diff --git a/backend/autoscaler.py b/backend/autoscaler.py
new file mode 100644
index 0000000000000000000000000000000000000000..d99214c932ac9cc79be9b4fc47c592a79b162a82
--- /dev/null
+++ b/backend/autoscaler.py
@@ -0,0 +1,65 @@
+import subprocess
+import time
+import requests
+
+# Configuration
+SERVICE_NAME = "mon_stack_backend"  # Change ce nom si différent
+BACKEND_URL = "http://localhost:5000/request_count"
+
+THRESHOLD_UP = 50      # Si plus de 50 requêtes dans l'intervalle -> scale up
+THRESHOLD_DOWN = 10    # Si moins de 10 requêtes -> scale down
+MIN_REPLICAS = 1
+MAX_REPLICAS = 10
+CHECK_INTERVAL = 10    # Secondes entre chaque vérification
+
+# Suivi local des requêtes précédentes
+last_request_count = 0
+
+def get_request_count():
+    """Récupère le nombre total de requêtes reçues (cumulé) depuis l'API backend"""
+    try:
+        response = requests.get(BACKEND_URL)
+        if response.status_code == 200:
+            return response.json()["data"]["request_received_count"]
+    except Exception as e:
+        print(f"❌ Erreur récupération requêtes : {e}")
+    return 0
+
+def get_current_replicas():
+    """Récupère le nombre actuel de réplicas du service"""
+    try:
+        result = subprocess.run(["docker", "service", "ls"], capture_output=True, text=True)
+        for line in result.stdout.splitlines():
+            if SERVICE_NAME in line:
+                return int(line.split()[3].split("/")[0])  # e.g., 3/3 -> 3
+    except Exception as e:
+        print(f"❌ Erreur récupération réplicas : {e}")
+    return MIN_REPLICAS
+
+def scale_service(replicas):
+    """Met à jour le nombre de réplicas via Docker Swarm"""
+    print(f"⚙️ Scaling {SERVICE_NAME} to {replicas} replicas...")
+    subprocess.run(["docker", "service", "scale", f"{SERVICE_NAME}={replicas}"])
+
+print("📈 Autoscaler démarré... (Ctrl+C pour arrêter)\n")
+
+while True:
+    current_request_count = get_request_count()
+    current_replicas = get_current_replicas()
+    
+    delta_requests = current_request_count - last_request_count
+    last_request_count = current_request_count
+
+    print(f"🔄 Requêtes sur les {CHECK_INTERVAL}s : {delta_requests} | Réplicas actuels : {current_replicas}")
+
+    if delta_requests > THRESHOLD_UP and current_replicas < MAX_REPLICAS:
+        print(f"📈 Augmentation de charge → scaling up !")
+        scale_service(current_replicas + 1)
+    elif delta_requests < THRESHOLD_DOWN and current_replicas > MIN_REPLICAS:
+        print(f"📉 Faible charge → scaling down.")
+        scale_service(current_replicas - 1)
+    else:
+        print("✅ Charge stable. Aucun scaling nécessaire.")
+
+    print("-" * 50)
+    time.sleep(CHECK_INTERVAL)
diff --git a/backend/stack.yml b/backend/stack.yml
index 3768f100b9b2cb3d9d69a3db3ffa0b472dda8d59..f8c7ab00772aa7f7d92b816af914eb7682891fa4 100644
--- a/backend/stack.yml
+++ b/backend/stack.yml
@@ -15,7 +15,7 @@ services:
         condition: any
 
   backend:
-    image: md5_backend:latest
+    image: mon-backend:latest
     deploy:
       replicas: 1
       restart_policy:
@@ -27,7 +27,8 @@ services:
       - target: 5000
         published: 5000
         protocol: tcp
-        mode: host
+        # ❌ SUPPRIMÉ : mode: host
+        # En mode ingress par défaut, ce port sera load balancé
     networks:
       - app_network
     environment: