Could not parse multipart form in transcribe

[stevegpope]

Description

I get this exception from time to time when calling the transcription API. It happens on my real device and on virtual devices.
I found some additional information about it here: https://community.openai.com/t/calling-whisper-api-using-curl-request-keeps-giving-error/81510

Caused by io.ktor.client.plugins.ClientRequestException: Client request(POST https://api.openai.com/v1/audio/transcriptions) invalid: 400 . Text: "{
"error": {
"message": "Could not parse multipart form",
"type": "invalid_request_error",
"param": null,
"code": null
}
}
"
at io.ktor.client.plugins.DefaultResponseValidationKt$addDefaultResponseValidation$1$1.invokeSuspend(DefaultResponseValidation.kt:54)
at io.ktor.client.plugins.DefaultResponseValidationKt$addDefaultResponseValidation$1$1.invoke(:8)
at io.ktor.client.plugins.DefaultResponseValidationKt$addDefaultResponseValidation$1$1.invoke(:4)
at io.ktor.client.plugins.HttpCallValidator.validateResponse(HttpCallValidator.kt:51)
at io.ktor.client.plugins.HttpCallValidator.access$validateResponse(HttpCallValidator.kt:43)
at io.ktor.client.plugins.HttpCallValidator$Companion$install$3.invokeSuspend(HttpCallValidator.kt:152)
at kotlin.coroutines.jvm.internal.BaseContinuationImpl.resumeWith(ContinuationImpl.kt:33)
at kotlinx.coroutines.DispatchedTask.run(DispatchedTask.kt:106)
at android.os.Handler.handleCallback(Handler.java:942)
at android.os.Handler.dispatchMessage(Handler.java:99)
at android.os.Looper.loopOnce(Looper.java:201)
at android.os.Looper.loop(Looper.java:288)
at android.app.ActivityThread.main(ActivityThread.java:7872)
at java.lang.reflect.Method.invoke(Method.java)
at com.android.internal.os.RuntimeInit$MethodAndArgsCaller.run(RuntimeInit.java:548)
at com.android.internal.os.ZygoteInit.main(ZygoteInit.java:936)

Steps to Reproduce

1. POST to https://api.openai.com/v1/audio/transcriptions with a valid mp3 file
2. most of time it succeeds, sometimes not

Environment

• openai-kotlin version: com.aallam.openai:openai-client-bom:3.3.0
• Kotlin version: kotlin = "1.8.20"
• OS: Android

Additional Info

My code to transcribe in kotlin:

`
public suspend fun transcribeText(mp3Path: String): String {
val path = mp3Path.toPath()
val audioSource = FileSource(path = path, fileSystem = FileSystem.SYSTEM)

    val request = TranscriptionRequest(
        audio = audioSource,
        model = ModelId("whisper-1"),
    )

    var result = openAI.transcription(request)
    return result.text
}

`

[wilinz]

I have come across this issue before, where the problem lies in the unsupported form boundary string for the Content-Type header (e.g., okhttp's default boundary string is not supported). To resolve this, you need to change the boundary string to a different value. I haven't extensively studied how to change this value specifically in ktor, but that should be the general solution.

Here's an example of the Content-Type header with a form boundary string:

// WebAppBoundar is the form boundary string
Content-Type: multipart/form-data; boundary=WebAppBoundar