Configure and diagnose native and Pdfium PDF extraction backends in Xberg.
npx skills add https://github.com/kreuzberg-dev/kreuzberg --skill pdf-backends
Please help me install this Agent Skill. Skill: pdf-backends Source: https://github.com/kreuzberg-dev/kreuzberg/tree/main/.ai-rulez/skills/pdf-backends Command: npx skills add https://github.com/kreuzberg-dev/kreuzberg --skill pdf-backends
Choosing and debugging the right PDF extraction engine in Xberg is error-prone: the native backend and the Pdfium backend have different capabilities, feature flags, and runtime requirements, and silent fallback between them produces misleading results.
PdfConfig.backend chooses between PdfBackend::Native (full pipeline: text, tables, annotations, images, attachments, OCR fallback) and PdfBackend::Pdfium (narrower surface: text, page count, Info-dictionary metadata).pdf-pdfium Cargo feature, PDFIUM_DYNAMIC_LIB_PATH runtime discovery, process-global binding serialization, and actionable errors when the feature or library is missing.force_ocr, force_ocr_pages, and the PdfRenderSession primitive for efficient repeated page rendering.Ask the assistant to switch the Xberg PDF extraction to the Pdfium backend and diagnose why the Pdfium library is not being found at runtime.
High-intent search queries and answers about installing and using this skill.
Set PdfConfig.backend to PdfBackend::Pdfium and build with the pdf-pdfium Cargo feature enabled. Selecting Pdfium without that feature returns an actionable error rather than silently falling back to the native engine.
The native backend supports the full pipeline: text, structure, tables, annotations, images, attachments, metadata, and OCR fallback. Pdfium intentionally offers a narrower surface limited to text, page count, and Info-dictionary metadata, warning about omitted capabilities.
Pdfium uses runtime dynamic loading that honors PDFIUM_DYNAMIC_LIB_PATH as a directory containing the platform library. Build-time static or dynamic link variables are not interchangeable with runtime discovery, so verify the path points to the correct library directory.
Use config.force_ocr to force OCR on the whole document or config.force_ocr_pages to select individual pages. Otherwise, pages without usable native text automatically enter the OCR stage inside PdfExtractor.
Yes, use PdfRenderSession in pdf/render.rs, which opens document bytes once and exposes page count and page rendering. It is a Rust primitive and does not need to be exposed by every language binding.
A missing Pdfium runtime is an infrastructure failure unless the job provisions it. When XBERG_REQUIRE_PDFIUM is set, skipping tests because the library or fixture is absent counts as a failure rather than a skip.
Miosa-osaImport workspaces from local, URL, or GitHub templates with collision handling.
weisbergAggregate cross-platform social performance CSVs into benchmarks and sentiment reports.
AlexKonceptExtract text from PDFs and scanned documents using pymupdf and marker-pdf.
kzinmrExtract text from PDFs and scanned images using OCR.