Function generate_speech

cake-core/src/cake/sharding/api/audio.rs:50–105

(
    state: web::Data<Arc<RwLock<Master<M>>>>,
    req: HttpRequest,
    body: web::Json<AudioSpeechRequest>,
)

}

pub async fn generate_speech<M: Model>(
    state: web::Data<Arc<RwLock<Master<M>>>>,
    req: HttpRequest,
    body: web::Json<AudioSpeechRequest>,
) -> impl Responder {
    let client = req
        .peer_addr()
        .map(|a| a.to_string())
        .unwrap_or_else(|| "unknown".to_string());

    log::info!("starting audio generation for {} ...", &client);

    let mut master = state.write().await;

    if !master.model.as_ref().is_some_and(|m| m.output_modality() == OutputModality::Audio) {
        return HttpResponse::NotFound()
            .json(serde_json::json!({"error": "No audio model loaded"}));
    }

    // Decode base64 voice data if provided
    let voice_data = match &body.voice_data {
        Some(b64) => match general_purpose::STANDARD.decode(b64) {
            Ok(bytes) => Some(bytes),
            Err(e) => {
                return HttpResponse::BadRequest()
                    .json(serde_json::json!({"error": format!("Invalid voice_data base64: {e}")}));
            }
        },
        None => None,
    };

    let args = AudioGenerationArgs {
        input: body.input.clone(),
        voice_data,
        voice_path: body.voice_path.clone(),
        cfg_scale: body.cfg_scale,
        max_frames: body.max_frames,
        diffusion_steps: body.diffusion_steps,
    };

    match master.generate_audio(&args).await {
        Ok(output) => {
            if body.response_format == "pcm" {
                HttpResponse::Ok()
                    .content_type("audio/pcm")
                    .body(output.to_pcm_bytes())
            } else {
                HttpResponse::Ok()
                    .content_type("audio/wav")
                    .body(output.to_wav_bytes())
            }
        }
        Err(e) => HttpResponse::InternalServerError()
            .json(serde_json::json!({"error": format!("{e}")})),
    }
}

#[cfg(test)]