Docling Studio — Open-Source Visual Inspection for Docling Pipelines

If you use Docling for document AI, you know the engine is solid — PDF parsing, layout analysis, table detection, OCR. But inspecting what the pipeline actually extracts means scrolling through JSON and mentally mapping bounding box coordinates back to the original pages.

Docling Studio fixes that.

What it is

An open-source visual layer on top of Docling ( ds4sd ). Upload a PDF, configure your pipeline options (OCR engine, table extraction, enrichment), run the conversion, and inspect every detected element directly on the rendered pages — bounding boxes, element types, content preview on click.

👉 Try the live demo: Docling Studio on Hugging Face Spaces

Two conversion modes:

• Local — Docling embedded, zero external dependency
• Remote — Docling Serve for heavier workloads

Stack: Vue 3 / TypeScript + FastAPI / Python, fully Dockerized (multi-arch image), 180+ tests, CI/CD.

Why it matters for the community

Docling is part of the Linux Foundation AI & Data ecosystem and powers a growing number of document processing workflows feeding into RAG pipelines. But without a visual feedback loop, it's hard to evaluate extraction quality, compare pipeline configs, or debug edge cases.

Docling Studio gives you that feedback loop. Especially useful for:

• Validating extraction before feeding chunks into your RAG pipeline
• Comparing pipeline configs — see how different Docling settings handle the same document
• Understanding extraction depth — you'll be surprised by what Docling actually captures

Try it

👉 github.com/scub-france/Docling-Studio

A Hugging Face Space is on the roadmap. Contributions and feedback welcome — if you're building RAG pipelines on top of Docling, what features would be most useful to you?