fix(search): wrap group_ids OR clause in parentheses for BM25 queries

[giulio-leone]

Summary

Fixes #1249

When searching with multiple group_ids, the BM25 fulltext query filter was missing parentheses around the OR clause. Due to Lucene operator precedence (AND binds tighter than OR), only the last group_id was effectively filtered.

Root Cause

# Before (buggy):
group_id:"g1" OR group_id:"g2" OR group_id:"g3" AND (hello)
# Parsed as: g1 OR g2 OR (g3 AND hello)  ← only g3 is filtered!

# After (fixed):
(group_id:"g1" OR group_id:"g2" OR group_id:"g3") AND (hello)
# Parsed correctly: all group_ids are filtered

Changes

• graphiti_core/search/search_utils.py (fulltext_query): Wrap the group_ids OR filter in parentheses before appending AND
• graphiti_core/driver/neo4j/operations/search_ops.py (_build_neo4j_fulltext_query): Same parenthesisation fix for the Neo4j-specific fulltext query builder
• tests/utils/search/search_utils_test.py: Added regression tests covering single group_id, multiple group_ids, no group_ids, and empty group_ids list

[danielchalef]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[giulio-leone]

I have read the CLA Document and I hereby sign the CLA

[giulio-leone]

Friendly ping — CI is green and this is ready for review. Happy to address any feedback. Thanks!

[Copilot]

Pull request overview

This PR aims to fix Lucene operator-precedence issues in BM25 fulltext queries when filtering by multiple group_ids, by ensuring the OR’d group_id clause is parenthesized before appending the main query with AND.

Changes:

• Wrap the group_id:"..." OR ... filter clause in parentheses when building Neo4j fulltext (Lucene) queries.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[giulio-leone]

All CI checks pass. Fixes BM25 query precedence bug with group_ids. Includes 7 regression tests. Ready for review.

[giulio-leone]

Hi! Gentle ping — this PR is rebased, CI passes, and ready for review. Happy to address any feedback. Thanks!

[giulio-leone]

Closing in favor of #1311 which addresses the same BM25 fulltext query parentheses bug with an identical fix. Keeping one PR to avoid duplicate review burden.