Kreuzberg Document Intelligence

4.8.5 · active · verified Fri Apr 17

Kreuzberg is a high-performance Python library for document intelligence, enabling extraction of text, metadata, and structured data from PDFs, Office documents, images, and over 88 other formats. It leverages a Rust core for significant speed improvements (10-50x faster) compared to pure Python alternatives. The current version is 4.8.5, with an active release cadence, typically releasing minor updates every few weeks.

Common errors

```
ModuleNotFoundError: No module named 'kreuzberg'
```
cause The Kreuzberg library is not installed in your current Python environment.

fix Run `pip install kreuzberg` to install the library.
```
TypeError: extract() got an unexpected keyword argument 'config'
```
cause This error typically indicates a version mismatch where the `extract` function's signature has changed, or you are using an older example with a newer library version (or vice-versa). The quickstart uses `config=config`.

fix Ensure you are using the latest `kreuzberg` version and refer to the official documentation or the quickstart code for correct `extract` function usage. The current API expects `extract(file_path, config=...)`.
```
ERROR: Package 'kreuzberg' requires a different Python: 3.9.x not in '>=3.10'
```
cause You are attempting to install or run Kreuzberg on a Python version older than 3.10, which is not supported.

fix Upgrade your Python installation to 3.10 or higher, or activate a virtual environment that uses a supported Python version.
```
My HTML output from Kreuzberg is suddenly styled with CSS, but I wanted plain HTML.
```
cause Starting with `v4.8.1`, HTML output gained default styling via `HtmlOutputConfig`. If you didn't specify `html_output`, it now defaults to a styled theme.

fix To explicitly get unstyled HTML, set `html_output=HtmlOutputConfig(theme="unstyled")` in your `ExtractionConfig`.

Warnings

breaking Kreuzberg requires Python 3.10 or newer. Installing or running the library on older Python versions will result in errors.
Fix: Upgrade your Python environment to 3.10 or a more recent version (e.g., 3.11, 3.12).
gotcha When extracting in HTML format, versions 4.8.1 and later introduced default styling. If you were expecting plain, unstyled HTML, your output will now include CSS and semantic classes.
Fix: To get unstyled HTML, explicitly set `html_output=HtmlOutputConfig(theme="unstyled")` in your `ExtractionConfig`.
gotcha Versions prior to 4.8.2 had a bug where legitimate repeated content (e.g., brand names, headers) in PDFs could be stripped, even if `strip_repeating_text` was not enabled or intended.
Fix: Upgrade to `kreuzberg v4.8.2` or newer to resolve the issue with overly aggressive content stripping.
gotcha Users on macOS ARM64 systems (e.g., M1/M2/M3 Macs) using `kreuzberg` versions older than `v4.7.3` might experience a `SIGBUS` (Bus error: 10) crash when processing archive files (ZIP, 7Z, TAR, GZIP).
Fix: Upgrade `kreuzberg` to `v4.7.3` or a newer version to fix the archive extraction crash.

Install

pip install kreuzberg Install latest version

Imports

extract
```
from kreuzberg import extract
```
ExtractionConfig
```
from kreuzberg import ExtractionConfig
```
OutputFormat
```
from kreuzberg import OutputFormat
```
HtmlOutputConfig
```
from kreuzberg import HtmlOutputConfig
```

Quickstart

This quickstart demonstrates how to perform basic text and styled HTML extraction using Kreuzberg's `extract` function with `ExtractionConfig` and `OutputFormat`. It shows how to specify the output format and customize HTML output with `HtmlOutputConfig`.

import os
from kreuzberg import extract, ExtractionConfig, OutputFormat, HtmlOutputConfig

# Create a dummy file for demonstration
with open("example.txt", "w") as f:
    f.write("This is a test document for Kreuzberg extraction.")

# Example 1: Basic text extraction
config_text = ExtractionConfig(
    output_format=OutputFormat.TEXT
)
result_text = extract("example.txt", config=config_text)
print("--- Text Extraction ---")
print(result_text.text)

# Example 2: HTML extraction with a specific theme
config_html = ExtractionConfig(
    output_format=OutputFormat.HTML,
    html_output=HtmlOutputConfig(theme="github")
)
result_html = extract("example.txt", config=config_html)
print("\n--- HTML Extraction (GitHub theme) ---")
print(result_html.html)

os.remove("example.txt") # Clean up the dummy file

view raw JSON →