usage.cost on a Sume video poll is nullable: guard it before you sum

The usage.cost field on a /v1/videos poll is nullable in the OpenAPI schema, so a ledger that does total += poll["usage"]["cost"] will crash or add zero for a job that has no figure yet. Sum cost only for completed jobs and keep the ids with no cost in their own list.

What the field is

Sume reserves the provider list price times 1.25 at submit, and usage.cost is the billable amount for the job. Your own estimate from pricing_skus is a forecast; the poll is the receipt.

Treat the two as separate columns. When they differ, the receipt wins and the estimate needs fixing.

A ledger that cannot lie

Decimal(str(cost)) avoids float noise when the value arrives as a JSON number. Jobs that are not completed, or that report no cost, go to the unknown list so a review step can look at them.

from decimal import Decimal

ledger = {"total": Decimal("0"), "unknown": []}

def add(poll):
    if poll["status"] != "completed":
        return
    cost = (poll.get("usage") or {}).get("cost")
    if cost is None:
        ledger["unknown"].append(poll["id"])
    else:
        ledger["total"] += Decimal(str(cost))

# after polling every job to a terminal status:
for poll in finished_polls:
    add(poll)
print(ledger["total"], len(ledger["unknown"]))

Gotchas

The poll body is a bare object, not a {data} envelope, so read poll["usage"] directly. Failed and cancelled jobs are skipped here on purpose; what a failed job was billed is a question for your invoice and the docs on credits, not for this sum.

Store the cost with the job id the moment you see it. Re-polling later costs a read, which is cheap on the 4,800 per minute Free read budget, but a missing row is hard to rebuild.