Sume STT metadata: tag a transcript job with your own ids

Put your own ids in the metadata object of the Sume STT request. The OpenAPI spec describes it as optional caller metadata stored with the Sume job request, and not sent to the provider. That makes it the right place for an episode id, a ticket number or a batch label, and the wrong place for anything you expect the transcription model to see.

It is useful because job ids come from Sume while the records you care about come from you. Metadata is how the two meet.

What does the spec guarantee?

In the Sume OpenAPI spec the STT request lists metadata as optional, stored with the job request, and not forwarded to the provider. The spec does not promise how it is echoed on every read, so confirm on a test job where it appears before you build a reconciliation on it.

What should you put in it?

Keep it small and boring. Use identifiers, not content.

• Your own record id, such as an episode or call id.
• A batch label so you can find a run later.
• A schema version for your own parser.
• Not secrets and not personal data you would not want stored.

How do you send it?

A submit call with tags:

import os, requests

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

r = requests.post("https://api.sume.com/v1/stt-1.0/transcribe",
    headers=H, timeout=60,
    json={
        "audio_url": os.environ["AUDIO_URL"],
        "duration_seconds": 240,
        "metadata": {"episode_id": "ep-0142", "batch": "oct-archive"},
    })
r.raise_for_status()
print(r.json()["data"]["job"]["id"])

Is there a related tag for end users?

Yes, the same idea applies across Sume jobs; see tagging voice and image jobs with an end-user id. For batches of recordings, pair it with the pattern in a podcast back catalog.