| --- |
| license: other |
| license_name: mixed-upstream-licenses |
| license_link: https://huggingface.co/superlinked/docling-artifacts/blob/main/SOURCES.json |
| library_name: docling |
| tags: |
| - docling |
| - document-ai |
| - ocr |
| --- |
| |
| # Superlinked Docling artifacts |
|
|
| This repository is an immutable composite artifact root for Docling's standard |
| PDF pipeline. It contains only the assets exercised by Superlinked's current |
| `docling` default and OCR runtime profiles: |
|
|
| - Docling Layout Heron; |
| - TableFormer in accurate mode; and |
| - RapidOCR PP-OCRv4 Chinese ONNX detection, classification, and recognition. |
|
|
| The directory layout is the one accepted by Docling's `artifacts_path`. The |
| snapshot was assembled and validated with Docling 2.99.0. It does not contain |
| Docling's Python distribution or any Superlinked serving code. |
|
|
| The RapidOCR visualization font is deliberately omitted. It is not required by |
| the validated inference path; consumers that render RapidOCR visualization |
| output must supply a suitably licensed font separately. |
|
|
| See `SOURCES.json` for the exact upstream revisions/releases, per-file sizes and |
| SHA-256 digests, and license evidence. Each upstream component retains its own |
| license; this composite repository does not relicense those assets. |
|
|
| For SIE, pin the resulting 40-character repository commit as the model's |
| `hf_revision`. `DoclingAdapter` resolves the snapshot and passes its root to |
| Docling without a model-specific staging path. |
|
|
