators of modern web frameworks and dynamic loading patterns."""
# scores = {
# 'framework_roots': 0,
# 'hydration': 0,
# 'routing': 0
# }
# # 1. Framework-specific root elements
# framework_roots = {
# 'react': ['react-root', 'react-app', 'root', '__next'],
# 'angular': ['ng-version', 'ng-app'],
# 'vue': ['v-app', '#app', 'nuxt-app'],
# 'modern': ['app-root', 'application', 'spa-root']
# }
# for framework, identifiers in framework_roots.items():
# for id_value in identifiers:
# if (soup.find(attrs={'id': re.compile(id_value, re.I)}) or
# soup.find(attrs={'class': re.compile(id_value, re.I)}) or
# soup.find(attrs={'data-': re.compile(id_value, re.I)})):
# scores['framework_roots'] += 1
# # 2. Check for hydration indicators
# hydration_patterns = [
# r'hydrate',
# r'createRoot',
# r'reactive',
# r'observable'
# ]
# scripts = soup.find_all('script')
# for script in scripts:
# if script.string:
# for pattern in hydration_patterns:
# if re.search(pattern, script.string):
# scores['hydration'] += 1
# # 3. Check for dynamic routing setup
# router_patterns = [
# 'router-view',
# 'router-link',
# 'route-link',
# 'history.push',
# 'navigation'
# ]
# for pattern in router_patterns:
# if soup.find(class_=re.compile(pattern, re.I)) or \
# soup.find(id=re.compile(pattern, re.I)):
# scores['routing'] += 1
# return scores
# def _check_dynamic_loading_patterns(soup: BeautifulSoup) -> Dict[str, int]:
# """Check for various dynamic content loading patterns."""
# scores = {
# 'infinite_scroll': 0,
# 'load_more_buttons': 0,
# 'pagination': 0,
# 'lazy_loading': 0,
# 'loading_indicators': 0
# }
# # 1. Check for infinite scroll indicators
# scroll_indicators = [
# 'infinite-scroll',
# 'data-infinite',
# 'data-virtualized',
# 'virtual-scroll',
# 'scroll-container',
# 'scroll-viewport'
# ]
# for indicator in scroll_indicators:
# elements = soup.find_all(
# lambda tag: any(indicator.lower() in str(v).lower()
# for v in tag.attrs.values())
# )
# if elements:
# scores['infinite_scroll'] += len(elements)
# # 2. Check for load more buttons
# button_patterns = [
# r'load[_-]?more',
# r'show[_-]?more',
# r'view[_-]?more',
# r'see[_-]?more',
# r'more[_-]?posts',
# r'more[_-]?results'
# ]
# for pattern in button_patterns:
# elements = soup.find_all(
# ['button', 'a', 'div', 'span'],
# text=re.compile(pattern, re.I)
# )
# if elements:
# scores['load_more_buttons'] += len(elements)
# # 3. Check for pagination
# pagination_patterns = [
# 'pagination',
# 'page-numbers',
# 'page-nav',
# 'page-links'
# ]
# for pattern in pagination_patterns:
# elements = soup.find_all(class_=re.compile(pattern, re.I))
# if elements:
# scores['pagination'] += len(elements)
# # 4. Check for lazy loading
# lazy_patterns = ['lazy', 'data-src', 'data-lazy']
# for pattern in lazy_patterns:
# elements = soup.find_all(
# lambda tag: any(pattern.lower() in str(v).lower()
# for v in tag.attrs.values())
# )
# if elements:
# scores['lazy_loading'] += len(elements)
# # 5. Check for loading indicators
# loading_patterns = [
# 'loading',
# 'spinner',
# 'skeleton',
# 'placeholder',
# 'shimmer'
# ]
# for pattern in loading_patterns:
# elements = soup.find_all(class_=re.compile(pattern, re.I))
# if elements:
# scores['loading_indicators'] += len(elements)
# return scores
# def _evaluate_dynamic_indicators(
# structural: Dict[str, int],
# framework: Dict[str, int],
# loading: Dict[str, int]
# ) -> Tuple[bool, Optional[str]]:
# """Evaluate dynamic indicators and return JavaScript instructions."""
# methods = []
# js_snippets = []
# # Infinite Scroll
# if loading['infinite_scroll'] > 0:
# methods.append("scroll")
# js_snippets.append(
# """
# window.scrollTo(0, document.body.scrollHeight);
# await new Promise(resolve => setTimeout(resolve, 1000));
# """.strip().replace('\n', '')
# )
# # Load More Buttons
# if loading['load_more_buttons'] > 0:
# methods.append("button")
# js_snippets.append(
# """
# const button = Array.from(document.querySelectorAll('button, a, div, span')).find(
# el => /load[_-]?more|show[_-]?more/i.test(el.textContent)
# );
# if (button) {
# button.click();
# await new Promise(resolve => setTimeout(resolve, 1000));
# } else {
# console.warn("No 'Load More' button found.");
# }
# """.strip().replace('\n', '')
# )
# # Paginated Interfaces
# if loading.get('pagination', 0) > 0:
# methods.append("pagination")
# js_snippets.append(
# """
# const nextPage = document.querySelector('a[rel="next"], .pagination-next, .page-next');
# if (nextPage) {
# nextPage.click();
# await new Promise(resolve => setTimeout(resolve, 1000));
# } else {
# console.warn("No pagination link found.");
# }
# """.strip().replace('\n', '')
# )
# # Lazy Loading
# if loading.get('lazy_loading', 0) > 0:
# methods.append("lazy")
# js_snippets.append(
# """
# if (window.__INITIAL_STATE__ || window.__REDUX_STATE__ || window.__NUXT__ || window.__NEXT_DATA__) {
# console.log('Framework state detected. Consider monitoring network requests for further actions.');
# }
# """.strip().replace('\n', '')
# )
# # Framework and State Management Indicators
# if framework['framework_roots'] > 0 or structural['state_management'] > 0:
# methods.append("stateful")
# js_snippets.append(
# """
# if (window.__INITIAL_STATE__ || window.__REDUX_STATE__ || window.__NUXT__ || window.__NEXT_DATA__) {
# console.log('Detected stateful framework data loading.');
# }
# """.strip().replace('\n', '')
# )
# # API-Driven Content
# if structural['api_endpoints'] > 0:
# methods.append("api")
# js_snippets.append(
# """
# console.log('API requests detected. Use browser devtools to inspect network activity for specific endpoints.');
# """.strip().replace('\n', '')
# )
# # Aggregate and finalize
# if methods:
# js_code = "\n".join(js_snippets)
# return True, js_code
# return False, None
# # Main execution
# soup = BeautifulSoup(html_content, 'html.parser')
# # Run all checks
# structural_scores = _check_structural_indicators(soup)
# framework_scores = _check_modern_framework_indicators(soup)
# loading_scores = _check_dynamic_loading_patterns(soup)
# # Evaluate results
# return _evaluate_dynamic_indicators(structural_scores, framework_scores, loading_scores)
# async def crawl(
# self,
# url,
# depth=2,
# max_pages=5,
# session_id=None,
# human_simulation=True,
# rotate_user_agent=True,
# rotate_proxy=True,
# return_html=False
# ):
# if not session_id:
# session_id = self.create_session()
# session_id = self.rotate_session(session_id)
# # List of rotating user agents
# user_agents = [
# 'Chrome/115.0.0.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/115.0.0.0 Safari/537.36',
# 'Chrome/115.0.0.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/115.0.0.0 Safari/537.36',
# 'Chrome/115.0.0.0 (iPad; CPU OS 12_2 like Mac OS X) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/115.0.0.0 Safari/537.36',
# 'Chrome/115.0.0.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/115.0.0.0 Safari/537.36',
# 'Chrome/115.0.0.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/115.0.0.0 Safari/537.36'
# ]
# # List of rotating proxies
# proxies = [
# "http://50.62.183.123:80",
# "http://104.129.60.84:6516",
# "http://156.228.118.163:3128",
# "http://142.111.104.97:6107",
# "http://156.228.99.99:3128"
# ]
# try:
# async with self.crawler as crawler:
# # Rotate user agent and optimize headers for each attempt
# headers = {
# "User-Agent": random.choice(user_agents) if rotate_user_agent else user_agents[0],
# "Accept": "text/html,application/xhtml+xml,application/xml;q=0.9,image/avif,image/webp,*/*;q=0.8",
# "Accept-Language": "en-US,en;q=0.5",
# "Accept-Encoding": "gzip, deflate",
# "Connection": "keep-alive",
# "Upgrade-Insecure-Requests": "1",
# "Sec-Fetch-Dest": "document",
# "Sec-Fetch-Mode": "navigate",
# "Sec-Fetch-Site": "none",
# "Sec-Fetch-User": "?1",
# "Cache-Control": "max-age=0"
# }
# # Update crawler headers for rotation
# crawler.crawler_strategy.headers = headers
# if rotate_proxy:
# # Update crawler proxy for rotation
# crawler.crawler_strategy.proxy = random.choice(proxies)
# result_1 = await crawler.arun(
# session_id=session_id,
# url=url,
# magic=True if human_simulation else False,
# simulate_user=True if human_simulation else False,
# override_navigator=True if human_simulation else False,
# depth=depth,
# max_pages=max_pages,
# bypass_cache=True,
# remove_overlay_elements=True,
# delay_before_retrieve_html=1.0,
# verbose=self.verbose
# )
# # Update session metrics
# self.session_pool[session_id]['requests_count'] += 1
# self.session_pool[session_id]['last_used'] = datetime.now()
# if result_1.success:
# if hasattr(result_1, 'html'):
# success, js_code = self.is_dynamic_page(result_1.html)
# if success:
# async with crawler as crawler:
# # Update crawler headers for rotation
# crawler.crawler_strategy.headers = headers
# if rotate_proxy:
# # Update crawler proxy for rotation
# crawler.crawler_strategy.proxy = random.choice(proxies)
# print(f"Executing JS code: {js_code}")
# result_2 = await crawler.arun(
# session_id=session_id,
# url=url,
# magic=True if human_simulation else False,
# simulate_user=True if human_simulation else False,
# override_navigator=True if human_simulation else False,
# depth=depth,
# max_pages=max_pages,
# js_code=js_code,
# bypass_cache=True,
# remove_overlay_elements=True,
# delay_before_retrieve_html=1.0,
# verbose=self.verbose
# )
# if result_2.success:
# result = result_2
# else:
# result = result_1
# # Update session metrics
# self.session_pool[session_id]['requests_count'] += 1
# self.session_pool[session_id]['last_used'] = datetime.now()
# else:
# result = result_1
# if return_html and hasattr(result, 'html'):
# return result.html
# elif hasattr(result, 'fit_markdown'):
# return result.fit_markdown
# elif hasattr(result, 'markdown'):
# return self.extract_content(result.markdown)
# except Exception as e:
# print(f"Error crawling {url}: {str(e)}")
# return None
# async def crawl_with_retry(
# self,
# url,
# depth=2,
# max_pages=5,
# max_retries=3,
# backoff_factor=1,
# session_id=None,
# human_simulation=True,
# rotate_user_agent=True,
# rotate_proxy=True,
# return_html=False,
# timeout=10.0
# ):
# """Crawl with retry logic and anti-blocking measures"""
# async def attempt_crawl(attempt):
# try:
# async with async_timeout.timeout(timeout):
# context = await self.get_browser_context(session_id)
# return await self.crawl(
# context,
# url,
# depth,
# max_pages,
# session_id,
# human_simulation,
# rotate_user_agent,
# rotate_proxy,
# return_html
# )
# except asyncio.TimeoutError:
# print(f"Timeout on attempt {attempt} for {url}")
# raise
# except Exception as e:
# print(f"Error on attempt {attempt} for {url}: {e}")
# raise
# if not self.is_valid_url(url) and not self.is_html_url(url):
# print(f"Invalid URL: {url}")
# return f"No web results found for query: {url}"
# for attempt in range(max_retries):
# try:
# if attempt > 0:
# # Add delay between retries with exponential backoff
# delay = backoff_factor * (2 ** (attempt - 1))
# await asyncio.sleep(delay)
# return await attempt_crawl(attempt + 1)
# except Exception as e:
# if attempt == max_retries - 1:
# print(f"Max retries ({max_retries}) reached for {url}")
# return f"Failed to crawl after {max_retries} attempts: {url}"
# continue
# return f"No content found after {max_retries} attempts for: {url}"
# def extract_content(self, html_content):
# soup = BeautifulSoup(html_content, 'html.parser')
# for script in soup(["script", "style"]):
# script.decompose()
# text = soup.get_text()
# lines = (line.strip() for line in text.splitlines())
# chunks = (phrase.strip() for line in lines for phrase in line.split(" "))
# text = '\n'.join(chunk for chunk in chunks if chunk)
# return text
# def cleanup_session(self, session_id):
# """Clean up a session"""
# print(f"Cleaning up session {session_id}")
# if session_id in self.session_pool:
# self.crawler.crawler_strategy.kill_session(session_id)
# del self.session_pool[session_id]
# def cleanup_expired_sessions(self):
# """Regular cleanup of expired sessions using proper time calculation"""
# try:
# current_time = datetime.now()
# expired_sessions = []
# for sid, data in self.session_pool.items():
# # Calculate time difference in seconds
# time_diff = (current_time - data['last_used']).total_seconds()
# # Check if more than 1 hour (3600 seconds)
# if time_diff > 3600:
# expired_sessions.append(sid)
# # Cleanup expired sessions
# for session_id in expired_sessions:
# self.cleanup_session(session_id)
# except Exception as e:
# if self.verbose:
# print(f"Error during session cleanup: {str(e)}")
# @staticmethod
# def is_valid_url(url):
# try:
# result = urlparse(url)
# return all([result.scheme, result.netloc])
# except ValueError:
# return False
# @staticmethod
# def is_html_url(url):
# return url.endswith(".html") or url.endswith(".htm")
class CustomCrawler:
def __init__(
self,
embed_model: str = "HIT-TMG/KaLM-embedding-multilingual-mini-instruct-v1",
max_concurrent_requests: int = 10,
verbose: bool = True
):
print(f"🦀 Initializing the crawler") if verbose else None
time.sleep(1)
self.embed_model = embed_model
self.max_concurrent_requests = max_concurrent_requests
self.verbose = verbose
self.ua = FakeUserAgent()
self.semaphore = asyncio.Semaphore(self.max_concurrent_requests)
self.sessions = {}
# Intilizing HTML Pruners and Tokenizer
print(f"🔃 Loading HTML Pruners and Tokenizer with {self.embed_model}") if self.verbose else None
self.bm25_html_pruner = BM25HTMLPruner()
self.embed_html_pruner = EmbedHTMLPruner(
embed_model=self.embed_model,
local_inference=True
)
self.tokenizer = AutoTokenizer.from_pretrained(
self.embed_model,
use_fast=True,
trust_remote_code=True,
device="cuda" if torch.cuda.is_available() else "cpu"
)
# Get the model config and set the max context length for the model
print(f"🛠️ Getting model configuration for {self.embed_model}") if self.verbose else None
self.config = AutoConfig.from_pretrained(self.embed_model)
self.tokenizer.max_seq_length = self.config.max_position_embeddings
print(f"📏 Setting max context length to {self.tokenizer.max_seq_length}") if self.verbose else None
async def create_session(self):
session_id = str(uuid.uuid4())
timeout = aiohttp.ClientTimeout(total=600) # Set a 10-minute timeout
connector = aiohttp.TCPConnector(limit=self.max_concurrent_requests) # Connection pool
self.sessions[session_id] = aiohttp.ClientSession(timeout=timeout, connector=connector)
print(f"🔗 Created session: {session_id}") if self.verbose else None
return session_id
async def close_session(self, session_id):
session = self.sessions.pop(session_id, None)
if session:
await session.close()
print(f"🌂 Closed session: {session_id}") if self.verbose else None
async def cleanup_expired_sessions(self, expiration_time: int = 600): # Default 10 minutes
current_time = time.time()
expired_sessions = []
print("🔍 Checking for expired sessions") if self.verbose else None
for session_id, (session, creation_time) in self.sessions.items():
if current_time - creation_time > expiration_time:
expired_sessions.append(session_id)
for session_id in expired_sessions:
await self.close_session(session_id)
print(f"🗑️ Successfully cleaned up all expired sessions") if self.verbose else None
@make_async
def html_rag(
self,
query: str,
html: str,
max_context_length: int = 32000,
buffer: int = 2000
) -> str:
if not html:
raise Exception("No HTML contents provided.")
# Validate HTML structure
try:
BeautifulSoup(html, 'html.parser')
except Exception as e:
raise Exception(f"Invalid HTML content: {e}")
prompt_for_retrieval = \
"""Given a query, your task is to retrieve the most relevant passages that answers and/or is relevant to the query.
Query:"""
self.embed_html_pruner.query_instruction_for_retrieval = prompt_for_retrieval
print(f"🧹 Pruning HTML for query: {query}") if self.verbose else None
cleaned_html = clean_html(html)
block_tree, cleaned_html = build_block_tree(cleaned_html, max_node_words=10)
block_rankings = self.bm25_html_pruner.calculate_block_rankings(query, cleaned_html, block_tree)
max_context_window = max_context_length - buffer
pruned_html = self.embed_html_pruner.prune_HTML(
cleaned_html,
block_tree,
block_rankings,
self.tokenizer,
max_context_window
)
print(f"👍 Successfully pruned HTML for query: {query}") if self.verbose else None
return pruned_html
async def fetch_page_contents(
self,
urls: List[str],
query: Optional[str] = None,
session_id: Optional[str] = None,
max_attempts: int = 3,
delay: float = 1.0,
timeout: float = 10.0,
return_type: str = "markdown",
rotate_headers: bool = True,
) -> List[Optional[str]]:
async def fetch_single_page(url, proxies, session=None, query=query):
for attempt in range(max_attempts):
print(f"🔍 Attempt {attempt + 1}/{max_attempts}: Fetching content from {url}") if self.verbose else None
content = await self._fetch_page_contents(
url=url,
query=query,
timeout=timeout,
return_type=return_type,
rotate_headers=rotate_headers,
proxies=proxies,
session=session
)
if content:
print(f"✅ Successfully fetched content from {url}") if self.verbose else None
return content
else:
if max_attempts > 1:
print(f"🚫 Failed to fetch content from {url}. Retrying in {delay} seconds...") if self.verbose else None
await asyncio.sleep(delay)
print(f"🚫 Failed to fetch content from {url} after {max_attempts} attempts.") if self.verbose else None
return None
proxy_list = self.load_proxies() # Load proxies from environment variables
if proxy_list:
proxies = proxy_list
else:
proxies = None
if not urls:
raise Exception("No URLs provided!")
if return_type == "fit_markdown" and query is None:
raise Exception("Query must be provided when return_type is 'fit_markdown'!")
if session_id: # Use existing session if provided
if session_id not in self.sessions:
raise ValueError(f"Invalid session ID: {session_id}")
session = self.sessions[session_id]
tasks = [fetch_single_page(url, proxies, session) for url in urls] # Pass session to tasks
else: # No session handling if session_id is None
tasks = [fetch_single_page(url, proxies) for url in urls] # No session passed
results = await asyncio.gather(*tasks)
return [result for result in results if result is not None]
async def _fetch_page_contents(
self,
url: str,
query: Optional[str] = None,
timeout: float = 5.0,
return_type: str = "markdown",
rotate_headers: bool = True,
proxies: Optional[List[str]] = None,
session: Optional[aiohttp.ClientSession] = None
) -> Optional[str]:
async def get_content(response, return_type=return_type):
print(f"📄 Getting content from {url}") if self.verbose else None
if return_type == "html":
return await response.text()
response.raise_for_status()
content_type = response.headers.get('Content-Type', '').lower()
if 'application/pdf' in content_type:
content = await response.read()
text = self.extract_text_from_pdf(content)
return text
elif 'text/html' in content_type:
html_content = await response.text()
if return_type == "fit_markdown":
html_content = self.html_rag(query, html_content).wait()
soup = BeautifulSoup(html_content, "html.parser")
for script_or_style in soup(["script", "style"]):
script_or_style.decompose()
text = self.html_to_markdown(soup)
return text.strip()
else:
print(f"🚫 Unsupported content type {content_type} for URL {url}") if self.verbose else None
return None
headers = self.get_headers() if rotate_headers else {}
proxy = self.get_proxy(proxies) if proxies else None
# Total connection timeout
timeout_config = aiohttp.ClientTimeout(total=timeout)
try:
# Use provided session if available
if session:
async with session.get(url, proxy=proxy, timeout=timeout_config, headers=headers) as response:
return await get_content(response)
# Otherwise, create a new session for each request
else:
async with aiohttp.ClientSession() as new_session:
async with new_session.get(url, proxy=proxy, timeout=timeout_config, headers=headers) as response:
return await get_content(response)
except aiohttp.ClientError as e:
print(f"🚫 Request Exception for {url}: {e}") if self.verbose else None
return None
except asyncio.TimeoutError as e:
print(f"🚫 Timeout error for {url}") if self.verbose else None
return None
except Exception as e:
print(f"🚫 Unexpected error fetching {url}: {e}") if self.verbose else None
return None
def load_proxies(self) -> Optional[List[str]]:
# Get all environment variables
env_vars = dict(os.environ)
# Load proxies from environment variables
proxy_pattern = re.compile(r"PROXY_\d+")
proxies = [env_vars[key] for key in env_vars if proxy_pattern.match(key)]
if proxies:
print(f"🔌 Loaded {len(proxies)} proxies from environment variables") if self.verbose else None
return proxies
else:
return None
def get_proxy(self, proxies: List[str]) -> str:
if proxies: # Check if the proxies list is not empty
return next(iter(proxies))
return None # Or raise an exception, handle differently, etc.
def get_headers(self) -> Dict[str, str]:
return {'User-Agent': self.ua.random}
def extract_text_from_pdf(self, pdf_content: bytes) -> str:
try:
print(f"📕 Extracting text from PDF") if self.verbose else None
pdf_reader = PyPDF2.PdfReader(BytesIO(pdf_content))
text = ''
for page in pdf_reader.pages:
text += page.extract_text()
print(f"💪 Successfully extracted text from PDF") if self.verbose else None
return text
except Exception as e:
print(f"🚫 Error extracting text from PDF: {e}") if self.verbose else None
return ""
def html_to_markdown(self, soup):
markdown_text = ""
print(f"📟 Converting HTML to Markdown") if self.verbose else None
def process_element(element, indent=0):
nonlocal markdown_text
if isinstance(element, NavigableString):
text = str(element).strip()
if text:
markdown_text += text + " "
return
tag = element.name
if tag == "h1":
markdown_text += "# " + element.text.strip() + "\n\n"
elif tag == "h2":
markdown_text += "## " + element.text.strip() + "\n\n"
elif tag == "h3":
markdown_text += "### " + element.text.strip() + "\n\n"
elif tag == "h4":
markdown_text += "#### " + element.text.strip() + "\n\n"
elif tag == "h5":
markdown_text += "##### " + element.text.strip() + "\n\n"
elif tag == "h6":
markdown_text += "###### " + element.text.strip() + "\n\n"
elif tag == "p":
markdown_text += element.text.strip() + "\n\n"
elif tag == "br":
markdown_text += "\n"
elif tag == "ul":
for li in element.find_all("li", recursive=False):
markdown_text += " " * indent + "- "
process_element(li, indent + 1)
markdown_text += "\n"
markdown_text += "\n"
elif tag == "ol":
for i, li in enumerate(element.find_all("li", recursive=False), 1):
markdown_text += " " * indent + f"{i}. "
process_element(li, indent + 1)
markdown_text += "\n"
markdown_text += "\n"
elif tag == "table":
rows = element.find_all("tr")
for row in rows:
cells = row.find_all(["td", "th"])
row_text = [cell.text.strip() for cell in cells]
markdown_text += "| " + " | ".join(row_text) + " |\n"
if row == rows[0]: # Header row separator
markdown_text += "| " + " | ".join(["---"] * len(cells)) + " |\n"
markdown_text += "\n"
elif tag == "blockquote":
markdown_text += "> " + element.text.strip().replace("\n", "\n> ") + "\n\n"
elif tag == "strong" or tag == "b":
markdown_text += "**" + element.text.strip() + "**"
elif tag == "em" or tag == "i":
markdown_text += "*" + element.text.strip() + "*"
elif tag == "code":
markdown_text += "`" + element.text.strip() + "`"
elif tag == "pre":
markdown_text += "```\n" + element.text + "\n```\n\n"
elif tag == "hr":
markdown_text += "---\n\n"
else:
for child in element.children:
process_element(child, indent)
process_element(soup)
print(f"👌 Successfully converted HTML to Markdown") if self.verbose else None
return markdown_text
if __name__ == "__main__":
import time
import winloop
URLS = [
"https://en.wikipedia.org/wiki/Treaty_Principles_Bill#:~:text=The%20Treaty%20Principles%20Bill%2C%20or,of%20the%20Treaty%20of%20Waitangi.",
"https://www.parliament.nz/en/pb/sc/make-a-submission/document/54SCJUST_SCF_227E6D0B-E632-42EB-CFFE-08DCFEB826C6/principles-of-the-treaty-of-waitangi-bill",
"https://en.wikipedia.org/wiki/Waitangi_Tribunal",
"https://aljazeera.com/news/2024/11/19/why-are-new-zealands-maori-protesting-over-colonial-era-treaty-bill",
"https://downiewenjack.ca/treaty-of-waitangi-treaty-principles-bill/"
]# * 10 # Make 50 requests
query = "What is the Treaty of Waitangi Bill?"
loop = asyncio.get_event_loop()
custom_crawler = CustomCrawler(max_concurrent_requests=1000)
session_id = loop.run_until_complete(custom_crawler.create_session())
start = time.perf_counter()
winloop.install()
result = loop.run_until_complete(custom_crawler.fetch_page_contents(
URLS,
query,
session_id=session_id,
timeout=20,
max_attempts=1,
return_type="fit_markdown",
)
)
end = time.perf_counter()
loop.run_until_complete(custom_crawler.close_session(session_id))
loop.run_until_complete(custom_crawler.cleanup_expired_sessions())
print("\n\n".join([f"Document {i+1}:\n\n{result[i]}" for i in range(len(result))]))
print(f"\n\nTime taken: {end - start} seconds")
|