v0.4.0

Features

CLI Improvements

• Align CLI with the refactored core architecture
• Introduce global --verbose / -v flag for debugging
• Add --version flag for quick version checks
• Improve structured logging across all CLI commands
• Ensure consistent exit codes and error handling

Performance Improvements

• Dependency checks are 90–95% faster via intelligent caching
• CLI initialization is 75–85% faster on subsequent runs
• Validation only evaluates the selected provider
• Lazy loading of configuration using caching decorators

Refactoring

Core Architecture

• Centralize LLM response handling to eliminate JSON parsing duplication
• Apply template method pattern to LLM base clients
• Simplify provider implementations with major code reduction
• Decompose monolithic pipeline logic into composable, testable stages
• Introduce YAML-based configuration for models and pipelines
• Replace multiple error-handling patterns with a unified exception hierarchy

Testing

• Update all unit tests to reflect the refactored clients, pipeline stages, and configuration model
• Align integration tests with new CLI behavior, exit codes, and error semantics
• Improve test isolation and reliability following pipeline decomposition

Documentation

Complete Documentation Overhaul

• Rebuild the entire docs/ folder into a 12-section, pipeline-oriented structure
• Add comprehensive examples, configuration guides, and API references
• Rewrite README to match new CLI usage and uv-based workflows
• Update MkDocs navigation and add a new documentation homepage
• Include Mermaid diagrams and clear “Next Steps” navigation throughout

Chores

• Fix PyPI deployment in the release workflow

Impact & Migration Notes

• This release introduces breaking changes due to core refactoring and API cleanup
• Users relying on internal APIs, custom pipelines, or CLI scripts may need to adjust usage
• Refer to the updated documentation for new configuration, CLI flags, and examples