lec12

assets/service/compose.yml

@@ -2,14 +2,13 @@ name: lec_app
 
 services:
   price-predict:
-    image: estate_model:1
+    image: estate_model:2
     ports:
       - "8001:8000"
     volumes:
       - ./models:/models
 
 
-
   requests:
     image: request_service:3
     deploy: 
@@ -19,7 +18,6 @@ services:
         delay: 5s
 
 
-
   prometheus:
     image: prom/prometheus
     ports: 
@@ -38,4 +36,25 @@ services:
     environment:
         - GF_SECURITY_ADMIN_USER=admin
         - GF_SECURITY_ADMIN_PASSWORD=admin
-    
+      
+  database:
+    image: postgres:17.2
+    ports:
+        - 5432:5432 
+    environment:
+        - POSTGRES_USER=admin
+        - POSTGRES_PASSWORD=admin
+        - POSTGRES_DB=my_db_name
+        - PGDATA=/var/lib/postgresql/data/pgdata
+    volumes:
+      - "./database/data:/var/lib/postgresql/data"
+
+  pgadmin:
+    image: dpage/pgadmin4
+    ports:
+        - 9091:80 
+    environment:
+        - PGADMIN_DEFAULT_EMAIL=email@example.com
+        - PGADMIN_DEFAULT_PASSWORD=admin
+    volumes:
+      - "./database/pgadmin:/var/lib/pgadmin"

assets/service/ml_service/api_handler.py

@@ -2,6 +2,8 @@
 import logging
 import pandas as pd
 import pickle as pkl
+import psycopg2
+from datetime import datetime
 
 logger = logging.getLogger("uvicorn.error")
 class FastAPIHandler():
@@ -14,7 +16,25 @@ class FastAPIHandler():
         except Exception as e:
             logger.error('Error loading model')
 
-    def predict(self, item_features:dict):
+    def predict(self, flat_id, item_features:dict):
         item_df = pd.DataFrame(data=item_features, index=[0])
         prediction = self.model.predict(item_df)
+
+        db_cred = {"dbname": "my_db_name",
+                   "user": 'admin',
+                   "password": 'admin',
+                   "host": "database"}
+        
+        db_conn = psycopg2.connect(**db_cred)
+        cur = db_conn.cursor()
+
+        now = datetime.now()
+        cur.execute(f"INSERT INTO public.features \
+                    (flat_id, ts, geo_lon, geo_lat, area) \
+                    VALUES ({flat_id}, '{now}', {item_features['geo_lon']}, {item_features['geo_lat']}, {item_features['area']});")
+        cur.execute(f"INSERT INTO public.predictions \
+                    (flat_id, ts, price) VALUES ({flat_id}, '{now}', {prediction[0]});")
+
+        db_conn.commit()
+
         return (prediction[0])

assets/service/ml_service/main.py

@@ -24,7 +24,7 @@ def root_dir():
 
 @app.post('/api/prediction')
 def make_prediction(flat_id: int, item_features: dict):
-    prediction = app.handler.predict(item_features)
+    prediction = app.handler.predict(flat_id, item_features)
 
     prediction_metric.observe(prediction)
 

assets/service/ml_service/requirements.txt

@@ -4,4 +4,5 @@ pandas
 pickle4
 scikit-learn
 requests
-prometheus_fastapi_instrumentator
+prometheus_fastapi_instrumentator
+psycopg2-binary