Metadata Extraction¶

Kreuzberg automatically extracts rich metadata from documents including titles, authors, dates, and format-specific properties.

How It Works¶

The ExtractionResult includes a metadata dictionary with available metadata properties. Each format uses specialized extractors:

• PDF: Uses Playa to extract document information, structure, and properties
• Office: Uses Pandoc for Word, PowerPoint, and other formats
• Images: Extracts dimensions and format properties
• HTML: Extracts meta tags and structured data

Metadata Fields¶

The Metadata dictionary is a TypedDict with optional fields. Only available fields are included in the results.

Usage Example¶

Accessing document metadata:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14

from kreuzberg import extract_file

# Extract from a file
result = await extract_file("document.pdf")

# Access metadata - use .get() with a default value for safety
title = result.metadata.get("title", "Untitled Document")
authors = result.metadata.get("authors", ["Unknown Author"])
created_date = result.metadata.get("created_at")

# Print available metadata
print("Available metadata fields:")
for key, value in result.metadata.items():
    print(f"{key}: {value}")

PDF-Specific Metadata¶

For PDF documents, Kreuzberg extracts a rich set of metadata including:

• Document information dictionary values (title, author, subject, keywords, etc.)
• Document structure information including page count, dimensions, and outline
• Font information from document pages
• PDF/A compliance status
• Encryption status and permissions
• Language information when available in document structure

If a PDF document contains UTF-16BE encoded strings (often present in PDF metadata with a byte order mark \xfe\xff), Kreuzberg will automatically detect and decode these properly.

Working with Multiple Document Types¶

When working with multiple document types, it's important to remember that different document formats may provide different metadata fields. Always use defensive programming (like using .get() with a default value) when accessing metadata fields:

1
2
3

# Safe way to access metadata across different document types
author = result.metadata.get("authors", ["Unknown"])[0] if "authors" in result.metadata else "Unknown"
creation_date = result.metadata.get("created_at", "Unknown date")

Viewing Available Metadata¶

To view all available metadata for a document:

1
2
3
4
5

from kreuzberg import extract_file
import json

result = await extract_file("your_file.pdf")
print(json.dumps(result.metadata, indent=2, default=str))

This will print all available metadata fields for the document in a readable format.