Verify source_ids field dependencies per PR #20 review feedback

[Copilot]

Addresses review comment r2637286239 from PR #20 requesting verification that no code depends on the source_ids field after its removal from test data.

Analysis Results

Confirmed the codebase correctly treats source_ids as a temporary internal field:

• BAML layer: source_ids tracks integer ID merges during entity resolution API calls
• Python ER (abzu/er/uuid.py): Explicitly removes source_ids from output (L402), converts to source_uuids for persistence
• Spark pipeline: No dependencies on source_ids found
• Design intent: source_uuids is the persistent provenance field; source_ids exists only within BAML processing scope

# Line 402 in abzu/er/uuid.py
# IMPORTANT: Do NOT include integer IDs (id, source_ids) in output
# These are internal to BAML processing and should not persist between iterations
resolved_dict: dict[str, Any] = {
    "uuid": master_uuid,
    "source_uuids": source_uuids_final,  # Persistent
    # source_ids intentionally omitted
}

Removal of source_ids from final_er.baml test data is correct—that function operates on UUIDs directly without integer ID mapping.

No code changes required. Implementation already handles this correctly.

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.