The library no longer supports Python 3.9 and so it was important to drop all references to python 3.9.
Following changes have been made:
- pyproject.toml: set requires-python to ">=3.10"; remove 3.9 classifier
- setup.py: set python_requires to ">=3.10"; remove 3.9 classifier
- docs: update Python version mentions
- deploy/docker/c4ai-doc-context.md: options -> 3.10, 3.11, 3.12, 3.13
Added a new `preserve_https_for_internal_links` configuration flag that preserves the original HTTPS scheme for same-domain links even when the server redirects to HTTP.
- Updated ProxyConfig.from_string to support multiple proxy formats, including URLs with credentials.
- Deprecated the 'proxy' parameter in BrowserConfig, replacing it with 'proxy_config' for better flexibility.
- Added warnings for deprecated usage and clarified behavior when both parameters are provided.
- Updated documentation and tests to reflect changes in proxy configuration handling.
- Return comprehensive error messages along with status codes for api internal errors.
- Fix fit_html property serialization issue in both /crawl and /crawl/stream endpoints
- Add sanitization to ensure fit_html is always JSON-serializable (string or None)
- Add comprehensive error handling test suite.
Implement hierarchical configuration for LLM parameters with support for:
- Temperature control (0.0-2.0) to adjust response creativity
- Custom base_url for proxy servers and alternative endpoints
- 4-tier priority: request params > provider env > global env > defaults
Add helper functions in utils.py, update API schemas and handlers,
support environment variables (LLM_TEMPERATURE, OPENAI_TEMPERATURE, etc.),
and provide comprehensive documentation with examples.
- Fix URLPatternFilter serialization by preventing private __slots__ from being serialized as constructor params
- Add public attributes to URLPatternFilter to store original constructor parameters for proper serialization
- Handle property descriptors in CrawlResult.model_dump() to prevent JSON serialization errors
- Ensure filter chains work correctly with Docker client and REST API
The issue occurred because:
1. Private implementation details (_simple_suffixes, etc.) were being serialized and passed as constructor arguments during deserialization
2. Property descriptors were being included in the serialized output, causing "Object of type property is not JSON serializable" errors
Changes:
- async_configs.py: Comment out __slots__ serialization logic (lines 100-109)
- filters.py: Add patterns, use_glob, reverse to URLPatternFilter __slots__ and store as public attributes
- models.py: Convert property descriptors to strings in model_dump() instead of including them directly
Fixes bug reported in issue #1405
[Bug]: Excluded selector (excluded_selector) doesn't work
This commit reintroduces the cssselect library which was removed by PR (https://github.com/unclecode/crawl4ai/pull/1368) and merged via (437395e490).
Integration tested against 0.7.4 Docker container. Reintroducing cssselector package eliminated errors seen in logs and excluded_selector functionality was restored.
Refs: #1405
This commit adds a complete, web scraping API example that demonstrates how to get structured data from any website and use it like an API using the crawl4ai library with a minimalist frontend interface.
Core Functionality
- AI-powered web scraping with plain English queries
- Dual scraping approaches: Schema-based (faster) and LLM-based (flexible)
- Intelligent schema caching for improved performance
- Custom LLM model support with API key management
- Automatic duplicate request prevention
Modern Frontend Interface
- Minimalist black-and-white design inspired by modern web apps
- Responsive layout with smooth animations and transitions
- Three main pages: Scrape Data, Models Management, API Request History
- Real-time results display with JSON formatting
- Copy-to-clipboard functionality for extracted data
- Toast notifications for user feedback
- Auto-scroll to results when scraping starts
Model Management System
- Web-based model configuration interface
- Support for any LLM provider (OpenAI, Gemini, Anthropic, etc.)
- Simplified configuration requiring only provider and API token
- Add, list, and delete model configurations
- Secure storage of API keys in local JSON files
API Request History
- Automatic saving of all API requests and responses
- Display of request history with URL, query, and cURL commands
- Duplicate prevention (same URL + query combinations)
- Request deletion functionality
- Clean, simplified display focusing on essential information
Technical Implementation
Backend (FastAPI)
- RESTful API with comprehensive endpoints
- Pydantic models for request/response validation
- Async web scraping with crawl4ai library
- Error handling with detailed error messages
- File-based storage for models and request history
Frontend (Vanilla JS/CSS/HTML)
- No framework dependencies - pure HTML, CSS, JavaScript
- Modern CSS Grid and Flexbox layouts
- Custom dropdown styling with SVG arrows
- Responsive design for mobile and desktop
- Smooth scrolling and animations
Core Library Integration
- WebScraperAgent class for orchestration
- ModelConfig class for LLM configuration management
- Schema generation and caching system
- LLM extraction strategy support
- Browser configuration with headless mode
- Replace BaseStrategy with CrawlStrategy in custom strategy examples (DomainSpecificStrategy, HybridStrategy)
- Remove “Custom Link Scoring” and “Caching Strategy” sections no longer aligned with current library
- Revise memory pruning example to use adaptive.get_relevant_content and index-based retention of top 500 docs
- Correct Quickstart note: default cache mode is CacheMode.BYPASS; instruct enabling with CacheMode.ENABLED
Previously, the system incorrectly used OPENAI_API_KEY for all LLM providers
due to a hardcoded api_key_env fallback in config.yml. This caused authentication
errors when using non-OpenAI providers like Gemini.
Changes:
- Remove api_key_env from config.yml to let litellm handle provider-specific env vars
- Simplify get_llm_api_key() to return None, allowing litellm to auto-detect keys
- Update validate_llm_provider() to trust litellm's built-in key detection
- Update documentation to reflect the new automatic key handling
The fix leverages litellm's existing capability to automatically find the correct
environment variable for each provider (OPENAI_API_KEY, GEMINI_API_TOKEN, etc.)
without manual configuration.
ref #1291
- Add comprehensive v0.7.4 release blog post with LLMTableExtraction feature highlight
- Update blog index to feature v0.7.4 as latest release
- Update README.md to showcase v0.7.4 features alongside v0.7.3
- Accurately describe dispatcher fix as bug fix rather than major enhancement
- Include practical code examples for new LLMTableExtraction capabilities
BREAKING CHANGE: Table extraction now uses Strategy Design Pattern
This epic commit introduces a game-changing approach to table extraction in Crawl4AI:
✨ NEW FEATURES:
- LLMTableExtraction: AI-powered extraction for complex HTML tables with rowspan/colspan
- Smart Chunking: Automatically splits massive tables into optimal chunks at row boundaries
- Parallel Processing: Processes multiple chunks simultaneously for blazing-fast extraction
- Intelligent Merging: Seamlessly combines chunk results into complete tables
- Header Preservation: Each chunk maintains context with original headers
- Auto-retry Logic: Built-in resilience with configurable retry attempts
🏗️ ARCHITECTURE:
- Strategy Design Pattern for pluggable table extraction strategies
- ThreadPoolExecutor for concurrent chunk processing
- Token-based chunking with configurable thresholds
- Handles tables without headers gracefully
⚡ PERFORMANCE:
- Process 1000+ row tables without timeout
- Parallel processing with up to 5 concurrent chunks
- Smart token estimation prevents LLM context overflow
- Optimized for providers like Groq for massive tables
🔧 CONFIGURATION:
- enable_chunking: Auto-handle large tables (default: True)
- chunk_token_threshold: When to split (default: 3000 tokens)
- min_rows_per_chunk: Meaningful chunk sizes (default: 10)
- max_parallel_chunks: Concurrent processing (default: 5)
📚 BACKWARD COMPATIBILITY:
- Existing code continues to work unchanged
- DefaultTableExtraction remains the default strategy
- Progressive enhancement approach
This is the future of web table extraction - handling everything from simple tables to massive, complex data grids with merged cells and nested structures. The chunking is completely transparent to users while providing unprecedented scalability.
- fix handling of special keys in Windows msvcrt implementation
- Guard against UnicodeDecodeError from multi-byte key sequences
- Filter out non-printable characters and control sequences
- Add error handling to prevent coroutine crashes
- Add unit test to verify keyboard input handling
Key changes:
- Safe UTF-8 decoding with try/except for special keys
- Skip non-printable and multi-byte character sequences
- Add broad exception handling in keyboard listener
Test runs on Windows only due to msvcrt dependency.
This commit introduces platform-specific handling for the 'q' key press to quit the browser profiler, ensuring compatibility with both Windows and Unix-like systems. It also adds a check to see if the browser process has already exited, terminating the input listener if so.
- Implemented `msvcrt` for Windows to capture keyboard input without requiring a newline.
- Retained `termios`, `tty`, and `select` for Unix-like systems.
- Added a check for browser process termination to gracefully exit the input listener.
- Updated logger messages to use colored output for better user experience.
BREAKING CHANGE: Table extraction now uses Strategy Design Pattern
This epic commit introduces a game-changing approach to table extraction in Crawl4AI:
✨ NEW FEATURES:
- LLMTableExtraction: AI-powered extraction for complex HTML tables with rowspan/colspan
- Smart Chunking: Automatically splits massive tables into optimal chunks at row boundaries
- Parallel Processing: Processes multiple chunks simultaneously for blazing-fast extraction
- Intelligent Merging: Seamlessly combines chunk results into complete tables
- Header Preservation: Each chunk maintains context with original headers
- Auto-retry Logic: Built-in resilience with configurable retry attempts
🏗️ ARCHITECTURE:
- Strategy Design Pattern for pluggable table extraction strategies
- ThreadPoolExecutor for concurrent chunk processing
- Token-based chunking with configurable thresholds
- Handles tables without headers gracefully
⚡ PERFORMANCE:
- Process 1000+ row tables without timeout
- Parallel processing with up to 5 concurrent chunks
- Smart token estimation prevents LLM context overflow
- Optimized for providers like Groq for massive tables
🔧 CONFIGURATION:
- enable_chunking: Auto-handle large tables (default: True)
- chunk_token_threshold: When to split (default: 3000 tokens)
- min_rows_per_chunk: Meaningful chunk sizes (default: 10)
- max_parallel_chunks: Concurrent processing (default: 5)
📚 BACKWARD COMPATIBILITY:
- Existing code continues to work unchanged
- DefaultTableExtraction remains the default strategy
- Progressive enhancement approach
This is the future of web table extraction - handling everything from simple tables to massive, complex data grids with merged cells and nested structures. The chunking is completely transparent to users while providing unprecedented scalability.
- Wrap all AsyncUrlSeeder usage with async context managers
- Update URL seeding adventure example to use "sitemap+cc" source, focus on course posts, and add stream=True parameter to fix runtime error
