Transcribe audio with a Parakeet endpoint

Upload an audio file to a Parakeet endpoint and get a transcription. The endpoint is OpenAI-compatible, so you send a multipart form request from curl or any OpenAI-compatible client.

Prerequisites

You need the following before you start:

A running Parakeet Managed Inference Job with a serving endpoint. See Create a Managed Inference Job (Parakeet).
Your API key, if the endpoint requires an authorization header. See Create an API key.
An audio file to transcribe.

Steps

Find your endpoint name

List your endpoints:

cosmicac models healthcheck

Each endpoint appears as Endpoint: <endpoint-name>. Copy the name of your Parakeet endpoint.

Get your inference URL

Copy the INFERENCE_APP_NODE_URL value:

cosmicac config show

Send the audio file

Replace <inference-url> with your INFERENCE_APP_NODE_URL, <endpoint-name> with the endpoint name you set, <api-key> with the key you created, and <audio-file> with the path to your file:

You need the API key only if you enabled Require Authorization header when you created the job. Otherwise, omit the Authorization header.

curl -X POST <inference-url>/v1/endpoints/<endpoint-name>/v1/audio/transcriptions \
  -H "Authorization: Bearer <api-key>" \
  -F "file=@<audio-file>" \
  -F "model=<endpoint-name>"

The endpoint name in the URL selects the model. The endpoint returns the transcription.