Add comprehensive documentation for the new streaming mode feature in arun_many():
- Update arun_many() API docs to reflect streaming return type
- Add streaming examples in quickstart and multi-url guides
- Document stream parameter in configuration classes
- Add clone() helper method documentation for configs
This change improves documentation for processing large numbers of URLs efficiently.
Add streaming capability to crawler configurations and introduce clone() methods
for both BrowserConfig and CrawlerRunConfig to support immutable config updates.
Move stream parameter from arun_many() method to CrawlerRunConfig.
BREAKING CHANGE: Removed stream parameter from AsyncWebCrawler.arun_many() method.
Use config.stream=True instead.
Add shared_data parameter to CrawlerRunConfig to allow data sharing between hooks.
Implement browser context reuse based on config signatures to improve memory usage.
Fix Firefox/Webkit channel settings.
Add config parameter to hook callbacks for better context access.
Remove debug print statements.
BREAKING CHANGE: Hook callback signatures now include config parameter
Add new streaming capability to the MemoryAdaptiveDispatcher and AsyncWebCrawler
to allow processing URLs with real-time result streaming. This enables
processing results as they become available rather than waiting for all
URLs to complete.
Key changes:
- Add run_urls_stream method to MemoryAdaptiveDispatcher
- Update AsyncWebCrawler.arun_many to support streaming mode
- Add result queue for better result handling
- Improve type hints and documentation
BREAKING CHANGE: The return type of arun_many now depends on the 'stream'
parameter, returning either List[CrawlResult] or AsyncGenerator[CrawlResult, None]
Add new LLMContentFilter class that uses LLMs to generate high-quality markdown content:
- Implement intelligent content filtering with customizable instructions
- Add chunk processing for handling large documents
- Support parallel processing of content chunks
- Include caching mechanism for filtered results
- Add usage tracking and statistics
- Update documentation with examples and use cases
Also includes minor changes:
- Disable Pydantic warnings in __init__.py
- Add new prompt template for content filtering
Implement more robust browser executable path handling using playwright's built-in browser management. This change:
- Adds async browser path resolution
- Implements path caching in the home folder
- Removes hardcoded browser paths
- Adds httpx dependency
- Removes obsolete test result files
This change makes the browser path resolution more reliable across different platforms and environments.
- Increase memory threshold from 70% to 90% for better resource utilization
- Remove incorrect self parameter from MemoryAdaptiveDispatcher initialization
These changes improve the crawler's performance by allowing more memory usage before throttling and fix a bug in dispatcher initialization.
Make fields in MediaItem and Link models optional with default values to prevent validation errors when data is incomplete. Also expose BaseDispatcher in __init__ and fix markdown field handling in database manager.
BREAKING CHANGE: MediaItem and Link model fields are now optional with default values which may affect existing code expecting required fields.
- Remove .pre-commit-config.yaml and duplicate mkdocs configuration files
- Add Optional type hint for proxy parameter in BrowserConfig
- Fix type annotation for results list in AsyncWebCrawler
- Move calculate_batch_size function import to model_loader
- Update prompt imports in extraction_strategy.py
No breaking changes.
Replace the ScrapingMode enum with a proper strategy pattern implementation for content scraping.
This change introduces:
- New ContentScrapingStrategy abstract base class
- Concrete WebScrapingStrategy and LXMLWebScrapingStrategy implementations
- New Pydantic models for structured scraping results
- Updated documentation reflecting the new strategy-based approach
BREAKING CHANGE: ScrapingMode enum has been removed. Users should now use ContentScrapingStrategy implementations instead.
Adds a new ScrapingMode enum to allow switching between BeautifulSoup and LXML parsing.
LXML mode offers 10-20x better performance for large HTML documents.
Key changes:
- Added ScrapingMode enum with BEAUTIFULSOUP and LXML options
- Implemented LXMLWebScrapingStrategy class
- Added LXML-based metadata extraction
- Updated documentation with scraping mode usage and performance considerations
- Added cssselect dependency
BREAKING CHANGE: None
Reorganize dispatcher functionality into separate components:
- Create dedicated dispatcher classes (MemoryAdaptive, Semaphore)
- Add RateLimiter for smart request throttling
- Implement CrawlerMonitor for real-time progress tracking
- Move dispatcher config from CrawlerRunConfig to separate classes
BREAKING CHANGE: Dispatcher configuration moved from CrawlerRunConfig to dedicated dispatcher classes. Users need to update their configuration approach for multi-URL crawling.
Implements a new MemoryAdaptiveDispatcher class to manage concurrent crawling operations with memory monitoring and rate limiting capabilities. Changes include:
- Added RateLimitConfig dataclass for configuring rate limiting behavior
- Extended CrawlerRunConfig with dispatcher-related settings
- Refactored arun_many to use the new dispatcher system
- Added memory threshold and session permit controls
- Integrated optional progress monitoring display
BREAKING CHANGE: The arun_many method now uses MemoryAdaptiveDispatcher by default, which may affect concurrent crawling behavior
Update all documentation URLs from crawl4ai.com/mkdocs to docs.crawl4ai.com
Improve badges styling and layout in documentation
Increase code font size in documentation CSS
BREAKING CHANGE: Documentation URLs have changed from crawl4ai.com/mkdocs to docs.crawl4ai.com
Revise the README's personal story section to better reflect the project's
origins, motivation, and vision for open-source data accessibility. Add more
detail about the creator's background and the project's mission to
democratize AI through open data access.
Also includes a minor TODO comment addition in async crawler strategy.
Reorganize documentation into core/advanced/extraction sections for better navigation.
Update terminal theme styles and add rich library for better CLI output.
Remove redundant tutorial files and consolidate content into core sections.
Add personal story to index page for project context.
BREAKING CHANGE: Documentation structure has been significantly reorganized
Add explanatory comments to JsonCssExtractionStrategy._get_elements() method to clarify that it returns all matching elements using select() instead of select_one(). This helps developers understand the method's behavior and its difference from single element selection.
Removed trailing whitespace at end of file.
- Fix JsonCssExtractionStrategy._get_elements to return all matching elements instead of just one
- Add robust error handling to page_need_scroll with default fallback
- Improve JSON extraction strategies documentation
- Refactor content scraping strategy
- Update version to 0.4.247
- Fixes critical memory leak issue where browser pages remained open
- Ensures proper cleanup of Playwright resources after page operations
- Improves resource management in browser farm implementation
This is an urgent fix to address resource leakage that could impact system stability.
- Set wait_for_images default to false for better performance
- Simplify response attribute copying in AsyncWebCrawler
- Update hello_world example with proper content filtering
BREAKING CHANGE: Updated `chrome_channel` to "chromium" to fix compatibility with the new Chromium headless implementation. This resolves the error `playwright._impl._errors.Error: BrowserType.launch: Chromium distribution 'chrome' is not found`, caused by the removal of the old headless mode in Chromium.
With this change, channels like "chrome" and "msedge" now default to the new headless mode, aligning with upstream updates in Playwright v1.49. The new headless mode uses the real Chrome browser, offering more authenticity, reliability, and feature parity with the full browser.
Additionally, simplified fallback logic by directly assigning `chrome_channel` based on `browser_type` or defaulting to "chromium".
Refer to:
- https://playwright.dev/python/docs/browsers#chromium
- https://github.com/microsoft/playwright/issues/33566
- Replace explicit package listing with setuptools.find
- Include all crawl4ai.* packages automatically
- Use `packages = {find = {where = ["."], include = ["crawl4ai*"]}}` syntax
- Bump version to 0.4.243
This change simplifies package maintenance by automatically discovering
all subpackages under crawl4ai namespace instead of listing them manually.
- Add --force flag to Playwright browser installation
- Add doctor command to test crawling functionality
- Install Chrome and Chromium browsers explicitly
- Add crawl4ai-doctor entry point in pyproject.toml
- Implement simple health check focused on crawling test
