Fix: query expander duplicate queries

[aquib8112]

Related Issues

• Fixes QueryExpander returns duplicate generated queries #12022

Proposed Changes:

QueryExpander could return duplicate queries when the chat generator repeated an expansion.

_parse_expanded_queries kept every non-empty string from the response, and the post-processing in run/run_async only checked whether the original query was already present. That check is a guard against appending the original twice, not deduplication, despite a comment claiming otherwise. Duplicates among the generated expansions were passed straight through.

Downstream, each duplicate makes MultiQueryTextRetriever and MultiQueryEmbeddingRetriever run another retrieval for the same query, adding latency and backend load while reducing the number of distinct expansions.

Generated queries are now deduplicated in _parse_expanded_queries, preserving first-seen order. The inaccurate comment was removed from both methods.

Two notes on the approach:

• Placement. The dedup lives in _parse_expanded_queries because both run and run_async already call it, so both paths are fixed by one change and cannot drift apart again. The method already strips whitespace and drops empty or non-string entries, so it normalizes the response rather than merely parsing it.
• Order. Deduplicating before truncation avoids spending the expansion budget on repeats. With n_expansions=3 and a response of ["a", "a", "b", "c"], truncating first yields ["a", "b"], while deduplicating first yields ["a", "b", "c"].

Deduplication can leave fewer than n_expansions queries. This matches the documented contract, since the class docstring already describes the output as "up to" N additional queries, and no existing test changes behaviour.

How did you test it?

Four unit tests added next to the existing test_run_query_deduplication, covering both the sync and async paths:

• duplicate expansions are removed
• deduplication happens before truncation (["a", "a", "b", "c"] with n_expansions=3 returns three queries, not two)

All four fail on main and pass with this change. Full file: 47 passed, 3 skipped.

Also verified manually that the reproduction snippet from the issue now returns {'queries': ['same query', 'other query']} on both run and run_async.

Notes for the reviewer

• Happy to move the logic into a shared _post_process_queries helper on the component instead, if you would prefer the parser to stay purely a parser. Kept the diff minimal here.

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 one of the conventional commit types for my PR title: fix:, feat:, build:, chore:, ci:, docs:, style:, refactor:, perf:, test: and added ! in case the PR includes breaking changes.
• I have documented my code.
• I have added a release note file, following the contributors guidelines.
• I have run pre-commit hooks and fixed any issue.

Disclaimer: This PR was assisted by AI. I have reviewed the changes and run the relevant tests.

[vercel]

@aquib8112 is attempting to deploy a commit to the deepset Team on Vercel.

A member of the Team first needs to authorize it.

[CLAassistant]

All committers have signed the CLA.

[github-actions]

Coverage report

Click to see where and how coverage changed

File	Statements	Missing	Coverage	Coverage (new stmts)	Lines missing
haystack/components/query
query_expander.py					298
haystack/tools
from_function.py
Project Total

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

[aquib8112]

Hey @sjrl, thank you for the feedback! I have removed the case-insensitivity logic and moved the async tests to TestQueryExpanderAsync with the @pytest.mark.asyncio marker.

[sjrl]

Thanks!