CLIENTS: Send Audio Transcription

[Catalin-Andronie]

Closes #212

[hassanhabib]

@Catalin-Andronie post a screenshot of this client running please

[Catalin-Andronie]

@Catalin-Andronie post a screenshot of this client running please

@hassanhabib Unfortunately this will not work as expected because of the next RESTFulSense missing features:

1. MEDIUM FIX: Enable creation of multipart-form with non-string types RESTFulSense#132
2. FOUNDATIONS: Allow nullable value properties to be skipped or be part on the multipart-form RESTFulSense#130

---

1. Without MEDIUM FIX: Enable creation of multipart-form with non-string types RESTFulSense#132 implemented we are forced to mark all AudioTranscriptionRequest properties as string. E.g: Temperature will become a string type instead of a double as it should be.

public class AudioTranscriptionRequest
{
-    public double Temperature { get; set; } = 0.2;
+    public string Temperature { get; set; } = "0.2";
}

2. Without FOUNDATIONS: Allow nullable value properties to be skipped or be part on the multipart-form RESTFulSense#130 implemented we are forced to set all the AudioTranscriptionRequest properties with some value even if some of them are optional. For the creation of a translation, OpenAI requires only the file and model to be present in the request body and the rest of the options are optional, and if we do so we are getting an exception for the other properties since they are null or have empty values.

var inputAudioTranscription = new AudioTranscription
{
    Request = new AudioTranscriptionRequest
    {
        Content = fileContent,
        FileName = fileName,
        Model = "whisper-1",
-       Prompt = null, // This is an optional value and I should not be forced to set its value.
+       Prompt = "Some prompt...",
-       Language = "", // This will throw an exception since RESTFulSense doesn't accept empty values.
+       Language = "en"
    }
};

AudioTranscription responseAudioTranscription =
    await this.openAIClient.AudioTranscriptions.SendAudioTranscriptionAsync(
        inputAudioTranscription);

@hassanhabib what's your suggestion?

[BrianLParker]

Reading the API the file in the request is a string. The file should be previously uploaded using the name (string) for this. RESTFulSense already supports this using the PostContentAsync method. If you look at the example on the API page they are only posting the name of the file "german.m4a". Sorry for the late response been quite ill for the last 4 days.

[Catalin-Andronie]

Reading the API the file in the request is a string. The file should be previously uploaded using the name (string) for this. RESTFulSense already supports this using the PostContentAsync method. If you look at the example on the API page they are only posting the name of the file "german.m4a". Sorry for the late response been quite ill for the last 4 days.

The API only accepts multipart/form-data content-type, which means we cannot use PostContentAsync since that uses "json" content-type (See example request).

curl https://api.openai.com/v1/audio/transcriptions \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: multipart/form-data" \
  -F file="@/path/to/file/audio.mp3" \
  -F model="whisper-1"

Actually, the API specifies to post the file itself as a stream of bytes, the example only illustrates what it should be there. Also, besides the file and model the API accepts other properties in the request body like temperature, language, and prompt (the latter 3 are optional).

Currently, we are using PostFormAsync to send the file stream to the server and we are obtaining its transcription, which works like a charm. Even so, we have two issues/features which need to be implemented before moving forward and completing the Audio Transcription. Please read this comment to understand the RESTFulSense features.