Skip to content

CHANGELOG.md

Latest commit

 

History

History
184 lines (145 loc) · 7.19 KB

CHANGELOG.md

File metadata and controls

184 lines (145 loc) · 7.19 KB

Changelog

All notable changes to LLM WebExtract will be documented in this file.

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

[1.2.2] - 2024-12-18

🆕 New Features

  • Added: extract_with_ollama() convenience function for simplified Ollama usage
  • Enhanced: Updated README with improved provider setup examples
  • Added: Quick one-liner examples for all providers in documentation

📚 Documentation

  • Updated: README.md with extract_with_ollama function examples
  • Improved: Provider setup section with both ConfigBuilder and convenience function usage
  • Enhanced: Quick start examples now show multiple usage patterns

[1.1.2] - 2024-12-16

🚀 Major Improvements

✅ Multi-Provider LLM Support (BREAKING FIX)

  • FIXED: Multi-provider support now actually works (was broken in previous versions)
  • Added: Full OpenAI GPT integration with proper error handling
  • Added: Complete Anthropic Claude integration
  • Added: LLM provider factory for dynamic client creation
  • Added: Provider availability checking and fallback strategies

🛡️ Robust Error Handling

  • Added: Comprehensive exception hierarchy (ExtractionError, ScrapingError, LLMError, AuthenticationError, ConfigurationError)
  • Fixed: Proper error propagation instead of silent failures
  • Added: Specific error messages for debugging different failure types
  • Added: Graceful handling of missing dependencies and invalid API keys

📧 Configuration Updates

  • Updated: Email address to himasha626@gmail.com
  • Fixed: Version consistency across all configuration files (1.1.2)
  • Fixed: Repository URLs pointing to correct GitHub repository
  • Fixed: Default model name consistency (llama3.2)

📚 Documentation Overhaul

📖 Enhanced Documentation

  • Added: Comprehensive README with badges, clear examples, and troubleshooting
  • Added: Complete API reference documentation (API_REFERENCE.md)
  • Added: Mermaid architecture diagrams showing how the system works
  • Added: Multiple usage patterns from basic to advanced
  • Added: Provider-specific setup instructions
  • Added: Real-world use cases and examples

🔧 Improved Examples

  • Updated: basic_usage.py with multi-provider examples
  • Added: multi_provider_example.py demonstrating fallback strategies
  • Added: Error handling examples throughout
  • Added: Cloud provider integration examples

🏗️ Infrastructure Improvements

⚡ GitHub Actions Workflows

  • Improved: CI workflow with better caching and parallel execution
  • Enhanced: Release workflow with automatic changelog generation
  • Added: Comprehensive code quality checks with auto-formatting
  • Added: Security analysis with safety and bandit
  • Added: Multi-Python version testing (3.8-3.12)
  • Added: Manual release triggers for better control

🧹 Project Cleanup

  • Removed: Build artifacts and temporary files
  • Organized: Clean project structure
  • Updated: All workflows to use latest GitHub Actions
  • Added: Better artifact management and retention policies

🔧 Code Quality

🏛️ Architecture Improvements

  • Refactored: LLM client architecture with proper base class
  • Added: Abstract base class for LLM clients (BaseLLMClient)
  • Improved: Configuration handling with proper validation
  • Added: Factory pattern for LLM client creation

🔒 Type Safety & Validation

  • Enhanced: Pydantic models with better validation
  • Added: Comprehensive type hints throughout codebase
  • Improved: Configuration validation and error reporting
  • Added: Runtime type checking for critical operations

🆕 New Features

🎛️ Enhanced Configuration

  • Added: ConfigBuilder fluent API for easy setup
  • Added: Environment variable support for all configuration options
  • Added: Pre-built profiles for different content types
  • Added: Custom prompt and schema support

📊 Better Data Handling

  • Added: Confidence scoring for extraction quality
  • Added: Smart caching with configurable policies
  • Added: Batch processing capabilities
  • Added: Content length limits and intelligent truncation

🐛 Bug Fixes

  • Fixed: JSON parsing reliability with multiple fallback strategies
  • Fixed: Content extraction edge cases
  • Fixed: Memory leaks in browser management
  • Fixed: Rate limiting implementation
  • Fixed: URL validation and error handling

💔 Breaking Changes

  • BREAKING: Exception handling - specific exception types instead of generic ones
  • BREAKING: Configuration API - some legacy config methods deprecated
  • BREAKING: LLM client initialization - now requires proper configuration

📦 Dependencies

  • Updated: All dependencies to latest compatible versions
  • Added: Optional dependencies for OpenAI and Anthropic
  • Improved: Dependency management with proper optional extras

🔄 Migration Guide

For users upgrading from 1.0.x:

# Old way (still works but deprecated)
from webextract import WebExtractor
extractor = WebExtractor()  # Used Ollama hardcoded

# New way (recommended)
from webextract import WebExtractor, ConfigBuilder
config = ConfigBuilder().with_ollama("llama3.2").build()
extractor = WebExtractor(config)

# Multi-provider support (now actually works!)
config = ConfigBuilder().with_openai("sk-...", "gpt-4o-mini").build()
extractor = WebExtractor(config)

[1.0.4] - 2024-06-14

Fixed

  • Fixed JSON parsing errors in LLM responses
  • Improved structured data validation and type checking
  • Enhanced error handling with better fallback responses
  • Resolved "Structured data missing required fields" warnings
  • Fixed single quotes to double quotes conversion in JSON
  • Added robust JSON string repair mechanisms

Added

  • Comprehensive example files for different use cases
  • Enhanced LLM prompts for more consistent JSON output
  • Better field type validation and normalization
  • Improved debugging and logging for JSON parsing issues

Changed

  • Enhanced default prompt template for clearer JSON instructions
  • Improved validation logic for structured data fields
  • More robust fallback response generation

[1.0.0] - 2024-12-13

Added

  • Initial release of LLM WebExtract
  • Support for multiple LLM providers (Ollama, OpenAI, Anthropic)
  • Modern web scraping with Playwright
  • Flexible configuration system with builder pattern
  • Pre-built configuration profiles for common use cases
  • CLI interface with rich output formatting
  • Structured data extraction with confidence scoring
  • Comprehensive documentation and examples

Package Features

  • Cross-platform support (Windows, macOS, Linux)
  • Python 3.8+ compatibility
  • Comprehensive test suite
  • GitHub Actions CI/CD pipeline
  • Automatic PyPI publishing on releases
  • Security scanning and code quality checks

Author

  • Created by Himasha Herath

Note: Version 1.1.2 represents a major stability and functionality improvement over previous versions. The multi-provider support that was advertised but broken in earlier versions now works correctly.