app.py

import os
import tempfile
import torch
from flask import Flask, request, jsonify, render_template, send_file
from transformers import WhisperProcessor, WhisperForConditionalGeneration
import librosa
import warnings

warnings.filterwarnings('ignore')

app = Flask(__name__)

# ========================
# CONFIGURATION
# ========================
STT_MODELS = {
    "yoruba": {
        "model_id": "ajibs75/whisper-small-yoruba",
        "language": "yo"
    },
    "hausa": {
        "model_id": "NCAIR1/Hausa-ASR",
        "language": "ha"
    }
}

device = "cuda" if torch.cuda.is_available() else "cpu"
print(f"🖥️ Device: {device}")

models = {}
processors = {}

# ========================
# LOAD MODELS AT STARTUP
# ========================
print("=" * 60)
print("🎤 Chargement des modèles STT...")
print("=" * 60)

for lang, cfg in STT_MODELS.items():
    print(f"📥 Chargement du modèle {lang}...")
    try:
        processor = WhisperProcessor.from_pretrained(cfg["model_id"])
        model = WhisperForConditionalGeneration.from_pretrained(cfg["model_id"])
        model.to(device)
        model.eval()
        
        processors[lang] = processor
        models[lang] = model
        print(f"✅ {lang.capitalize()} prêt")
    except Exception as e:
        print(f"❌ Erreur {lang}: {e}")

print("=" * 60)
print("✅ Tous les modèles sont chargés")
print("=" * 60)

# ========================
# UTILITY FUNCTIONS
# ========================
def transcribe_audio(audio_path: str, language: str) -> str:
    """Transcrit un fichier audio avec le modèle approprié"""
    
    if language not in models:
        raise ValueError(f"Langue non supportée: {language}")
    
    processor = processors[language]
    model = models[language]
    
    # Charger l'audio avec librosa (resample à 16kHz)
    audio, sr = librosa.load(audio_path, sr=16000)
    
    # Prétraiter l'audio
    input_features = processor(
        audio,
        sampling_rate=16000,
        return_tensors="pt"
    ).input_features.to(device)
    
    # Générer la transcription
    with torch.no_grad():
        predicted_ids = model.generate(
            input_features,
            max_length=448,
            num_beams=5,
            temperature=0.0
        )
    
    # Décoder
    transcription = processor.batch_decode(
        predicted_ids,
        skip_special_tokens=True
    )[0]
    
    return transcription.strip()

# ========================
# ROUTES
# ========================
@app.route("/")
def home():
    """Page d'accueil"""
    return render_template("index.html")

@app.route("/stt", methods=["POST"])
def stt():
    """Endpoint de transcription"""
    try:
        # Vérifier le fichier audio
        if "audio" not in request.files:
            return jsonify({"error": "Aucun fichier audio fourni"}), 400
        
        audio_file = request.files["audio"]
        
        if audio_file.filename == "":
            return jsonify({"error": "Nom de fichier vide"}), 400
        
        # Récupérer la langue
        language = request.form.get("language", "").lower()
        
        if language not in models:
            return jsonify({"error": f"Langue non supportée: {language}"}), 400
        
        # Sauvegarder temporairement le fichier
        with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_audio:
            audio_file.save(temp_audio.name)
            temp_path = temp_audio.name
        
        print(f"📝 Transcription en cours ({language})...")
        
        # Transcrire
        transcription = transcribe_audio(temp_path, language)
        
        # Nettoyer
        os.remove(temp_path)
        
        print(f"✅ Transcription: {transcription}")
        
        return jsonify({
            "success": True,
            "language": language,
            "transcription": transcription
        })
    
    except Exception as e:
        print(f"❌ Erreur: {str(e)}")
        return jsonify({
            "success": False,
            "error": str(e)
        }), 500

@app.route("/health")
def health():
    """Health check"""
    return jsonify({
        "status": "healthy",
        "device": device,
        "models": list(models.keys())
    })

@app.route("/languages")
def languages():
    """Liste des langues disponibles"""
    return jsonify({
        "languages": [
            {"code": "yoruba", "name": "Yoruba", "flag": "🇳🇬"},
            {"code": "hausa", "name": "Hausa", "flag": "🇳🇬"}
        ]
    })

# ========================
# RUN
# ========================
if __name__ == "__main__":
    port = int(os.environ.get("PORT", 7860))
    app.run(host="0.0.0.0", port=port, debug=False)