release: prepare v0.8.0 - Dependency Pattern Refactoring

[AdemBoukhris457]

🚀 Release v0.8.0

This release introduces a major refactoring of OCR and VLM engine configurations to use a dependency injection pattern. This change improves API clarity, enables engine reuse, and provides better separation of concerns. This is a breaking change release - users will need to update their code to use the new API.

⚠️ Breaking Changes

• OCR Engine Configuration: Parser classes now require pre-initialized OCR engine instances instead of individual OCR parameters

  • PytesseractOCREngine and PaddleOCREngine must be instantiated separately and passed to parsers
  • Removed individual OCR parameters (ocr_lang, ocr_psm, ocr_oem, paddleocr_*, etc.) from parser constructors
  • Default behavior: If no ocr_engine is provided, PytesseractOCREngine() is used automatically

• VLM Configuration: Parser classes now require pre-initialized VLM engine instances instead of individual VLM parameters

  • VLMStructuredExtractor must be instantiated separately and passed to parsers
  • Removed individual VLM parameters (use_vlm, vlm_provider, vlm_model, vlm_api_key) from parser constructors
  • Default behavior: If no vlm is provided, VLM features are disabled

✨ Improvements

• Dependency Injection Pattern: Cleaner API design with explicit engine initialization

  • Engines can be created once and reused across multiple parsers
  • Better error handling - engine initialization errors are caught before parser creation
  • Prevents mixed configurations and improves code maintainability

• Code Quality: Removed unnecessary comments from codebase for cleaner code

• Consistent API: Both OCR and VLM engines now follow the same dependency pattern

🔧 Technical Changes

• Refactored all parser classes:

  • StructuredPDFParser
  • EnhancedPDFParser
  • ChartTablePDFParser
  • StructuredDOCXParser

• Updated CLI commands to create engine instances before parser initialization

• Updated all UI components (Gradio apps) to use new engine initialization pattern

• Enhanced error handling for engine initialization failures

📚 Documentation

• Updated README with new OCR and VLM engine configuration examples
• Added comprehensive migration guides in documentation
• Updated all API reference documentation
• Updated all usage examples and tutorials
• Added engine configuration sections in user guides

🔄 Migration Guide

Before (v0.7.1):

parser = StructuredPDFParser(
    ocr_lang="eng",
    ocr_psm=6,
    use_vlm=True,
    vlm_provider="openai",
    vlm_api_key="your-key"
)

After (v0.8.0):

from doctra.engines.ocr.pytesseract import PytesseractOCREngine
from doctra.engines.vlm.service import VLMStructuredExtractor

# Initialize engines
ocr_engine = PytesseractOCREngine(lang="eng", psm=6)
vlm_engine = VLMStructuredExtractor(
    vlm_provider="openai",
    api_key="your-key"
)

# Pass to parser
parser = StructuredPDFParser(
    ocr_engine=ocr_engine,
    vlm=vlm_engine
)

For CLI users:

• No changes needed - CLI automatically handles engine initialization
• All existing CLI commands work the same way

For UI users:

• No changes needed - UI components automatically handle engine initialization
• All existing UI workflows work the same way

📦 Compatibility

• Python 3.8+ (unchanged)
• All existing dependencies maintained
• No new dependencies added

Related PRs

• refactor VLM configuration to use dependency pattern #78: refactor VLM configuration to use dependency pattern
• refactor OCR engine configuration to use dependency pattern #77: refactor OCR engine configuration to use dependency pattern
• refactor: remove unnecessary comments from codebase #76: refactor: remove unnecessary comments from codebase