fix(pptx): keep soft line breaks with markdown and plain link formats

[Rainmemery]

Related Issues

• fixes PPTXToDocument drops soft line breaks when link_format is markdown or plain #13043

Proposed Changes:

With link_format="markdown" or "plain", _process_paragraph built its output from
paragraph.runs, which yields only the <a:r> children. A soft line break (<a:br/>,
Shift+Enter) standing between two runs was therefore dropped and the two sides were glued
together: a text box holding Account: Alice, a soft line break, and Balance: 100 was
converted into the single token Account: AliceBalance: 100, which retrieval cannot match.
The default link_format="none", which returns paragraph.text, already kept the break.

The link-formatting branch now walks the paragraph's children in document order instead of
just its runs, and emits each <a:br/> as the vertical tab python-pptx itself uses in
paragraph.text. All three link formats therefore return the same text, and hyperlinked
runs keep their existing formatting.

How did you test it?

Unit tests: three regressions in test/components/converters/test_pptx_to_document.py —

• test_soft_line_break_with_link_format — the report's exact repro, for both formats;
• test_soft_line_break_matches_link_format_none — the formatted output equals what
  link_format="none" already returns;
• test_hyperlink_still_formatted_next_to_soft_line_break — a break next to a hyperlinked
  run still yields [example](https://example.com) / example (https://example.com).

Red/green: the new tests fail on unpatched main (6 cases; the vertical tab is missing) and
pass with the change. The whole PPTX module is green (17 passed). A before/after run of
test/components/converters goes from 133 failed / 189 passed to 127 failed / 195 passed —
the difference is exactly these 6 cases, so nothing else moved; the remaining failures are
optional dependencies missing in my environment (pypdf, pdfminer, beautifulsoup) and are
identical before and after. ruff check and ruff format --check are clean on both files.

I could not run hatch run test:types: my sandbox cannot reach PyPI to install mypy. The new
code narrows paragraph._p.iterchildren() with isinstance against the annotated
pptx.oxml.text classes rather than casting, so it should type-check cleanly.

Notes for the reviewer

• The earlier attempt at this issue (fix(pptx): preserve soft line breaks with link formats #13073) was closed by the first-PR-limit automation, not
  on review grounds, and nothing has referenced the issue since.
• Only <a:r> and <a:br/> children are handled, matching the previous behaviour for every
  other paragraph child (fields were already dropped in the formatted modes).
• Transparency: this contribution was prepared with AI assistance, then reviewed, tested and
  understood by me before submitting.

Checklist

• I have read the contributors guidelines and the code of conduct.
• I have updated the related issue with new insights and changes.
• I have added unit tests and updated the docstrings.
• I've used a conventional commit type for my PR title: fix:.
• I have documented my code.
• I have added a release note file, following the contributors guidelines.
• I have run linters/formatters (ruff check, ruff format) and fixed any issue.

[vercel]

@Rainmemery is attempting to deploy a commit to the deepset Team on Vercel.

A member of the Team first needs to authorize it.

[github-actions]

Coverage report

Click to see where and how coverage changed

File	Statements	Missing	Coverage	Coverage (new stmts)	Lines missing
haystack/components/converters
pptx.py
haystack/core/pipeline
pipeline.py
Project Total

This report was generated by python-coverage-comment-action