From 024290e4c0e29fd65648647603c4883dd4a5bf47 Mon Sep 17 00:00:00 2001
From: Urtzi Alfaro <urtzialfaro@MacBook-Pro-de-Urtzi.local>
Date: Wed, 30 Jul 2025 08:41:47 +0200
Subject: [PATCH] Start fixing forecast service  18

---
 .../app/services/forecasting_service.py       |  8 +++--
 services/training/app/ml/data_processor.py    | 31 ++++++++++++-------
 2 files changed, 25 insertions(+), 14 deletions(-)

diff --git a/services/forecasting/app/services/forecasting_service.py b/services/forecasting/app/services/forecasting_service.py
index 6b8c4a85..a0fe7540 100644
--- a/services/forecasting/app/services/forecasting_service.py
+++ b/services/forecasting/app/services/forecasting_service.py
@@ -122,15 +122,17 @@ class ForecastingService:
                 
                 # Metadata
                 created_at=forecast.created_at,
-                processing_time_ms=forecast.processing_time_ms,
+                processing_time_ms=int((datetime.now() - start_time).total_seconds() * 1000),
                 features_used=forecast.features_used
             )
             
         except Exception as e:
+            processing_time = int((datetime.now() - start_time).total_seconds() * 1000)
             logger.error("Error generating forecast", 
                         error=str(e),
                         product=request.product_name,
-                        tenant_id=tenant_id)
+                        tenant_id=tenant_id,
+                        processing_time=processing_time)
             raise
     
     async def _get_latest_model_with_fallback(
@@ -422,6 +424,8 @@ class ForecastingService:
     ) -> Forecast:
         """Save forecast to database"""
         
+        start_time = datetime.now()
+        
         forecast = Forecast(
             tenant_id=tenant_id,
             product_name=request.product_name,
diff --git a/services/training/app/ml/data_processor.py b/services/training/app/ml/data_processor.py
index 045c3f13..57dedeed 100644
--- a/services/training/app/ml/data_processor.py
+++ b/services/training/app/ml/data_processor.py
@@ -324,15 +324,18 @@ class BakeryDataProcessor:
             if 'date' not in weather_clean.columns and 'ds' in weather_clean.columns:
                 weather_clean = weather_clean.rename(columns={'ds': 'date'})
             
-            # ✅ FIX: Ensure timezone consistency
+            # 🔧 CRITICAL FIX: Ensure both DataFrames have compatible datetime formats
             weather_clean['date'] = pd.to_datetime(weather_clean['date'])
             daily_sales['date'] = pd.to_datetime(daily_sales['date'])
             
-            # Remove timezone info from both to make them compatible
+            # ✅ NEW FIX: Normalize both to timezone-naive datetime for merge compatibility
             if weather_clean['date'].dt.tz is not None:
-                weather_clean['date'] = weather_clean['date'].dt.tz_localize(None)
+                # Convert timezone-aware to UTC then remove timezone info
+                weather_clean['date'] = weather_clean['date'].dt.tz_convert('UTC').dt.tz_localize(None)
+            
             if daily_sales['date'].dt.tz is not None:
-                daily_sales['date'] = daily_sales['date'].dt.tz_localize(None)
+                # Convert timezone-aware to UTC then remove timezone info
+                daily_sales['date'] = daily_sales['date'].dt.tz_convert('UTC').dt.tz_localize(None)
             
             # Map weather columns to standard names
             weather_mapping = {
@@ -390,15 +393,19 @@ class BakeryDataProcessor:
             if 'date' not in traffic_clean.columns and 'ds' in traffic_clean.columns:
                 traffic_clean = traffic_clean.rename(columns={'ds': 'date'})
             
-            # 🔧 FIX: Ensure timezone awareness before merge
+            # 🔧 CRITICAL FIX: Ensure both DataFrames have compatible datetime formats
             traffic_clean['date'] = pd.to_datetime(traffic_clean['date'])
-        
-            # If timezone-naive, localize to UTC
-            if traffic_clean['date'].dt.tz is None:
-                traffic_clean['date'] = traffic_clean['date'].dt.tz_localize('UTC')
-            # If already timezone-aware but not UTC, convert to UTC
-            elif str(traffic_clean['date'].dt.tz) != 'UTC':
-                traffic_clean['date'] = traffic_clean['date'].dt.tz_convert('UTC')
+            daily_sales['date'] = pd.to_datetime(daily_sales['date'])
+            
+            # ✅ NEW FIX: Normalize both to timezone-naive datetime for merge compatibility
+            # This prevents the "datetime64[ns] and datetime64[ns, UTC]" merge error
+            if traffic_clean['date'].dt.tz is not None:
+                # Convert timezone-aware to UTC then remove timezone info
+                traffic_clean['date'] = traffic_clean['date'].dt.tz_convert('UTC').dt.tz_localize(None)
+            
+            if daily_sales['date'].dt.tz is not None:
+                # Convert timezone-aware to UTC then remove timezone info  
+                daily_sales['date'] = daily_sales['date'].dt.tz_convert('UTC').dt.tz_localize(None)
             
             # Map traffic columns to standard names
             traffic_mapping = {