Bold list entries are not converted correctly

[sabotrax]

Bug

Using docling as the content extraction engine in open-webui (a llm web frontend).
docling is running in docker. docling has been installed as advised in https://docs.openwebui.com/features/document-extraction/docling/
when uploading md documents into openwebui knowledge (collection of documents the llm should be working on),
the original md document (excerpt):

- **Kategorie**: Rückläufe/ Abfälle HM-Material in der Produktion
- **Zustand**: PR trocken, Kleinmengen

will be markdown-rendered to:

-  [<RawText children='Kategorie'>]
-  [<RawText children='Zustand'>]

Subsequently, the llm is missing some of the information it should know about.
...

Steps to reproduce

I skipped open-webui and uploaded the document to the docling web ui running on port 5001 and processed the file with default settings. I also tested the different engines tesseract, easyocr, ocrmac, rapidocr and tesserocr with the same result.
The source document:

# Rücklaufmaterial - Pressrücklauf PR, Pressabfall PA

## Hauptinformationen
- **Kategorie**: Rückläufe/ Abfälle HM-Material in der Produktion
- **Zustand**: PR trocken, Kleinmengen

## Entsorgungsanweisung
Material nach HM-Sorten getrennt - Auswiegen - Eintrag in Übergabeliste an Übergabestation RU

## Behälter & Kennzeichnung
- **Sammelbehältnis**: Kleinmengen: Plastikeimer rot mit Deckel weiß verschlossen
- **Kennzeichnung**: Behälter mit HM-Sorte und RU-Art ("VR" oder "PR") kennzeichnen

## Übergabe
Übergabestation Schwerlastregal Bereich 300t-Presse Keine Direktentsorgung!

The output after processing in markdown:

# Rücklaufmaterial - Pressrücklauf PR, Pressabfall PA

## Hauptinformationen

- [<RawText children='Kategorie'>]
- [<RawText children='Zustand'>]

## Entsorgungsanweisung

Material nach HM-Sorten getrennt - Auswiegen - Eintrag in Übergabeliste an Übergabestation RU

## Behälter & Kennzeichnung

- [<RawText children='Sammelbehältnis'>]
- [<RawText children='Kennzeichnung'>]

## Übergabe

Übergabestation Schwerlastregal Bereich 300t-Presse Keine Direktentsorgung!

ID: 1

...

Docling version

docling version: 2.30.0
...

Python version

Python 3.12.10
...

[pyaaaa]

我也碰到了这个问题，

[bverhoeve1]

I am experiencing a similar issue. Example markdown input:

# My Heading 1
Some text here

## My Subheading 1: Some Heading
1. ***Some List Item.*** This text seems to disappear 
1. ***Some other list item*** this also disappears 

## My Subheading 2: Some Other Heading

Calling export_to_dict() on the result gives json that looks like this (snippet below) - it seems to drop everything after the bold ends.

        {
            "self_ref": "#/texts/3",
            "parent": {
                "$ref": "#/groups/0"
            },
            "children": [],
            "content_layer": "body",
            "label": "list_item",
            "prov": [],
            "orig": "[<StrongEmphasis children=[<RawText children='Some List Item.'>]>]",
            "text": "[<StrongEmphasis children=[<RawText children='Some List Item.'>]>]",
            "enumerated": false,
            "marker": "-"
        },
        {
            "self_ref": "#/texts/4",
            "parent": {
                "$ref": "#/groups/0"
            },
            "children": [],
            "content_layer": "body",
            "label": "list_item",
            "prov": [],
            "orig": "[<StrongEmphasis children=[<RawText children='Some other list item'>]>]",
            "text": "[<StrongEmphasis children=[<RawText children='Some other list item'>]>]",
            "enumerated": false,
            "marker": "-"
        },

[vagenas]

Relevant PRs:

• feat(markdown): add formatting & improve inline support #1804 (parsing of formatted MD — most points)
• feat(markdown): add heading formatting, fix code & formula formatting docling-core#336 (serialization of rich headings / code / formulas)
• fix(markdown): fix single-formatted headings & list items #1820 (parsing of single-formatted list items / headings)