Xberg v1: Rust-core multilingual document intelligence engine with 107 formats

Xberg is a rebuild and rename of Kreuzberg, positioned as a Rust-core multilingual document intelligence engine. It unifies format detection, reading, OCR, and structured extraction, covering 107 formats, 141 file extensions, and code intelligence for 371 programming languages, with 6 output formats. It offers 15 language bindings from Python to C FFI and runs as a library, CLI, REST API, or MCP server. The core change is how capabilities are packaged: audio and video transcription runs on Whisper ONNX, OCR supports swappable backends like Tesseract and PaddleOCR, layout and table structure are reconstructed by ML models, and structured extraction outputs JSON schema via local Ollama or hosted LLMs without prompt engineering. Docker images bundle common features by default, while source builds enable them on demand. For teams running document pipelines on Kreuzberg, upgrading means all package names and dependency coordinates change, including pip, npm, and cargo import paths. Developers integrating with Claude Desktop or Cursor can configure MCP directly, avoiding the work of stitching together PDF parsing, OCR, and transcription libraries themselves. Go users should note the repo root is not a Go module and must point to the packages/go subdirectory.