Advanced Topics¶

This section explores advanced capabilities of the Kreuzberg document intelligence framework, including extensibility mechanisms, performance optimization, and custom integration patterns.

Topics¶

• Performance Guide - Sync vs async performance analysis and optimization
• Error Handling - Detailed strategies for handling extraction errors
• Custom Hooks - Creating custom validation and post-processing hooks
• Custom Extractors - Adding and removing custom extractors

Advanced Configuration¶

Kreuzberg allows you to deeply customize the extraction process through:

1. Validation Hooks: Apply custom validation rules to extraction results
2. Post-Processing Hooks: Transform extraction results before they're returned
3. Custom OCR Settings: Fine-tune OCR behavior for specific document types

Custom Validation Example¶

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18

from kreuzberg import extract_file, ExtractionConfig, ExtractionResult
from kreuzberg import ValidationError

# Define a custom validator that requires a minimum text length
def validate_text_length(result: ExtractionResult) -> None:
    if len(result.content) < 100:
        raise ValidationError("Extracted text is too short", context={"content_length": len(result.content)})

# Use the validator in extraction
async def extract_with_validation(file_path):
    config = ExtractionConfig(validators=[validate_text_length])

    try:
        result = await extract_file(file_path, config=config)
        return result.content
    except ValidationError as e:
        print(f"Validation failed: {e}")
        return None