8000 Release PyVisionAI v0.3.0: Now with Claude Vision Support · MDGrey33/pyvisionai · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

PyVisionAI v0.3.0: Now with Claude Vision Support
Compare
Choose a tag to compare
@MDGrey33 MDGrey33 released this 20 Feb 14:16
· 2 commits to main since this release
v0.3.0
751749e

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog,
and this project adheres to Semantic Versioning.

[0.3.0] - 2024-03-25

Added

Claude Vision Integration

  • Added ClaudeVisionModel class for Anthropic's Claude Vision API integration
  • Implemented robust retry logic and error handling for Claude API calls
    • Added handling for rate limits and server errors
    • Added specific handling for API overload conditions (Error 529)
    • Implemented exponential backoff for retries
  • Added support for custom prompts with Claude Vision
  • Added describe_image_claude function to main API

Testing Framework

  • Added Claude-specific test markers (@pytest.mark.claude)
  • Added comprehensive test suite for Claude Vision model:
    • Unit tests for initialization, configuration, and error handling
    • Integration tests with real API calls
    • Rate limit and retry logic tests
    • Custom prompt handling tests
    • CLI interface tests

Documentation

  • Added Claude Vision model documentation in docs/getting_started.md
  • Updated API documentation with Claude Vision integration details
  • Added environment setup instructions for Anthropic API key
  • Enhanced testing documentation with Claude-specific examples
  • Updated CLI help messages with Claude Vision options

Configuration

  • Added ANTHROPIC_API_KEY environment variable support
  • Added Claude Vision model configuration in factory system
  • Added retry strategy configuration for API calls

CLI Updates

  • Added Claude Vision support to describe-image CLI command
  • Added -u claude option for model selection
  • Added support for custom prompts with Claude Vision

Enhanced

  • Improved error handling with specific error types for API issues
  • Enhanced retry logic for rate limits and server errors
  • Updated model factory to support Claude Vision
  • Improved test fixtures for better test isolation
  • Enhanced documentation with more comprehensive examples

Fixed

  • Proper handling of empty responses from Claude API
  • Correct error propagation for authentication issues
  • Improved rate limit handling with exponential backoff

[0.2.8] - 2024-01-31

Added

  • Added Homebrew support for easy installation:
    • Created Homebrew formula with all dependencies
    • Added support for both cloud (OpenAI) and local (Ollama) models
    • Automated installation of system dependencies (poppler, libreoffice)
    • Added post-installation verification and helpful setup instructions
    • Comprehensive documentation for Homebrew users

Changed

  • Improved installation process with better dependency management
  • Enhanced system compatibility checks
  • Updated documentation with Homebrew installation instructions

[0.2.7] - 2024-03-22

Added

  • Added retry mechanism for handling transient failures:
    • Implemented RetryManager with configurable strategies
    • Added support for exponential, linear, and constant backoff
    • Added comprehensive logging for retry attempts
    • Added proper error handling and delay management

Changed

  • Improved error handling in model selection:
    • Enhanced connection error handling for API calls
    • Added graceful fallback when default model is unavailable
    • Improved error messages with detailed failure context
  • Enhanced test coverage:
    • Added tests for retry mechanism with various strategies
    • Added tests for model fallback scenarios
    • Added mocked API tests for connection failures

Fixed

  • Fixed model selection to properly handle connection failures
  • Fixed retry delays to prevent excessive wait times
  • Fixed logging to capture all retry and fallback attempts

[0.2.6] - 2024-01-25

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

[0.2.5] - 2024-01-21

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

[0.2.4] - 2024-03-21

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

[0.2.3] - 2024-03-20

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

[0.2.2] - 2024-03-20

Added

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

[0.2.1] - 2024-03-19

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

[0.2.0] - 2024-01-07

Fixed

  • Fixed PDF image extraction where images were being extracted as black (#11)
    • Added proper color space handling for ICC and other PDF color spaces
    • Implemented data decompression and size verification for image data
    • Added validation to detect and skip corrupted or completely black images
    • Improved error handling and logging for image extraction process

Changed

  • Improved image extraction reliability across all supported formats
  • Enhanced error reporting during image processing
  • Implemented parallel processing for image extraction and description to improve performance
  • Updated documentation with more detailed command parameters
  • Restructured README with comprehensive sections on CLI parameters and usage examples

[0.1.1] - 2024-01-07

Added

  • Initial release with support for PDF, DOCX, and PPTX file processing
  • Text and image extraction capabilities
  • Image description using Vision LLMs
  • Command-line interface for file extraction and image description
Assets 4
Loading
0