fix: allow AutoMergingRetriever to pass through parentless matched documents - #13042

@0xamlab is attempting to deploy a commit to the deepset Team on Vercel.

A member of the Team first needs to authorize it.

All committers have signed the CLA.

Hi @0xamlab, 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.

I independently reproduced #13029 and prepared a local fix before seeing this PR. The production change here matches the minimal fix I tested: remove the parent-ID truthiness guard while retaining the __level and __block_size presence checks. I won't open a competing PR.

One test-coverage suggestion: the new mixed parentless/leaf tests describe verifying that merging still works, but each passes only leaf1 from a parent with two children at threshold=0.6. The ratio is 1/2, so no merge occurs; the assertions verify leaf/root pass-through instead. Could you also add paired sync/async cases passing both siblings alongside an unrelated parentless root, and assert the parent replaces the leaves while the root is preserved?

Additional edge cases I covered locally are real HierarchicalDocumentSplitter roots with absent, null, and empty-string __parent_id, both alone and mixed with leaves. I also retained mixed-valid/invalid input validation coverage by making the invalid document lack __level rather than deleting that behavior test entirely.

My local retriever and hierarchical-splitter suites pass together: 62 tests. That result is for my local patch, not this PR's checkout; I have reviewed your diff but have not run your branch's suite. This is community feedback, not a maintainer approval. The investigation and tests were prepared with Perplexity Computer AI assistance.

Thanks @saiprasanth-git — those were exactly the coverage gaps. I added sync + async cases where both siblings merge into the parent while a parentless root is preserved, real HierarchicalDocumentSplitter roots with absent/None/"" __parent_id (alone and mixed with leaves), and a mixed valid/invalid case that keeps the __level validation pinned. All 39 retriever tests pass and hatch run fmt is clean.

Thanks for signing the CLA, @0xamlab! 🎉 This PR is now ready for review again and the reviewer has been re-assigned.

Thanks for addressing the requested cases. The added sync/async coverage, splitter-root cases, and mixed-validity validation test cover the gaps I identified. Looks good from my side.