Fix Entity Source IDs Tracking ProblemDuring Relationship Processing

Problem

Missing source_id updates for entities created via relationships - In_merge_edges_then_upsert(), when processing relationships, existing entities weren't getting theirsource_id lists updated with chunks when source or target node missing in new relationships of current document.

Impact:

• Entity-chunk tracking became incomplete
• Miss deleting nodes or edges in documnent deleting process because of lack of chunk tracking information
• Rebuilding knowledge graph from chunks would miss some entity occurrences

Solution

When an existing entity is encountered during relationship processing, the code now:

1. Retrieves existing full source_ids fromentity_chunks_storage
2. Merges with new source_ids from the current relationship
3. Conditionally updatesentity_chunks_storage (only when data actually changes)
4. Applies source_ids limit usingapply_source_ids_limit()
5. Conditionally updates graph database and vector database (only when limited source_ids differ)

Testing Recommendations

1. Entity Source IDs: Process a document with relationships, verify entities get source_ids from both direct entity extraction and relationship processing
2. Deletion Order: Delete documents and verify no orphaned edges remain in graph database
3. Performance: Measure database operation count before/after on large datasets