[FEATURE REQ] Expose used tokens on transcription

[bw-flagship]

Describe the feature or improvement you are requesting

I use AudioClient.TranscribeAudioAsync and would like to know how many tokens I consumed.

The API itself returns this info:

{
  "text": "some text.",
  "usage": {
    "type": "tokens",
    "input_tokens": 14,
    "input_token_details": {
      "text_tokens": 0,
      "audio_tokens": 14
    },
    "output_tokens": 45,
    "total_tokens": 59
  }
}

See https://platform.openai.com/docs/api-reference/audio/createTranscription

But the library only returns ClientResult<AudioTranscription> which does not contain information about the used tokens.

Additional context

Other methods (e.g. in the responses api) return the information about used tokens.

[jsquire]

Hi @bw-flagship. Thanks for reaching out and for your inquiry. The usage structure is a is not yet represented in the REST API spec used for client generation. Once the REST API spec is updated to include this area and we're able to ingest, we'll include it in the client.

In the meantime, you can access the necessary additional properties by reading the raw JSON data. For example:

var client = new AudioClient ("whisper-1","<< YOUR API KEY >>");
var result = await client.TranscribeAudioAsync("<< SOME FILE PATH >>");
var json = result.GetRawResponse().Content;

// Read usage properties
var responseWithUsage = JsonSerializer.Deserialize<ResponseUsage>(json);



public record ResponseUsage(Usage usage);

public record Usage(
    string type,
    int input_tokens,
    int output_tokens,
    int total_tokens,
    InputDetails input_token_details);
    
public record InputDetails(
    int text_tokens,
    int audio_tokens);