Kreuzberg: Document Intelligence Framework

2026-05-14 · source · source ↗ · raw capture · wiki page

Overview

Kreuzberg is a polyglot document processing system built on a Rust foundation. The framework extracts text, metadata, images, and structured information from PDFs, Office documents, images, and 97+ formats while supporting 14+ programming languages through native bindings.

Core Capabilities

Text & Metadata Extraction Processes PDF, Office documents, images, HTML, XML, emails, archives, academic formats with intelligent format detection and metadata recovery.

Code Intelligence Extracts functions, classes, imports, symbols, and docstrings from 306 programming languages via tree-sitter. Supports Google, NumPy, Sphinx, JSDoc, RustDoc, and 10+ docstring formats.

OCR Capabilities Supports Tesseract, PaddleOCR, EasyOCR, VLM OCR (GPT-4o, Claude, Gemini, Ollama). Features table detection, reconstruction, and configurable accuracy thresholds.

Language Support

Scripting: Python, Ruby, PHP, Elixir, R, Dart
JavaScript/TypeScript: Node.js, WebAssembly
Compiled: Rust, Go, Java, Kotlin, C#, Swift, Zig, C (FFI)

Format Support

91+ file formats across eight categories: Office documents, images with OCR, web content, archives, email, academic formats.

Deployment Options

Library, CLI tool, REST API server, or MCP server. Docker containers for core and full builds with OCR.

Licensing

Elastic License 2.0 (ELv2).