Starstruk Logo MarkStarstruk

v2.118.0

docling-project/doclingv2.118.0Aug 3, 2026by docling-ops[bot]

AI Summary

A major feature release adding support for EBCDIC documents and agent skills. It significantly improves parsing accuracy for Word documents, PowerPoint presentations, and PDF outlines through various backend refactors and fixes.

Key Highlights

  • Add the ebcdic backend for new document support
  • Add agent skills for using docling
  • Refactor RapidOcrModel to support all PP-OCR languages
  • Fix for DocX section header/footer content dropping
  • Fix for DocX ordered-list numbering after intervening lists

New Features

  • EBCDIC backend
  • Expose PDF heading-level inference in service API
  • Agent skills
  • RapidOcrModel refactor

Full Release Notes

### Feature

* Add the ebcdic backend ([#3926](https://github.com/docling-project/docling/issues/3926)) ([`9bdb2af`](https://github.com/docling-project/docling/commit/9bdb2af94e93cab47fe96105850438eb663d7953))
* **service datamodel:** Expose PDF heading-level inference in the service API ([#3874](https://github.com/docling-project/docling/issues/3874)) ([`a47f76a`](https://github.com/docling-project/docling/commit/a47f76a871b678321b54c8aa37a7d1490d6e4849))
* **skills:** Add agent skills for using docling ([#3913](https://github.com/docling-project/docling/issues/3913)) ([`5dc8c19`](https://github.com/docling-project/docling/commit/5dc8c1989c8e14323b723b20b9287bf32b5c90ca))
* Refactor RapidOcrModel to support all PP-OCR languages via version/backbone resolution ([#3863](https://github.com/docling-project/docling/issues/3863)) ([`f7eb9b4`](https://github.com/docling-project/docling/commit/f7eb9b4e32daf32fc0f0b60600e4dde239efa106))

### Fix

* **docx:** Stop dropping section header/footer content in Docx ([#3843](https://github.com/docling-project/docling/issues/3843)) ([`2ec4f64`](https://github.com/docling-project/docling/commit/2ec4f64b073018c1a1132639029e49fe9021cff8))
* **docx:** Resume ordered-list numbering after an intervening list (#3896) ([#3897](https://github.com/docling-project/docling/issues/3897)) ([`62f6a51`](https://github.com/docling-project/docling/commit/62f6a51694c8e41d690001716c06bbb289a0fdbf))
* **docx:** Keep body text after a blank spacer inside a list (#3898) ([#3902](https://github.com/docling-project/docling/issues/3902)) ([`7883aaa`](https://github.com/docling-project/docling/commit/7883aaade00981b49a363db29c2da31c80938a10))
* **pdf:** Rewind stream inputs in docling-parse ([#3895](https://github.com/docling-project/docling/issues/3895)) ([`995ae2a`](https://github.com/docling-project/docling/commit/995ae2a452b2c804d318f2990eecb2bc69185474))
* **odf:** Preserve text after inline elements ([#3850](https://github.com/docling-project/docling/issues/3850)) ([`955d579`](https://github.com/docling-project/docling/commit/955d5797a0211852b66ff4351d380c5f8d608804))
* **reading-order:** Dehyphenate hard continuations ([#3888](https://github.com/docling-project/docling/issues/3888)) ([`52d8a6f`](https://github.com/docling-project/docling/commit/52d8a6f24de7318a9ad4be2a7361ba93fc81a5c1))
* **pptx:** Process slide shapes in visual reading order ([#3393](https://github.com/docling-project/docling/issues/3393)) ([`91fa745`](https://github.com/docling-project/docling/commit/91fa745b3228fa0df0510d76eb94956b063054e1))