Read the live STT rate from GET /v1/catalog before you price a batch

Read the live Sume STT rate from GET https://api.sume.com/v1/catalog instead of hard-coding it. The catalog entry for sume/stt-1.0 carries model_pricing rows with a price and a unit; the docs say to confirm live prices there, because the numbers in prose can change. At the time of reading, the STT rate is $0.01 per audio minute, prorated by the second. The catalog is described in Core workflow and the price pointer is in Video inspect, read 2026-10-06.

Which numbers to trust

The catalog is the contract for rates and limits. An estimate you computed last month is a guess. For a spend estimate with your workspace's own terms, use the admission preview, which is read-only and returns the figure for your account.

Read it in code

The script looks for the sume/stt-1.0 row anywhere in the catalog and prints its price and unit. It makes no assumption about how many catalog entries there are.

import os, requests

H = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}


def stt_prices():
    r = requests.get("https://api.sume.com/v1/catalog", headers=H, timeout=30)
    r.raise_for_status()
    body = r.json()
    for item in body.get("data", []):
        for row in item.get("model_pricing") or []:
            if row.get("model") == "sume/stt-1.0":
                yield row.get("price"), row.get("unit")


for price, unit in stt_prices():
    print(price, unit)

Put the number in your estimate

Feed the price into the ffprobe folder estimate in place of the constant, and refresh it at the start of each run.