fund-tracer/backend/app/rules/dedup_rules.py

"""Transaction deduplication rules.

Only marks records as duplicate when there is strong deterministic evidence.
Highly similar records (same amount/time/counterparty) are intentionally kept
for manual review to avoid filtering out potential scam brushing transactions.
"""

from app.models.transaction import TransactionRecord


def is_duplicate_pair(a: TransactionRecord, b: TransactionRecord) -> bool:
    # Rule 1: exact order_no match (strong deterministic signal).
    if a.order_no and b.order_no and a.order_no == b.order_no:
        return True

    # Intentionally do NOT deduplicate by amount/time similarity.
    # Those records should enter the review stage for human confirmation.
    return False