Fix: Remove Duplicate Entity/Realtion Tracking Deletion in adelete_by_doc_id

[danielaskdd]

🐛 Fix: Remove Duplicate Entity/Realtion Tracking Deletion in adelete_by_doc_id

Problem

In the adelete_by_doc_id function, there were two separate calls to relation_chunks.delete that could potentially delete the same relation chunk tracking records:

1. First deletion (around line 2656): During batch chunk tracking updates, relations with no remaining chunks were deleted
2. Second deletion (around line 2729): When deleting relationship graph edges, corresponding chunk tracking records were deleted again

While this didn't cause errors (delete operations are idempotent), it resulted in:

• Unnecessary performance overhead
• Code redundancy and reduced maintainability
• Unclear separation of concerns

Solution

Refactored the chunk tracking update logic to eliminate duplicate deletions:

Changes in the first section (Chunk Tracking Updates):

• Removed the relation_delete_ids set and its associated deletion logic
• Modified the loop to skip empty relations using continue instead of deleting them
• Added clear comment: "Empty relations are deleted alongside graph edges later"
• Now only performs upsert operations for relations with remaining chunks

Preserved the second section (Graph Element Deletion):

• Kept this as the single source of truth for deleting relation chunk tracking records
• Ensures consistent deletion across all storage layers (chunk tracking, vector DB, graph DB)

Note: The same optimization pattern was applied to entity_chunks for code consistency.

Benefits

• ✅ Eliminates redundant operations: No duplicate deletion attempts
• ✅ Improves performance: Reduces unnecessary async operations
• ✅ Clearer code logic: Single responsibility - updates in one place, deletions in another
• ✅ Better maintainability: Clear comments explain the design rationale
• ✅ Consistent pattern: Applied the same optimization to entity_chunks for consistency

Testing Recommendations

1. Test document deletion with relations that:
   • Are fully deleted (all chunks removed)
   • Are partially affected (some chunks remain)
   • Have no chunk references
2. Verify chunk tracking storage consistency after deletion
3. Confirm no regression in deletion functionality

Impact

• Scope: Document deletion workflow (adelete_by_doc_id)
• Risk: Low - logic simplification with same functional outcome
• Backward compatibility: ✅ Fully compatible - no API changes

[danielaskdd]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. 👍

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".