Merge pull request #4 from DataScientest-Studio/refactor/load-recomme…

…nder-model-def-in-api-predict

Refactor and Docs

README.md

@@ -15,7 +15,7 @@ System of recommendation of movies based on the user's profile and preferences.
 
 ## ℹ️ About the Project
 
-This project is a starting pack for MLOps projects focused on the subject of "movie recommendation". It provides a structured framework to develop, train, and deploy machine learning models for recommending movies to users. It uses Supabase for the backend, Airflow for the orchestration, MLflow for the tracking, Minio for the storage of the models, Prometheus and Grafana for the monitoring.
+This project is a starting pack for MLOps projects focused on the subject of "movie recommendation". It provides a fully integrated local development environment where all tools and applications are containerized and managed through Docker Compose. The stack includes Supabase for the backend, Airflow for workflow orchestration, MLflow for experiment tracking, Minio for model storage, and Prometheus/Grafana for monitoring - all configured to work together seamlessly in your local environment. This allows you to develop, train, and deploy locally machine learning models for recommending movies with a complete MLOps toolchain running entirely on your machine.
 
 ## 🏞 Showcase
 

app/api/predict/main.py

@@ -30,29 +30,20 @@
 model = None
 model_infos = None
 
+MLFLOW_URI = os.getenv("MLFLOW_TRACKING_URI", "http://tracking_server:5000")
+
+
+def load_model_from_mlflow():
+    print(f"Tentative de connexion à MLflow sur : {MLFLOW_URI}")
+
+    mlflow.set_tracking_uri(MLFLOW_URI)
 
-def load_recommender_model():
     try:
-        # Essayer d'abord MLflow
-        mlflow_uri = os.getenv("MLFLOW_TRACKING_URI", "http://tracking_server:5000")
-        print(f"Tentative de connexion à MLflow sur : {mlflow_uri}")
-
-        mlflow.set_tracking_uri(mlflow_uri)
-        print("URI MLflow configuré")
-
-        # Vérifier la connexion à MLflow avec un délai d'attente
-        try:
-            import requests
-            response = requests.get(mlflow_uri, timeout=5)  # Délai d'attente de 5 secondes
-            response.raise_for_status()
-            mlflow.search_runs()
-            print("Connexion à MLflow réussie")
-        except requests.exceptions.RequestException as e:
-            print(f"Erreur de connexion à MLflow: {str(e)}")
-            raise
-
-        # Tenter de charger le modèle
-        print("Tentative de chargement du modèle 'movie_recommender'")
+        response = requests.get(MLFLOW_URI, timeout=5)
+        response.raise_for_status()
+        mlflow.search_runs()
+        print("Connexion à MLflow réussie")
+
         client = mlflow.tracking.MlflowClient()
         model_champion = client.get_model_version_by_alias(
             name="movie_recommender", alias="champion"
@@ -61,38 +52,42 @@ def load_recommender_model():
         model = mlflow.sklearn.load_model(f"models:/movie_recommender/{model_version}")
         print("Modèle chargé avec succès depuis MLflow")
 
-        MODEL_INFO.info({"model_name": "movie_recommender", "source": "mlflow"})
+        return model, model_version
+    except requests.exceptions.RequestException as e:
+        print(f"Erreur de connexion à MLflow: {str(e)}")
+        raise
 
-        model_infos = {
-            "model_name": "movie_recommender",
-            "model_version": model_version,
-            "source": "mlflow",
-        }
 
-        return model, model_infos
+def load_model_locally():
+    print("Tentative de chargement du modèle local")
 
+    try:
+        with open("model.pkl", "rb") as f:
+            model = pickle.load(f)
+        print("Modèle local chargé avec succès")
+        return model, "local"
     except Exception as e:
-        print("Tentative de chargement du modèle local de secours")
+        print(f"Erreur lors du chargement du modèle local: {str(e)}")
+        raise
 
-        try:
-            # Charger le modèle local
-            with open("model.pkl", "rb") as f:
-                model = pickle.load(f)
-            print("Modèle local chargé avec succès")
 
-            MODEL_INFO.info({"model_name": "movie_recommender", "source": "local"})
+def load_recommender_model():
+    try:
+        model, model_version = load_model_from_mlflow()
+        source = "mlflow"
+    except Exception:
+        model, model_version = load_model_locally()
+        source = "local"
 
-            model_infos = {
-                "model_name": "movie_recommender",
-                "model_version": "local",
-                "source": "local",
-            }
+    model_infos = {
+        "model_name": "movie_recommender",
+        "model_version": model_version,
+        "source": source,
+    }
 
-            return model, model_infos
+    MODEL_INFO.info(model_infos)
 
-        except Exception as e:
-            print(f"Erreur lors du chargement du modèle local: {str(e)}")
-            raise
+    return model, model_infos
 
 
 def make_predictions(genres, model):