rig-core 0.40.0

use bytes::Bytes;

use crate::http_client::multipart::Part;
use crate::http_client::{HttpClientExt, MultipartForm};
use crate::providers::openai::{Client, client::ApiResponse};
use crate::transcription;
use crate::transcription::TranscriptionError;
use serde::Deserialize;

// ================================================================
// OpenAI Transcription API
// ================================================================

pub const WHISPER_1: &str = "whisper-1";

#[derive(Debug, Deserialize)]
pub struct TranscriptionResponse {
    pub text: String,
}

impl TryFrom<TranscriptionResponse>
    for transcription::TranscriptionResponse<TranscriptionResponse>
{
    type Error = TranscriptionError;

    fn try_from(value: TranscriptionResponse) -> Result<Self, Self::Error> {
        Ok(transcription::TranscriptionResponse {
            text: value.text.clone(),
            response: value,
        })
    }
}

#[derive(Clone)]
pub struct TranscriptionModel<T = reqwest::Client> {
    client: Client<T>,
    pub model: String,
}

impl<T> TranscriptionModel<T> {
    pub fn new(client: Client<T>, model: impl Into<String>) -> Self {
        Self {
            client,
            model: model.into(),
        }
    }
}

impl<T> transcription::TranscriptionModel for TranscriptionModel<T>
where
    T: HttpClientExt + Clone + std::fmt::Debug + Default + Send + 'static,
{
    type Response = TranscriptionResponse;

    type Client = Client<T>;

    fn make(client: &Self::Client, model: impl Into<String>) -> Self {
        Self::new(client.clone(), model)
    }

    async fn transcription(
        &self,
        request: transcription::TranscriptionRequest,
    ) -> Result<
        transcription::TranscriptionResponse<Self::Response>,
        transcription::TranscriptionError,
    > {
        let data = request.data;

        let mut body = MultipartForm::new()
            .text("model", self.model.clone())
            .part(Part::bytes("file", data).filename(request.filename.clone()));

        if let Some(language) = request.language {
            body = body.text("language", language);
        }

        if let Some(prompt) = request.prompt {
            body = body.text("prompt", prompt.clone());
        }

        if let Some(ref temperature) = request.temperature {
            body = body.text("temperature", temperature.to_string());
        }

        if let Some(ref additional_params) = request.additional_params {
            let params = additional_params.as_object().ok_or_else(|| {
                TranscriptionError::RequestError(Box::new(std::io::Error::new(
                    std::io::ErrorKind::InvalidInput,
                    "additional transcription parameters must be a JSON object",
                )))
            })?;

            for (key, value) in params {
                body = body.text(key.to_owned(), value.to_string());
            }
        }

        let req = self
            .client
            .post("/audio/transcriptions")?
            .body(body)
            .map_err(|e| TranscriptionError::HttpError(e.into()))?;

        let response = self.client.send_multipart::<Bytes>(req).await?;

        let status = response.status();
        let response_body = response.into_body().into_future().await?.to_vec();
        if status.is_success() {
            match serde_json::from_slice::<ApiResponse<TranscriptionResponse>>(&response_body)? {
                ApiResponse::Ok(response) => response.try_into(),
                ApiResponse::Err(api_error_response) => {
                    tracing::warn!(message = %api_error_response.message, "provider returned an error response");
                    Err(TranscriptionError::from_http_response(
                        status,
                        String::from_utf8_lossy(&response_body).into_owned(),
                    ))
                }
            }
        } else {
            let str = String::from_utf8_lossy(&response_body).to_string();
            Err(TranscriptionError::from_http_response(status, str))
        }
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::client::transcription::TranscriptionClient;
    use crate::test_utils::RecordingHttpClient;
    use crate::transcription::TranscriptionModel as _;

    #[tokio::test]
    async fn transcription_http_non_success_preserves_status_and_body() {
        let body = r#"{"error":{"message":"bad audio","type":"invalid_request_error"}}"#;
        let http_client =
            RecordingHttpClient::with_error_response(http::StatusCode::BAD_REQUEST, body);
        let client = Client::builder()
            .api_key("test-key")
            .http_client(http_client)
            .build()
            .expect("build client");
        let model = client.transcription_model(WHISPER_1);

        let error = match model
            .transcription_request()
            .data(vec![0u8; 16])
            .send()
            .await
        {
            Err(error) => error,
            Ok(_) => panic!("transcription should fail with non-success status"),
        };

        assert!(matches!(error, TranscriptionError::HttpError(_)));
        assert_eq!(
            error.provider_response_status(),
            Some(http::StatusCode::BAD_REQUEST)
        );
        assert_eq!(error.provider_response_body(), Some(body));
    }
}