Pyvisionai

Added
- Implemented Model Factory pattern for vision models:
- Added VisionModel base class with abstract methods
- Added ModelFactory for centralized model management
- Added concrete implementations for GPT4 and Llama models
- Added comprehensive logging for model lifecycle
- Added configuration validation for each model type

Changed
- Refactored model initialization to use factory pattern
- Improved error handling in model creation and validation
- Standardized model interface across all implementations
- Enhanced logging with model-specific context

Documentation
- Added docstrings for new model classes
- Updated logging documentation
- Added model factory usage examples

Added
- Implemented comprehensive logging across all extractors:
- Added structured logging for PDF processing stages
- Added progress tracking for DOCX file conversions and page processing
- Added detailed logging for PPTX slide extraction and conversion
- Added HTML processing status and element detection logging

Changed
- Standardized logging patterns across all extractors:
- Consistent start/completion messages
- Clear error reporting with context
- Progress indicators for multi-step operations
- Performance metrics logging
- Replaced print statements with proper logger calls
- Added logging initialization in all core modules
- Standardized log message format and levels:
- INFO for progress and success
- WARNING for non-critical issues
- ERROR for operation failures

Improved
- Enhanced benchmark testing reliability:
- Added self-contained benchmark test fixtures
- Improved test independence from environment
- Added comprehensive validation of benchmark metrics
- Removed dependency on pre-existing log files
- Added performance metrics logging for both CLI and API interfaces

Documentation
- Added logging configuration examples
- Updated docstrings with logging details
- Added benchmark metrics documentation

Changed
- Implemented parallel processing for DOCX text and images extraction
- Added concurrent processing of paragraphs and images
- Improved performance through ThreadPoolExecutor implementation
- Maintained document structure and content order
- Fixed image placement to ensure correct positioning within text
- Added proper error handling and cleanup
- Performance results: ~72% reduction in processing time (189s → 53s)

- Implemented parallel processing for DOCX page-as-image extraction
- Added PageTask dataclass for encapsulating page processing data
- Introduced process_page method for individual page handling
- Modified extract method to use ThreadPoolExecutor with 4 workers
- Maintained page order using indexed results collection

Fixed
- Added docstring to PDF extractor explaining sequential processing decision
- Fixed test infrastructure to properly use poetry run in CLI tests

Changed
- Implemented parallel processing for PDF page-as-image extraction
- Improved performance by ~68% (from 4 minutes to 1.3 minutes on a 27-page PDF)
- Added ThreadPoolExecutor with 4 workers for concurrent page processing
- Maintained page order while processing in parallel

Added
- Support for custom prompts in image description
- Added support for custom prompts in file extraction

Added
- Support for HTML file extraction using Playwright
- Capability to handle interactive HTML pages with JavaScript rendering
- HTML to image conversion for consistent extraction results
- Simplified the test suite with V2