v2.109.0

docling-project/doclingv2.109.0Jul 3, 2026by docling-ops[bot]

AI Summary

This release enhances PDF parsing by inferring heading levels from bookmarks and introduces new OCR defaults, while improving text extraction from HTML and Office formats like Strict OOXML.

Key Highlights

  • Infer heading levels from PDF bookmarks and ToC
  • Use PP-OCRv6 RapidOCR defaults
  • Expose native Whisper decoding options
  • Support Strict OOXML (.docx) files

New Features

  • Allow overriding same-font enforcement
  • Use PP-OCRv6 RapidOCR defaults
  • Expose native Whisper decoding options
  • Add code language detection
  • Infer heading levels from PDF bookmarks/ToC
  • Parse tables nested inside list items

Full Release Notes

### Feature

* Allow overriding same-font enforcement ([#3737](https://github.com/docling-project/docling/issues/3737)) ([`0064a45`](https://github.com/docling-project/docling/commit/0064a4558bf820c1af0ef425fe0dbf43036ac7cc))
* Use PP-OCRv6 RapidOCR defaults ([#3677](https://github.com/docling-project/docling/issues/3677)) ([`01235ce`](https://github.com/docling-project/docling/commit/01235ceef710d822dc5ad1c4173e60990f1d1e9f))
* **asr:** Expose and forward native Whisper decoding options ([#3717](https://github.com/docling-project/docling/issues/3717)) ([`4f37eee`](https://github.com/docling-project/docling/commit/4f37eeee07a1741526671eff5329b3cca2e685b8))
* Add code language detection ([#3721](https://github.com/docling-project/docling/issues/3721)) ([`648fcf1`](https://github.com/docling-project/docling/commit/648fcf15cd42b4b7500ba0578a559cf848d37bee))
* **pdf:** Infer heading levels from PDF bookmarks/ToC ([#3688](https://github.com/docling-project/docling/issues/3688)) ([`9cbef42`](https://github.com/docling-project/docling/commit/9cbef428c63cf42b0fada976e07dbcf62e0eab88))

### Fix

* **html:** Parse tables nested inside list items (#3508) ([#3738](https://github.com/docling-project/docling/issues/3738)) ([`8c162bd`](https://github.com/docling-project/docling/commit/8c162bd4c70055ba2b959e6aea17435b1b948172))
* Forward VLM convert runtime generation settings ([#3322](https://github.com/docling-project/docling/issues/3322)) ([`1c03fd4`](https://github.com/docling-project/docling/commit/1c03fd47b22c80deb1f2c9359614e891221c4045))
* **tests:** Correct XBRL source path to point to 'sources/' subdirectory ([#3734](https://github.com/docling-project/docling/issues/3734)) ([`b77612c`](https://github.com/docling-project/docling/commit/b77612ce4770d052aa90af23681b55ab0be38971))
* **docx:** Support Strict OOXML (.docx) files ([#3729](https://github.com/docling-project/docling/issues/3729)) ([`528eb5b`](https://github.com/docling-project/docling/commit/528eb5b4b7bbb5b8253df0cf130feb24c9fd1fc4))
* **msword:** Don't abort conversion on a malformed hyperlink address ([#3744](https://github.com/docling-project/docling/issues/3744)) ([`c979c6a`](https://github.com/docling-project/docling/commit/c979c6a6bca2568ab7a0e392a87b46e0bf6d2f8b))

### Documentation

* **asr:** Document backend selection and WhisperS2T ([#3730](https://github.com/docling-project/docling/issues/3730)) ([`e0b1ca5`](https://github.com/docling-project/docling/commit/e0b1ca5964b11fb71f7e747b75ad4be8500d8f39))