Fix issues 3

services/sales/app/services/data_import_service.py

@@ -865,34 +865,65 @@ class DataImportService:
         return mapping
     
     def _parse_date(self, date_str: str) -> Optional[datetime]:
-        """Enhanced date parsing with pandas and multiple format support"""
+        """Enhanced date parsing with explicit format handling for CSV dates"""
         if not date_str or str(date_str).lower() in ['nan', 'null', 'none']:
             return None
         
         date_str = str(date_str).strip()
         
-        # Try pandas first (most robust)
+        # For CSV format like "2024/10/01", try specific formats first to avoid ambiguity
+        # Priority order: YYYY/MM/DD (most likely for machine-generated data)
+        priority_formats = [
+            '%Y/%m/%d',     # 2024/10/01 (October 1, 2024) - most likely for CSV exports
+            '%Y-%m-%d',     # 2024-10-01 
+            '%d/%m/%Y',     # 01/10/2024 (European format)
+            '%m/%d/%Y',     # 10/01/2024 (US format)
+        ]
+        
+        # Try priority formats first
+        for fmt in priority_formats:
+            try:
+                parsed_dt = datetime.strptime(date_str, fmt)
+                if parsed_dt.tzinfo is None:
+                    parsed_dt = parsed_dt.replace(tzinfo=timezone.utc)
+                logger.debug(f"Successfully parsed date '{date_str}' using format '{fmt}' -> {parsed_dt}")
+                return parsed_dt
+            except ValueError:
+                continue
+        
+        # Try pandas as fallback with explicit format inference
         try:
-            parsed_dt = pd.to_datetime(date_str, dayfirst=True)
+            # For YYYY/MM/DD format, disable dayfirst to prevent misinterpretation
+            if '/' in date_str and len(date_str.split('/')[0]) == 4:
+                # Looks like YYYY/MM/DD format, so don't use dayfirst
+                parsed_dt = pd.to_datetime(date_str, dayfirst=False)
+            else:
+                # For other formats, use dayfirst=True for European-style dates
+                parsed_dt = pd.to_datetime(date_str, dayfirst=True)
+                
             if hasattr(parsed_dt, 'to_pydatetime'):
                 parsed_dt = parsed_dt.to_pydatetime()
             
             if parsed_dt.tzinfo is None:
                 parsed_dt = parsed_dt.replace(tzinfo=timezone.utc)
             
+            logger.debug(f"Successfully parsed date '{date_str}' using pandas -> {parsed_dt}")
             return parsed_dt
-        except Exception:
+        except Exception as e:
+            logger.debug(f"Pandas date parsing failed for '{date_str}': {e}")
             pass
         
-        # Try specific formats as fallback
+        # Try remaining formats as last fallback
         for fmt in self.DATE_FORMATS:
-            try:
-                parsed_dt = datetime.strptime(date_str, fmt)
-                if parsed_dt.tzinfo is None:
-                    parsed_dt = parsed_dt.replace(tzinfo=timezone.utc)
-                return parsed_dt
-            except ValueError:
-                continue
+            if fmt not in priority_formats:  # Skip already tried formats
+                try:
+                    parsed_dt = datetime.strptime(date_str, fmt)
+                    if parsed_dt.tzinfo is None:
+                        parsed_dt = parsed_dt.replace(tzinfo=timezone.utc)
+                    logger.debug(f"Successfully parsed date '{date_str}' using fallback format '{fmt}' -> {parsed_dt}")
+                    return parsed_dt
+                except ValueError:
+                    continue
         
         logger.warning(f"Could not parse date: {date_str}")
         return None

services/training/app/services/training_orchestrator.py

@@ -176,36 +176,65 @@ class TrainingDataOrchestrator:
             logger.error(f"Training data preparation failed: {str(e)}")
             raise ValueError(f"Failed to prepare training data: {str(e)}")
     
-    def _extract_sales_date_range(self, sales_data: List[Dict[str, Any]]) -> DateRange:
-        """Extract date range from sales data with timezone handling and strict date format."""
-        if not sales_data:
-            raise ValueError("No sales data provided")
+    def extract_sales_date_range_utc_localize(sales_data_df: pd.DataFrame):
+        """
+        Extracts the UTC-aware date range from a sales DataFrame using tz_localize.
         
-        dates = []
-        
-        for record in sales_data:
-            date_value = record.get('date')
-            if not date_value:
-                continue  # Skip records with missing date
+        Args:
+            sales_data_df: A pandas DataFrame containing a 'date' column.
             
-            if isinstance(date_value, str):
-                # Parse string with explicit format
-                dt = pd.to_datetime(date_value, format='mixed', errors='raise')
-                if dt.tz is None:
-                    dt = dt.tz_localize('UTC')  # Assign UTC timezone if none
-                dates.append(dt.to_pydatetime())
-            elif isinstance(date_value, datetime):
-                if date_value.tzinfo is None:
-                    date_value = date_value.replace(tzinfo=timezone.utc)
-                dates.append(date_value)
-            else:
-                continue
+        Returns:
+            A tuple of timezone-aware start and end dates in UTC.
+        """
+        if 'date' not in sales_data_df.columns:
+            raise ValueError("DataFrame does not contain a 'date' column.")
         
-        if not dates:
+        # Convert the 'date' column to datetime objects
+        sales_data_df['date'] = pd.to_datetime(sales_data_df['date'])
+        
+        # Localize the naive datetime objects to UTC
+        sales_data_df['date'] = sales_data_df['date'].tz_localize('UTC')
+        
+        # Find the minimum and maximum dates
+        start_date = sales_data_df['date'].min()
+        end_date = sales_data_df['date'].max()
+        
+        return DateRange(start_date, end_date, DataSourceType.BAKERY_SALES)
+
+    def _extract_sales_date_range(self, sales_data: List[Dict[str, Any]]) -> 'DateRange':
+        """
+        Extract date range from sales data with proper date parsing
+        
+        Args:
+            sales_data: List of sales records
+            
+        Returns:
+            DateRange object with timezone-aware start and end dates
+        """
+        if not sales_data:
+            raise ValueError("No sales data provided for date range extraction")
+        
+        # Convert to DataFrame for easier processing
+        sales_df = pd.DataFrame(sales_data)
+        
+        if 'date' not in sales_df.columns:
+            raise ValueError("Sales data does not contain a 'date' column")
+        
+        # Convert dates to datetime with proper parsing
+        # This will use the improved date parsing from the data import service
+        sales_df['date'] = pd.to_datetime(sales_df['date'], utc=True, errors='coerce')
+        
+        # Remove any rows with invalid dates
+        sales_df = sales_df.dropna(subset=['date'])
+        
+        if len(sales_df) == 0:
             raise ValueError("No valid dates found in sales data")
         
-        start_date = min(dates)
-        end_date = max(dates)
+        # Find the minimum and maximum dates
+        start_date = sales_df['date'].min()
+        end_date = sales_df['date'].max()
+        
+        logger.info(f"Extracted sales date range: {start_date} to {end_date}")
         
         return DateRange(start_date, end_date, DataSourceType.BAKERY_SALES)
 

services/training/app/services/training_service.py

@@ -6,7 +6,8 @@ Main training service that uses the repository pattern for data access
 from typing import Dict, List, Any, Optional
 import uuid
 import structlog
-from datetime import datetime
+from datetime import datetime, date, timezone
+from decimal import Decimal
 from sqlalchemy.ext.asyncio import AsyncSession
 import json
 import numpy as np
@@ -37,10 +38,26 @@ logger = structlog.get_logger()
 
 def make_json_serializable(obj):
     """Convert numpy/pandas types, datetime, and UUID objects to JSON-serializable Python types"""
-    import uuid
-    from decimal import Decimal
-    from datetime import datetime, date
     
+    # Handle None values
+    if obj is None:
+        return None
+    
+    # Handle basic datetime types first (most common)
+    if isinstance(obj, datetime):
+        return obj.isoformat()
+    elif isinstance(obj, date):
+        return obj.isoformat()
+    
+    # Handle pandas timestamp types
+    if hasattr(pd, 'Timestamp') and isinstance(obj, pd.Timestamp):
+        return obj.isoformat()
+    
+    # Handle numpy datetime types
+    if hasattr(np, 'datetime64') and isinstance(obj, np.datetime64):
+        return pd.Timestamp(obj).isoformat()
+    
+    # Handle numeric types
     if isinstance(obj, (np.integer, pd.Int64Dtype)):
         return int(obj)
     elif isinstance(obj, (np.floating, pd.Float64Dtype)):
@@ -51,19 +68,36 @@ def make_json_serializable(obj):
         return obj.tolist()
     elif isinstance(obj, pd.DataFrame):
         return obj.to_dict('records')
+    elif isinstance(obj, Decimal):
+        return float(obj)
+    
+    # Handle UUID types
     elif isinstance(obj, uuid.UUID):
         return str(obj)
     elif hasattr(obj, '__class__') and 'UUID' in str(obj.__class__):
         # Handle any UUID-like objects (including asyncpg.pgproto.pgproto.UUID)
         return str(obj)
-    elif isinstance(obj, Decimal):
-        return float(obj)
+    
+    # Handle collections recursively
     elif isinstance(obj, dict):
         return {k: make_json_serializable(v) for k, v in obj.items()}
-    elif isinstance(obj, list):
+    elif isinstance(obj, (list, tuple)):
         return [make_json_serializable(item) for item in obj]
-    else:
+    elif isinstance(obj, set):
+        return [make_json_serializable(item) for item in obj]
+    
+    # Handle other common types
+    elif isinstance(obj, (str, int, float, bool)):
         return obj
+    
+    # Last resort: try to convert to string
+    else:
+        try:
+            # For any other object, try to convert to string
+            return str(obj)
+        except Exception:
+            # If all else fails, return None
+            return None
 
 
 class EnhancedTrainingService: