AI dashboard data check: the audit prompt and a practice kit for the duplicate order trap

import csv, sys
from collections import Counter

path = sys.argv[1]
write = "--write" in sys.argv
rows = list(csv.DictReader(open(path, newline="")))

# 1. Unit of a row and assumptions, stated before any number.
print("Assumption: each row is one complete order (not a line item).")
currencies = {r["currency"] for r in rows}
dates = sorted(r["order_date"] for r in rows)
print(f"Currency: {', '.join(sorted(currencies))}. Date window: {dates[0]} to {dates[-1]}.")
if len(currencies) != 1:
    sys.exit("STOP: more than one currency. Never merge or guess exchange rates.")

# 2. Parse amounts; fail loudly on blanks or text.
for i, r in enumerate(rows, start=2):
    try:
        r["amount"] = float(r["amount"])
    except ValueError:
        sys.exit(f"STOP: row {i} amount is not a number: {r['amount']!r}")

original_total = sum(r["amount"] for r in rows)
print(f"Rows: {len(rows)}. Original total: {currencies.pop()} {original_total:,.0f}.")

# 3. Repeated IDs: exact duplicates vs conflicting records.
by_id = Counter(r["order_id"] for r in rows)
exact, conflicting = [], []
seen = {}
for i, r in enumerate(rows, start=2):
    key = tuple(r.values())
    if r["order_id"] in seen:
        (exact if seen[r["order_id"]] == key else conflicting).append((i, r))
    else:
        seen[r["order_id"]] = key
for i, r in exact:
    print(f"EXACT DUPLICATE: row {i}, {r['order_id']}, amount {r['amount']:,.0f} (identical in every field).")
for i, r in conflicting:
    print(f"CONFLICT: row {i}, {r['order_id']} repeats with different fields. Review by hand; do not delete.")
print(f"Repeated IDs: {[k for k, n in by_id.items() if n > 1] or 'none'}.")

# 4. Reconcile. Only exact duplicates are removed, and only on --write.
removed = sum(r["amount"] for _, r in exact)
kept = [r for i, r in enumerate(rows, start=2) if i not in {i for i, _ in exact}]
cleaned_total = sum(r["amount"] for r in kept)
print(f"Reconcile: {original_total:,.0f} - {removed:,.0f} = {cleaned_total:,.0f} "
      f"({'OK' if original_total - removed == cleaned_total else 'MISMATCH'}). Clean rows: {len(kept)}.")

if write and exact and not conflicting:
    with open("cleaned-orders.csv", "w", newline="") as f:
        w = csv.DictWriter(f, fieldnames=rows[0].keys())
        w.writeheader()
        w.writerows(kept)
    print("Wrote cleaned-orders.csv. The original was not changed.")