Uploaded decisions

decisions.create builds the multipart upload automatically. Pass audio, state, and questions as keyword arguments. model defaults to tazo-large.

Audio input	Behavior
String or `Path`	SDK opens and closes the file.
Bytes	Pass `filename` and `content_type` when known.
Binary file object	Read from its current position; never closed by the SDK.

Seekable streams are restored to their original position. Nonseekable streams are consumed once. The SDK buffers one immutable upload of at most 16 MiB so every retry uses identical audio, JSON, and MIME type. Model duration limits still apply. The SDK imposes no playground word-count restrictions. Questions can be dictionaries, as in the quickstart, or typed models:

from tazo import ChoiceQuestion, NoulQuestion

choice = ChoiceQuestion(
    instructions="Which action should be taken?",
    criteria={"get_weather": "Get weather.", "NO_TOOL": "No tool needed."},
)
yes_no = NoulQuestion(instructions="Is the speaker asking for help?")

Tazo Large supports one choice or noul. Tazo Small is unavailable through the API; contact support@tazolabs.com for assistance. Check client.models.list() for current capabilities. This SDK does not expose score questions. Each create call generates a new operation key. Supply idempotency_key to resume an earlier operation with its original input. The SDK does not expose a job-status API. Read retries and pending work.

Get started

Guides

Python reference

Uploaded decisions