fix: distinguish empty document content IDs

[daleselaji-dev]

Problem

Document(content="") and Document(content=None) currently receive the same automatic ID. This can raise a misleading duplicate error for a same-batch write and can silently violate ID uniqueness across separate writes.

Root Cause

Document._create_id() normalizes self.content with self.content or None, collapsing the valid empty string into None before hashing.

Solution

Hash the validated content value directly so an empty string remains distinct from None.

Changes

• Added a focused regression test proving empty-string and None content produce different IDs.
• Added a release note for the automatic-ID behavior correction.

Testing

• Baseline: PYTHONUTF8=1 PYTHONIOENCODING=utf-8 uvx hatch run test:unit test/dataclasses/test_document.py — 38 passed.
• Focused: same command after the change — 39 passed.
• Risk-matched: PYTHONUTF8=1 PYTHONIOENCODING=utf-8 uvx hatch run test:unit test/document_stores/test_in_memory.py — 370 passed, 8 skipped.
• Format/lint: PYTHONUTF8=1 PYTHONIOENCODING=utf-8 uvx hatch run fmt-check haystack/dataclasses/document.py test/dataclasses/test_document.py — passed.
• Typecheck: PYTHONUTF8=1 PYTHONIOENCODING=utf-8 uvx hatch run test:types haystack/dataclasses/document.py test/dataclasses/test_document.py — passed.
• git diff --check — passed.
• Full test matrix was not run locally.

Compatibility/Risk

Automatic IDs for documents whose content is exactly "" will change. IDs for None content and other content values remain unchanged. Persisted stores containing empty-string documents may need an ID migration or re-write.

Notes for Reviewer

The implementation intentionally changes only the falsey-content normalization in _create_id; metadata and other ID inputs retain their existing serialization behavior. The change was verified on the current main snapshot and includes AI-assisted development disclosure per the repository contribution guide.

Linked Issue

Fixes #13234

[vercel]

@daleselaji-dev is attempting to deploy a commit to the deepset Team on Vercel.

A member of the Team first needs to authorize it.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[HaystackBot]

Hi @daleselaji-dev, thanks a lot for your contribution! 🙏

We noticed that the Contributor License Agreement (CLA) check (license/cla) hasn't passed yet, so we've temporarily moved this PR to draft and paused the review assignment.

To get your PR reviewed, please sign the CLA via the link in the license/cla check below (or in the CLA bot comment). As soon as the check turns green, this PR will automatically be marked ready for review again and a reviewer will be re-assigned.

[sjrl]

Thanks for the PR! We've decided not to change this behavior since it would change the auto-generated IDs of existing documents with empty content. See #13234 for details. Closing.