What it does
Xberg, formerly Kreuzberg, processes office files, PDFs, images, audio, archives, and other sources. It provides a CLI, batch processing, format detection, OCR options, and several output representations.
Pass a path, URL or standard input to extract, or use batch for parallel documents; library APIs and a REST server expose the same core engine. Outputs include plain text, Markdown, Djot, HTML, JSON trees and DocTags. Format detection, OCR/layout options, per-file timeouts and archive nesting/compression safeguards shape extraction. URL ingestion and transcription need their build features; prebuilt packages bundle a common set, while source builds choose features explicitly. Local model routes and provider-hosted OCR/embeddings/LLM extraction have different dependencies and costs. Xberg's MIT engine is separate from the commercial Pro and Enterprise backends.
Strengths
- Combines format detection, batch extraction, and several output representations.
Watch out for
- Some capabilities require optional build features or dependencies.
Best for
- Technical teams normalizing heterogeneous document collections