Improve AI logic

services/training/Dockerfile

@@ -9,10 +9,13 @@ FROM python:3.11-slim
 
 WORKDIR /app
 
-# Install system dependencies
+# Install system dependencies including cmdstan requirements
 RUN apt-get update && apt-get install -y \
     gcc \
+    g++ \
+    make \
     curl \
+    build-essential \
     && rm -rf /var/lib/apt/lists/*
 
 # Copy requirements
@@ -36,6 +39,13 @@ COPY services/training/ .
 # Add shared libraries to Python path
 ENV PYTHONPATH="/app:/app/shared:${PYTHONPATH:-}"
 
+# Set TMPDIR for cmdstan (directory will be created at runtime)
+ENV TMPDIR=/tmp/cmdstan
+
+# Install cmdstan for Prophet (required for model optimization)
+# Suppress verbose output to reduce log noise
+RUN python -m pip install --no-cache-dir cmdstanpy && \
+    python -m cmdstanpy.install_cmdstan
 
 # Expose port
 EXPOSE 8000

services/training/app/api/models.py

@@ -39,7 +39,8 @@ router = APIRouter()
 training_service = EnhancedTrainingService()
 
 @router.get(
-    route_builder.build_base_route("models") + "/{inventory_product_id}/active"
+    route_builder.build_base_route("models") + "/{inventory_product_id}/active",
+    response_model=TrainedModelResponse
 )
 async def get_active_model(
     tenant_id: str = Path(..., description="Tenant ID"),
@@ -90,21 +91,25 @@ async def get_active_model(
         await db.commit()
         
         return {
-            "model_id": str(model_record.id),  # ✅ This is the correct field name
+            "model_id": str(model_record.id),
+            "tenant_id": str(model_record.tenant_id),
+            "inventory_product_id": str(model_record.inventory_product_id),
+            "model_type": model_record.model_type,
             "model_path": model_record.model_path,
-            "features_used": model_record.features_used,
-            "hyperparameters": model_record.hyperparameters,
+            "version": 1,  # Default version
+            "training_samples": model_record.training_samples or 0,
+            "features": model_record.features_used or [],
+            "hyperparameters": model_record.hyperparameters or {},
             "training_metrics": {
-                "mape": model_record.mape,
-                "mae": model_record.mae,
-                "rmse": model_record.rmse,
-                "r2_score": model_record.r2_score
+                "mape": model_record.mape or 0.0,
+                "mae": model_record.mae or 0.0,
+                "rmse": model_record.rmse or 0.0,
+                "r2_score": model_record.r2_score or 0.0
             },
-            "created_at": model_record.created_at.isoformat() if model_record.created_at else None,
-            "training_period": {
-                "start_date": model_record.training_start_date.isoformat() if model_record.training_start_date else None,
-                "end_date": model_record.training_end_date.isoformat() if model_record.training_end_date else None
-            }
+            "is_active": model_record.is_active,
+            "created_at": model_record.created_at,
+            "data_period_start": model_record.training_start_date,
+            "data_period_end": model_record.training_end_date
         }
         
     except HTTPException:

services/training/app/ml/data_processor.py

@@ -17,6 +17,8 @@ from shared.database.base import create_database_manager
 from shared.database.transactions import transactional
 from shared.database.exceptions import DatabaseError
 from app.core.config import settings
+from app.ml.enhanced_features import AdvancedFeatureEngineer
+import holidays
 
 logger = structlog.get_logger()
 
@@ -26,16 +28,67 @@ class EnhancedBakeryDataProcessor:
     Integrates date alignment, data cleaning, feature engineering, and preparation for ML models.
     """
     
-    def __init__(self, database_manager=None):
+    def __init__(self, database_manager=None, region: str = 'MD'):
         self.database_manager = database_manager or create_database_manager(settings.DATABASE_URL, "training-service")
         self.scalers = {}  # Store scalers for each feature
         self.imputers = {}  # Store imputers for missing value handling
         self.date_alignment_service = DateAlignmentService()
+        self.feature_engineer = AdvancedFeatureEngineer()
+        self.region = region  # Region for holidays (MD=Madrid, PV=Basque, etc.)
+        self.spain_holidays = holidays.Spain(prov=region)  # Initialize holidays library
     
     def get_scalers(self) -> Dict[str, Any]:
         """Return the scalers/normalization parameters for use during prediction"""
         return self.scalers.copy()
-        
+
+    @staticmethod
+    def _extract_numeric_from_dict(value: Any) -> Optional[float]:
+        """
+        Robust extraction of numeric values from complex data structures.
+        Handles various dict structures that might come from external APIs.
+
+        Args:
+            value: Any value that might be a dict, numeric, or other type
+
+        Returns:
+            Numeric value as float, or None if extraction fails
+        """
+        # If already numeric, return it
+        if isinstance(value, (int, float)) and not isinstance(value, bool):
+            return float(value)
+
+        # If it's a dict, try multiple extraction strategies
+        if isinstance(value, dict):
+            # Strategy 1: Try common keys
+            for key in ['value', 'data', 'result', 'amount', 'count', 'number', 'val']:
+                if key in value:
+                    extracted = value[key]
+                    # Recursively extract if nested
+                    if isinstance(extracted, dict):
+                        return EnhancedBakeryDataProcessor._extract_numeric_from_dict(extracted)
+                    elif isinstance(extracted, (int, float)) and not isinstance(extracted, bool):
+                        return float(extracted)
+
+            # Strategy 2: Try to find first numeric value in dict
+            for v in value.values():
+                if isinstance(v, (int, float)) and not isinstance(v, bool):
+                    return float(v)
+                elif isinstance(v, dict):
+                    # Recursively try nested dicts
+                    result = EnhancedBakeryDataProcessor._extract_numeric_from_dict(v)
+                    if result is not None:
+                        return result
+
+        # Strategy 3: Try to convert string to numeric
+        if isinstance(value, str):
+            try:
+                return float(value)
+            except (ValueError, TypeError):
+                pass
+
+        # If all strategies fail, return None (will be converted to NaN)
+        return None
+
     async def _get_repositories(self, session):
         """Initialize repositories with session"""
         return {
@@ -117,9 +170,12 @@ class EnhancedBakeryDataProcessor:
                 daily_sales = self._merge_weather_features(daily_sales, weather_data)
                 daily_sales = self._merge_traffic_features(daily_sales, traffic_data)
                 
-                # Step 6: Engineer additional features
+                # Step 6: Engineer basic features
                 daily_sales = self._engineer_features(daily_sales)
-                
+
+                # Step 6b: Add advanced features (lagged, rolling, cyclical, interactions, trends)
+                daily_sales = self._add_advanced_features(daily_sales)
+
                 # Step 7: Handle missing values
                 daily_sales = self._handle_missing_values(daily_sales)
                 
@@ -177,52 +233,73 @@ class EnhancedBakeryDataProcessor:
     async def prepare_prediction_features(self,
                                          future_dates: pd.DatetimeIndex,
                                          weather_forecast: pd.DataFrame = None,
-                                         traffic_forecast: pd.DataFrame = None) -> pd.DataFrame:
+                                         traffic_forecast: pd.DataFrame = None,
+                                         historical_data: pd.DataFrame = None) -> pd.DataFrame:
         """
         Create features for future predictions with proper date handling.
-        
+
         Args:
             future_dates: Future dates to predict
             weather_forecast: Weather forecast data
             traffic_forecast: Traffic forecast data
-            
+            historical_data: Historical data for creating lagged and rolling features
+
         Returns:
             DataFrame with features for prediction
         """
         try:
             # Create base future dataframe
             future_df = pd.DataFrame({'ds': future_dates})
-            
+
             # Add temporal features
             future_df = self._add_temporal_features(
                 future_df.rename(columns={'ds': 'date'})
             ).rename(columns={'date': 'ds'})
-            
+
             # Add weather features
             if weather_forecast is not None and not weather_forecast.empty:
                 weather_features = weather_forecast.copy()
                 if 'date' in weather_features.columns:
                     weather_features = weather_features.rename(columns={'date': 'ds'})
-                
+
                 future_df = future_df.merge(weather_features, on='ds', how='left')
-            
-            # Add traffic features  
+
+            # Add traffic features
             if traffic_forecast is not None and not traffic_forecast.empty:
                 traffic_features = traffic_forecast.copy()
                 if 'date' in traffic_features.columns:
                     traffic_features = traffic_features.rename(columns={'date': 'ds'})
-                
+
                 future_df = future_df.merge(traffic_features, on='ds', how='left')
-            
-            # Engineer additional features
+
+            # Engineer basic features
             future_df = self._engineer_features(future_df.rename(columns={'ds': 'date'}))
+
+            # Add advanced features if historical data is provided
+            if historical_data is not None and not historical_data.empty:
+                # Combine historical and future data to calculate lagged/rolling features
+                combined_df = pd.concat([
+                    historical_data.rename(columns={'ds': 'date'}),
+                    future_df
+                ], ignore_index=True).sort_values('date')
+
+                # Apply advanced features to combined data
+                combined_df = self._add_advanced_features(combined_df)
+
+                # Extract only the future rows
+                future_df = combined_df[combined_df['date'].isin(future_df['date'])].copy()
+            else:
+                # Without historical data, add advanced features with NaN for lags
+                logger.warning("No historical data provided, lagged features will be NaN")
+                future_df = self._add_advanced_features(future_df)
+
             future_df = future_df.rename(columns={'date': 'ds'})
-            
+
             # Handle missing values in future data
             future_df = self._handle_missing_values_future(future_df)
-            
+
             return future_df
-            
+
         except Exception as e:
             logger.error("Error creating prediction features", error=str(e))
             # Return minimal features if error
@@ -428,19 +505,40 @@ class EnhancedBakeryDataProcessor:
             for standard_name, possible_names in weather_mapping.items():
                 for possible_name in possible_names:
                     if possible_name in weather_clean.columns:
-                        weather_clean[standard_name] = pd.to_numeric(weather_clean[possible_name], errors='coerce')
+                        # Extract numeric values using robust helper function
+                        try:
+                            # Check if column contains dict-like objects
+                            has_dicts = weather_clean[possible_name].apply(lambda x: isinstance(x, dict)).any()
+
+                            if has_dicts:
+                                logger.warning(f"Weather column {possible_name} contains dict objects, extracting numeric values")
+                                # Use robust extraction for all values
+                                weather_clean[standard_name] = weather_clean[possible_name].apply(
+                                    self._extract_numeric_from_dict
+                                )
+                            else:
+                                # Direct numeric conversion for simple values
+                                weather_clean[standard_name] = pd.to_numeric(weather_clean[possible_name], errors='coerce')
+                        except Exception as e:
+                            logger.warning(f"Error converting weather column {possible_name}: {e}")
+                            # Fallback: try to extract from each value
+                            weather_clean[standard_name] = weather_clean[possible_name].apply(
+                                self._extract_numeric_from_dict
+                            )
                         weather_features.append(standard_name)
                         break
-            
+
             # Keep only the features we found
             weather_clean = weather_clean[weather_features].copy()
-            
+
             # Merge with sales data
             merged = daily_sales.merge(weather_clean, on='date', how='left')
-            
+
             # Fill missing weather values with Madrid-appropriate defaults
             for feature, default_value in weather_defaults.items():
                 if feature in merged.columns:
+                    # Ensure the column is numeric before filling
+                    merged[feature] = pd.to_numeric(merged[feature], errors='coerce')
                     merged[feature] = merged[feature].fillna(default_value)
             
             return merged
@@ -494,16 +592,35 @@ class EnhancedBakeryDataProcessor:
             for standard_name, possible_names in traffic_mapping.items():
                 for possible_name in possible_names:
                     if possible_name in traffic_clean.columns:
-                        traffic_clean[standard_name] = pd.to_numeric(traffic_clean[possible_name], errors='coerce')
+                        # Extract numeric values using robust helper function
+                        try:
+                            # Check if column contains dict-like objects
+                            has_dicts = traffic_clean[possible_name].apply(lambda x: isinstance(x, dict)).any()
+
+                            if has_dicts:
+                                logger.warning(f"Traffic column {possible_name} contains dict objects, extracting numeric values")
+                                # Use robust extraction for all values
+                                traffic_clean[standard_name] = traffic_clean[possible_name].apply(
+                                    self._extract_numeric_from_dict
+                                )
+                            else:
+                                # Direct numeric conversion for simple values
+                                traffic_clean[standard_name] = pd.to_numeric(traffic_clean[possible_name], errors='coerce')
+                        except Exception as e:
+                            logger.warning(f"Error converting traffic column {possible_name}: {e}")
+                            # Fallback: try to extract from each value
+                            traffic_clean[standard_name] = traffic_clean[possible_name].apply(
+                                self._extract_numeric_from_dict
+                            )
                         traffic_features.append(standard_name)
                         break
-            
+
             # Keep only the features we found
             traffic_clean = traffic_clean[traffic_features].copy()
-            
+
             # Merge with sales data
             merged = daily_sales.merge(traffic_clean, on='date', how='left')
-            
+
             # Fill missing traffic values with reasonable defaults
             traffic_defaults = {
                 'traffic_volume': 100.0,
@@ -511,9 +628,11 @@ class EnhancedBakeryDataProcessor:
                 'congestion_level': 1.0,  # Low congestion
                 'average_speed': 30.0     # km/h typical for Madrid
             }
-            
+
             for feature, default_value in traffic_defaults.items():
                 if feature in merged.columns:
+                    # Ensure the column is numeric before filling
+                    merged[feature] = pd.to_numeric(merged[feature], errors='coerce')
                     merged[feature] = merged[feature].fillna(default_value)
             
             return merged
@@ -530,17 +649,23 @@ class EnhancedBakeryDataProcessor:
         
         # Weather-based features
         if 'temperature' in df.columns:
+            # Ensure temperature is numeric (defensive check)
+            df['temperature'] = pd.to_numeric(df['temperature'], errors='coerce').fillna(15.0)
+
             df['temp_squared'] = df['temperature'] ** 2
             df['is_hot_day'] = (df['temperature'] > 25).astype(int)  # Hot days in Madrid
             df['is_cold_day'] = (df['temperature'] < 10).astype(int)  # Cold days
             df['is_pleasant_day'] = ((df['temperature'] >= 18) & (df['temperature'] <= 25)).astype(int)
-            
+
             # Temperature categories for bakery products
-            df['temp_category'] = pd.cut(df['temperature'], 
-                                    bins=[-np.inf, 5, 15, 25, np.inf], 
+            df['temp_category'] = pd.cut(df['temperature'],
+                                    bins=[-np.inf, 5, 15, 25, np.inf],
                                     labels=[0, 1, 2, 3]).astype(int)
-        
+
         if 'precipitation' in df.columns:
+            # Ensure precipitation is numeric (defensive check)
+            df['precipitation'] = pd.to_numeric(df['precipitation'], errors='coerce').fillna(0.0)
+
             df['is_rainy_day'] = (df['precipitation'] > 0.1).astype(int)
             df['is_heavy_rain'] = (df['precipitation'] > 10).astype(int)
             df['rain_intensity'] = pd.cut(df['precipitation'],
@@ -549,10 +674,13 @@ class EnhancedBakeryDataProcessor:
         
         # Traffic-based features with NaN protection
         if 'traffic_volume' in df.columns:
+            # Ensure traffic_volume is numeric (defensive check)
+            df['traffic_volume'] = pd.to_numeric(df['traffic_volume'], errors='coerce').fillna(100.0)
+
             # Calculate traffic quantiles for relative measures
             q75 = df['traffic_volume'].quantile(0.75)
             q25 = df['traffic_volume'].quantile(0.25)
-            
+
             df['high_traffic'] = (df['traffic_volume'] > q75).astype(int)
             df['low_traffic'] = (df['traffic_volume'] < q25).astype(int)
             
@@ -578,7 +706,15 @@ class EnhancedBakeryDataProcessor:
             
             # Fill any remaining NaN values
             df['traffic_normalized'] = df['traffic_normalized'].fillna(0.0)
-        
+
+        # Ensure other weather features are numeric if they exist
+        for weather_col in ['humidity', 'wind_speed', 'pressure', 'pedestrian_count', 'congestion_level', 'average_speed']:
+            if weather_col in df.columns:
+                df[weather_col] = pd.to_numeric(df[weather_col], errors='coerce').fillna(
+                    {'humidity': 60.0, 'wind_speed': 5.0, 'pressure': 1013.0,
+                     'pedestrian_count': 50.0, 'congestion_level': 1.0, 'average_speed': 30.0}.get(weather_col, 0.0)
+                )
+
         # Interaction features - bakery specific
         if 'is_weekend' in df.columns and 'temperature' in df.columns:
             df['weekend_temp_interaction'] = df['is_weekend'] * df['temperature']
@@ -619,7 +755,39 @@ class EnhancedBakeryDataProcessor:
                               column=col,
                               nan_count=nan_count)
                 df[col] = df[col].fillna(0.0)
-        
+
+        return df
+
+    def _add_advanced_features(self, df: pd.DataFrame) -> pd.DataFrame:
+        """
+        Add advanced features using AdvancedFeatureEngineer.
+        Includes lagged features, rolling statistics, cyclical encoding, and trend features.
+        """
+        df = df.copy()
+
+        logger.info("Adding advanced features (lagged, rolling, cyclical, trends)")
+
+        # Reset feature engineer to clear previous features
+        self.feature_engineer = AdvancedFeatureEngineer()
+
+        # Create all advanced features at once
+        df = self.feature_engineer.create_all_features(
+            df,
+            date_column='date',
+            include_lags=True,
+            include_rolling=True,
+            include_interactions=True,
+            include_cyclical=True
+        )
+
+        # Fill NA values from lagged and rolling features
+        df = self.feature_engineer.fill_na_values(df, strategy='forward_backward')
+
+        # Store created feature columns for later reference
+        created_features = self.feature_engineer.get_feature_columns()
+        logger.info(f"Added {len(created_features)} advanced features",
+                   features=created_features[:10])  # Log first 10 for brevity
+
         return df
     
     def _handle_missing_values(self, df: pd.DataFrame) -> pd.DataFrame:
@@ -733,46 +901,83 @@ class EnhancedBakeryDataProcessor:
             return 4  # Autumn
     
     def _is_spanish_holiday(self, date: datetime) -> bool:
-        """Check if a date is a major Spanish holiday"""
-        month_day = (date.month, date.day)
-        
-        # Major Spanish holidays that affect bakery sales
-        spanish_holidays = [
-            (1, 1),   # New Year
-            (1, 6),   # Epiphany (Reyes)
-            (5, 1),   # Labour Day
-            (8, 15),  # Assumption
-            (10, 12), # National Day
-            (11, 1),  # All Saints
-            (12, 6),  # Constitution
-            (12, 8),  # Immaculate Conception
-            (12, 25), # Christmas
-            (5, 15),  # San Isidro (Madrid patron saint)
-            (5, 2),   # Madrid Community Day
-        ]
-        
-        return month_day in spanish_holidays
+        """
+        Check if a date is a Spanish holiday using holidays library.
+        Supports dynamic Easter calculation and regional holidays.
+        """
+        try:
+            # Convert to date if datetime
+            if isinstance(date, datetime):
+                date = date.date()
+            elif isinstance(date, pd.Timestamp):
+                date = date.date()
+
+            # Check if date is in holidays
+            return date in self.spain_holidays
+        except Exception as e:
+            logger.warning(f"Error checking holiday status for {date}: {e}")
+            # Fallback to checking basic holidays
+            month_day = (date.month, date.day)
+            basic_holidays = [
+                (1, 1), (1, 6), (5, 1), (8, 15), (10, 12),
+                (11, 1), (12, 6), (12, 8), (12, 25)
+            ]
+            return month_day in basic_holidays
     
     def _is_school_holiday(self, date: datetime) -> bool:
-        """Check if a date is during school holidays (approximate)"""
-        month = date.month
-        
-        # Approximate Spanish school holiday periods
-        # Summer holidays (July-August)
-        if month in [7, 8]:
-            return True
-        
-        # Christmas holidays (mid December to early January)
-        if month == 12 and date.day >= 20:
-            return True
-        if month == 1 and date.day <= 10:
-            return True
-        
-        # Easter holidays (approximate - early April)
-        if month == 4 and date.day <= 15:
-            return True
-        
-        return False
+        """
+        Check if a date is during school holidays in Spain.
+        Uses dynamic Easter calculation and standard Spanish school calendar.
+        """
+        try:
+            from datetime import timedelta
+            import holidays as hol
+
+            # Convert to date if datetime
+            if isinstance(date, datetime):
+                check_date = date.date()
+            elif isinstance(date, pd.Timestamp):
+                check_date = date.date()
+            else:
+                check_date = date
+
+            month = check_date.month
+            day = check_date.day
+
+            # Summer holidays (July 1 - August 31)
+            if month in [7, 8]:
+                return True
+
+            # Christmas holidays (December 23 - January 7)
+            if (month == 12 and day >= 23) or (month == 1 and day <= 7):
+                return True
+
+            # Easter/Spring break (Semana Santa)
+            # Calculate Easter for this year
+            year = check_date.year
+            spain_hol = hol.Spain(years=year, prov=self.region)
+
+            # Find Easter dates (Viernes Santo - Good Friday, and nearby days)
+            # Easter break typically spans 1 week before and after Easter Sunday
+            for holiday_date, holiday_name in spain_hol.items():
+                if 'viernes santo' in holiday_name.lower() or 'easter' in holiday_name.lower():
+                    # Easter break: 7 days before and 7 days after
+                    easter_start = holiday_date - timedelta(days=7)
+                    easter_end = holiday_date + timedelta(days=7)
+                    if easter_start <= check_date <= easter_end:
+                        return True
+
+            return False
+
+        except Exception as e:
+            logger.warning(f"Error checking school holiday for {date}: {e}")
+            # Fallback to simple approximation
+            month = date.month if hasattr(date, 'month') else date.month
+            day = date.day if hasattr(date, 'day') else date.day
+            return (month in [7, 8] or
+                   (month == 12 and day >= 23) or
+                   (month == 1 and day <= 7) or
+                   (month == 4 and 1 <= day <= 15))  # Approximate Easter
     
     async def calculate_feature_importance(self, 
                                    model_data: pd.DataFrame,

services/training/app/ml/enhanced_features.py

@@ -0,0 +1,347 @@
+"""
+Enhanced Feature Engineering for Hybrid Prophet + XGBoost Models
+Adds lagged features, rolling statistics, and advanced interactions
+"""
+
+import pandas as pd
+import numpy as np
+from typing import Dict, List, Optional
+import structlog
+
+logger = structlog.get_logger()
+
+
+class AdvancedFeatureEngineer:
+    """
+    Advanced feature engineering for hybrid forecasting models.
+    Adds lagged features, rolling statistics, and complex interactions.
+    """
+
+    def __init__(self):
+        self.feature_columns = []
+
+    def add_lagged_features(self, df: pd.DataFrame, lag_days: List[int] = None) -> pd.DataFrame:
+        """
+        Add lagged demand features for capturing recent trends.
+
+        Args:
+            df: DataFrame with 'quantity' column
+            lag_days: List of lag periods (default: [1, 7, 14])
+
+        Returns:
+            DataFrame with added lagged features
+        """
+        if lag_days is None:
+            lag_days = [1, 7, 14]
+
+        df = df.copy()
+
+        for lag in lag_days:
+            col_name = f'lag_{lag}_day'
+            df[col_name] = df['quantity'].shift(lag)
+            self.feature_columns.append(col_name)
+
+        logger.info(f"Added {len(lag_days)} lagged features", lags=lag_days)
+        return df
+
+    def add_rolling_features(
+        self,
+        df: pd.DataFrame,
+        windows: List[int] = None,
+        features: List[str] = None
+    ) -> pd.DataFrame:
+        """
+        Add rolling statistics (mean, std, max, min).
+
+        Args:
+            df: DataFrame with 'quantity' column
+            windows: List of window sizes (default: [7, 14, 30])
+            features: List of statistics to calculate (default: ['mean', 'std', 'max', 'min'])
+
+        Returns:
+            DataFrame with rolling features
+        """
+        if windows is None:
+            windows = [7, 14, 30]
+
+        if features is None:
+            features = ['mean', 'std', 'max', 'min']
+
+        df = df.copy()
+
+        for window in windows:
+            for feature in features:
+                col_name = f'rolling_{feature}_{window}d'
+
+                if feature == 'mean':
+                    df[col_name] = df['quantity'].rolling(window=window, min_periods=max(1, window // 2)).mean()
+                elif feature == 'std':
+                    df[col_name] = df['quantity'].rolling(window=window, min_periods=max(1, window // 2)).std()
+                elif feature == 'max':
+                    df[col_name] = df['quantity'].rolling(window=window, min_periods=max(1, window // 2)).max()
+                elif feature == 'min':
+                    df[col_name] = df['quantity'].rolling(window=window, min_periods=max(1, window // 2)).min()
+
+                self.feature_columns.append(col_name)
+
+        logger.info(f"Added rolling features", windows=windows, features=features)
+        return df
+
+    def add_day_of_week_features(self, df: pd.DataFrame, date_column: str = 'date') -> pd.DataFrame:
+        """
+        Add enhanced day-of-week features.
+
+        Args:
+            df: DataFrame with date column
+            date_column: Name of date column
+
+        Returns:
+            DataFrame with day-of-week features
+        """
+        df = df.copy()
+
+        # Day of week (0=Monday, 6=Sunday)
+        df['day_of_week'] = df[date_column].dt.dayofweek
+
+        # Is weekend
+        df['is_weekend'] = (df['day_of_week'] >= 5).astype(int)
+
+        # Is Friday (often higher demand due to weekend prep)
+        df['is_friday'] = (df['day_of_week'] == 4).astype(int)
+
+        # Is Monday (often lower demand after weekend)
+        df['is_monday'] = (df['day_of_week'] == 0).astype(int)
+
+        # Add to feature list
+        for col in ['day_of_week', 'is_weekend', 'is_friday', 'is_monday']:
+            if col not in self.feature_columns:
+                self.feature_columns.append(col)
+
+        return df
+
+    def add_calendar_enhanced_features(self, df: pd.DataFrame, date_column: str = 'date') -> pd.DataFrame:
+        """
+        Add enhanced calendar features beyond basic temporal features.
+
+        Args:
+            df: DataFrame with date column
+            date_column: Name of date column
+
+        Returns:
+            DataFrame with enhanced calendar features
+        """
+        df = df.copy()
+
+        # Month and quarter (if not already present)
+        if 'month' not in df.columns:
+            df['month'] = df[date_column].dt.month
+
+        if 'quarter' not in df.columns:
+            df['quarter'] = df[date_column].dt.quarter
+
+        # Day of month
+        df['day_of_month'] = df[date_column].dt.day
+
+        # Is month start/end
+        df['is_month_start'] = (df['day_of_month'] <= 3).astype(int)
+        df['is_month_end'] = (df[date_column].dt.is_month_end).astype(int)
+
+        # Week of year
+        df['week_of_year'] = df[date_column].dt.isocalendar().week
+
+        # Payday indicators (15th and last day of month - high bakery traffic)
+        df['is_payday'] = ((df['day_of_month'] == 15) | df[date_column].dt.is_month_end).astype(int)
+
+        # Add to feature list
+        for col in ['month', 'quarter', 'day_of_month', 'is_month_start', 'is_month_end',
+                    'week_of_year', 'is_payday']:
+            if col not in self.feature_columns:
+                self.feature_columns.append(col)
+
+        return df
+
+    def add_interaction_features(self, df: pd.DataFrame) -> pd.DataFrame:
+        """
+        Add interaction features between variables.
+
+        Args:
+            df: DataFrame with base features
+
+        Returns:
+            DataFrame with interaction features
+        """
+        df = df.copy()
+
+        # Weekend × Temperature (people buy more cold drinks in hot weekends)
+        if 'is_weekend' in df.columns and 'temperature' in df.columns:
+            df['weekend_temp_interaction'] = df['is_weekend'] * df['temperature']
+            self.feature_columns.append('weekend_temp_interaction')
+
+        # Rain × Weekend (bad weather reduces weekend traffic)
+        if 'is_weekend' in df.columns and 'precipitation' in df.columns:
+            df['rain_weekend_interaction'] = df['is_weekend'] * (df['precipitation'] > 0).astype(int)
+            self.feature_columns.append('rain_weekend_interaction')
+
+        # Friday × Traffic (high Friday traffic means weekend prep buying)
+        if 'is_friday' in df.columns and 'traffic_volume' in df.columns:
+            df['friday_traffic_interaction'] = df['is_friday'] * df['traffic_volume']
+            self.feature_columns.append('friday_traffic_interaction')
+
+        # Month × Temperature (seasonal temperature patterns)
+        if 'month' in df.columns and 'temperature' in df.columns:
+            df['month_temp_interaction'] = df['month'] * df['temperature']
+            self.feature_columns.append('month_temp_interaction')
+
+        # Payday × Weekend (big shopping days)
+        if 'is_payday' in df.columns and 'is_weekend' in df.columns:
+            df['payday_weekend_interaction'] = df['is_payday'] * df['is_weekend']
+            self.feature_columns.append('payday_weekend_interaction')
+
+        logger.info(f"Added {len([c for c in self.feature_columns if 'interaction' in c])} interaction features")
+        return df
+
+    def add_trend_features(self, df: pd.DataFrame, date_column: str = 'date') -> pd.DataFrame:
+        """
+        Add trend-based features.
+
+        Args:
+            df: DataFrame with date and quantity
+            date_column: Name of date column
+
+        Returns:
+            DataFrame with trend features
+        """
+        df = df.copy()
+
+        # Days since start (linear trend proxy)
+        df['days_since_start'] = (df[date_column] - df[date_column].min()).dt.days
+
+        # Momentum indicators (recent change vs. older change)
+        if 'lag_1_day' in df.columns and 'lag_7_day' in df.columns:
+            df['momentum_1_7'] = df['lag_1_day'] - df['lag_7_day']
+            self.feature_columns.append('momentum_1_7')
+
+        if 'rolling_mean_7d' in df.columns and 'rolling_mean_30d' in df.columns:
+            df['trend_7_30'] = df['rolling_mean_7d'] - df['rolling_mean_30d']
+            self.feature_columns.append('trend_7_30')
+
+        # Velocity (rate of change)
+        if 'lag_1_day' in df.columns and 'lag_7_day' in df.columns:
+            df['velocity_week'] = (df['lag_1_day'] - df['lag_7_day']) / 7
+            self.feature_columns.append('velocity_week')
+
+        self.feature_columns.append('days_since_start')
+
+        return df
+
+    def add_cyclical_encoding(self, df: pd.DataFrame) -> pd.DataFrame:
+        """
+        Add cyclical encoding for periodic features (day_of_week, month).
+        Helps models understand that Monday follows Sunday, December follows January.
+
+        Args:
+            df: DataFrame with day_of_week and month columns
+
+        Returns:
+            DataFrame with cyclical features
+        """
+        df = df.copy()
+
+        # Day of week cyclical encoding
+        if 'day_of_week' in df.columns:
+            df['day_of_week_sin'] = np.sin(2 * np.pi * df['day_of_week'] / 7)
+            df['day_of_week_cos'] = np.cos(2 * np.pi * df['day_of_week'] / 7)
+            self.feature_columns.extend(['day_of_week_sin', 'day_of_week_cos'])
+
+        # Month cyclical encoding
+        if 'month' in df.columns:
+            df['month_sin'] = np.sin(2 * np.pi * df['month'] / 12)
+            df['month_cos'] = np.cos(2 * np.pi * df['month'] / 12)
+            self.feature_columns.extend(['month_sin', 'month_cos'])
+
+        logger.info("Added cyclical encoding for temporal features")
+        return df
+
+    def create_all_features(
+        self,
+        df: pd.DataFrame,
+        date_column: str = 'date',
+        include_lags: bool = True,
+        include_rolling: bool = True,
+        include_interactions: bool = True,
+        include_cyclical: bool = True
+    ) -> pd.DataFrame:
+        """
+        Create all enhanced features in one go.
+
+        Args:
+            df: DataFrame with base data
+            date_column: Name of date column
+            include_lags: Whether to include lagged features
+            include_rolling: Whether to include rolling statistics
+            include_interactions: Whether to include interaction features
+            include_cyclical: Whether to include cyclical encoding
+
+        Returns:
+            DataFrame with all enhanced features
+        """
+        logger.info("Creating comprehensive feature set for hybrid model")
+
+        # Reset feature list
+        self.feature_columns = []
+
+        # Day of week and calendar features (always needed)
+        df = self.add_day_of_week_features(df, date_column)
+        df = self.add_calendar_enhanced_features(df, date_column)
+
+        # Optional features
+        if include_lags:
+            df = self.add_lagged_features(df)
+
+        if include_rolling:
+            df = self.add_rolling_features(df)
+
+        if include_interactions:
+            df = self.add_interaction_features(df)
+
+        if include_cyclical:
+            df = self.add_cyclical_encoding(df)
+
+        # Trend features (depends on lags and rolling)
+        if include_lags or include_rolling:
+            df = self.add_trend_features(df, date_column)
+
+        logger.info(f"Created {len(self.feature_columns)} enhanced features for hybrid model")
+
+        return df
+
+    def get_feature_columns(self) -> List[str]:
+        """Get list of all created feature column names."""
+        return self.feature_columns.copy()
+
+    def fill_na_values(self, df: pd.DataFrame, strategy: str = 'forward_backward') -> pd.DataFrame:
+        """
+        Fill NA values in lagged and rolling features.
+
+        Args:
+            df: DataFrame with potential NA values
+            strategy: 'forward_backward', 'zero', 'mean'
+
+        Returns:
+            DataFrame with filled NA values
+        """
+        df = df.copy()
+
+        if strategy == 'forward_backward':
+            # Forward fill first (use previous values)
+            df = df.fillna(method='ffill')
+            # Backward fill remaining (beginning of series)
+            df = df.fillna(method='bfill')
+
+        elif strategy == 'zero':
+            df = df.fillna(0)
+
+        elif strategy == 'mean':
+            df = df.fillna(df.mean())
+
+        return df

services/training/app/ml/event_feature_generator.py

@@ -0,0 +1,253 @@
+"""
+Event Feature Generator
+Converts calendar events into features for demand forecasting
+"""
+
+import pandas as pd
+import numpy as np
+from typing import List, Dict, Any, Optional
+from datetime import date, timedelta
+import structlog
+
+logger = structlog.get_logger()
+
+
+class EventFeatureGenerator:
+    """
+    Generate event-related features for demand forecasting.
+
+    Features include:
+    - Binary flags for event presence
+    - Event impact multipliers
+    - Event type indicators
+    - Days until/since major events
+    """
+
+    # Event type impact weights (default multipliers)
+    EVENT_IMPACT_WEIGHTS = {
+        'promotion': 1.3,
+        'festival': 1.8,
+        'holiday': 0.7,  # Bakeries often close or have reduced demand
+        'weather_event': 0.8,  # Bad weather reduces foot traffic
+        'school_break': 1.2,
+        'sport_event': 1.4,
+        'market': 1.5,
+        'concert': 1.3,
+        'local_event': 1.2
+    }
+
+    def __init__(self):
+        pass
+
+    def generate_event_features(
+        self,
+        dates: pd.DatetimeIndex,
+        events: List[Dict[str, Any]]
+    ) -> pd.DataFrame:
+        """
+        Generate event features for given dates.
+
+        Args:
+            dates: Dates to generate features for
+            events: List of event dictionaries with keys:
+                - event_date: date
+                - event_type: str
+                - impact_multiplier: float (optional)
+                - event_name: str
+
+        Returns:
+            DataFrame with event features
+        """
+        df = pd.DataFrame({'date': dates})
+
+        # Initialize feature columns
+        df['has_event'] = 0
+        df['event_impact'] = 1.0  # Neutral impact
+        df['is_promotion'] = 0
+        df['is_festival'] = 0
+        df['is_local_event'] = 0
+        df['days_to_next_event'] = 365
+        df['days_since_last_event'] = 365
+
+        if not events:
+            logger.debug("No events provided, returning default features")
+            return df
+
+        # Convert events to DataFrame for easier processing
+        events_df = pd.DataFrame(events)
+        events_df['event_date'] = pd.to_datetime(events_df['event_date'])
+
+        for idx, row in df.iterrows():
+            current_date = pd.to_datetime(row['date'])
+
+            # Check if there's an event on this date
+            day_events = events_df[events_df['event_date'] == current_date]
+
+            if not day_events.empty:
+                df.at[idx, 'has_event'] = 1
+
+                # Use custom impact multiplier if provided, else use default
+                if 'impact_multiplier' in day_events.columns and not day_events['impact_multiplier'].isna().all():
+                    impact = day_events['impact_multiplier'].max()
+                else:
+                    # Use default impact based on event type
+                    event_types = day_events['event_type'].tolist()
+                    impacts = [self.EVENT_IMPACT_WEIGHTS.get(et, 1.0) for et in event_types]
+                    impact = max(impacts)
+
+                df.at[idx, 'event_impact'] = impact
+
+                # Set event type flags
+                event_types = day_events['event_type'].tolist()
+                if 'promotion' in event_types:
+                    df.at[idx, 'is_promotion'] = 1
+                if 'festival' in event_types:
+                    df.at[idx, 'is_festival'] = 1
+                if 'local_event' in event_types or 'market' in event_types:
+                    df.at[idx, 'is_local_event'] = 1
+
+            # Calculate days to/from nearest event
+            future_events = events_df[events_df['event_date'] > current_date]
+            if not future_events.empty:
+                next_event_date = future_events['event_date'].min()
+                df.at[idx, 'days_to_next_event'] = (next_event_date - current_date).days
+
+            past_events = events_df[events_df['event_date'] < current_date]
+            if not past_events.empty:
+                last_event_date = past_events['event_date'].max()
+                df.at[idx, 'days_since_last_event'] = (current_date - last_event_date).days
+
+        # Cap days values at 365
+        df['days_to_next_event'] = df['days_to_next_event'].clip(upper=365)
+        df['days_since_last_event'] = df['days_since_last_event'].clip(upper=365)
+
+        logger.debug("Generated event features",
+                    total_days=len(df),
+                    days_with_events=df['has_event'].sum())
+
+        return df
+
+    def add_event_features_to_forecast_data(
+        self,
+        forecast_data: pd.DataFrame,
+        event_features: pd.DataFrame
+    ) -> pd.DataFrame:
+        """
+        Add event features to forecast input data.
+
+        Args:
+            forecast_data: Existing forecast data with 'date' column
+            event_features: Event features from generate_event_features()
+
+        Returns:
+            Enhanced forecast data with event features
+        """
+        forecast_data = forecast_data.copy()
+        forecast_data['date'] = pd.to_datetime(forecast_data['date'])
+        event_features['date'] = pd.to_datetime(event_features['date'])
+
+        # Merge event features
+        enhanced_data = forecast_data.merge(
+            event_features[[
+                'date', 'has_event', 'event_impact', 'is_promotion',
+                'is_festival', 'is_local_event', 'days_to_next_event',
+                'days_since_last_event'
+            ]],
+            on='date',
+            how='left'
+        )
+
+        # Fill missing with defaults
+        enhanced_data['has_event'].fillna(0, inplace=True)
+        enhanced_data['event_impact'].fillna(1.0, inplace=True)
+        enhanced_data['is_promotion'].fillna(0, inplace=True)
+        enhanced_data['is_festival'].fillna(0, inplace=True)
+        enhanced_data['is_local_event'].fillna(0, inplace=True)
+        enhanced_data['days_to_next_event'].fillna(365, inplace=True)
+        enhanced_data['days_since_last_event'].fillna(365, inplace=True)
+
+        return enhanced_data
+
+    def get_event_summary(self, events: List[Dict[str, Any]]) -> Dict[str, Any]:
+        """
+        Get summary statistics about events.
+
+        Args:
+            events: List of event dictionaries
+
+        Returns:
+            Summary dict with counts by type, avg impact, etc.
+        """
+        if not events:
+            return {
+                'total_events': 0,
+                'events_by_type': {},
+                'avg_impact': 1.0
+            }
+
+        events_df = pd.DataFrame(events)
+
+        summary = {
+            'total_events': len(events),
+            'events_by_type': events_df['event_type'].value_counts().to_dict(),
+            'date_range': {
+                'start': events_df['event_date'].min().isoformat() if not events_df.empty else None,
+                'end': events_df['event_date'].max().isoformat() if not events_df.empty else None
+            }
+        }
+
+        if 'impact_multiplier' in events_df.columns:
+            summary['avg_impact'] = float(events_df['impact_multiplier'].mean())
+
+        return summary
+
+
+def create_event_calendar_features(
+    dates: pd.DatetimeIndex,
+    tenant_id: str,
+    event_repository = None
+) -> pd.DataFrame:
+    """
+    Convenience function to fetch events from database and generate features.
+
+    Args:
+        dates: Dates to generate features for
+        tenant_id: Tenant UUID
+        event_repository: EventRepository instance (optional)
+
+    Returns:
+        DataFrame with event features
+    """
+    if event_repository is None:
+        logger.warning("No event repository provided, using empty events")
+        events = []
+    else:
+        # Fetch events from database
+        from datetime import date
+        start_date = dates.min().date()
+        end_date = dates.max().date()
+
+        try:
+            import asyncio
+            from uuid import UUID
+
+            loop = asyncio.get_event_loop()
+            events_objects = loop.run_until_complete(
+                event_repository.get_events_by_date_range(
+                    tenant_id=UUID(tenant_id),
+                    start_date=start_date,
+                    end_date=end_date,
+                    confirmed_only=False
+                )
+            )
+
+            # Convert to dict format
+            events = [event.to_dict() for event in events_objects]
+
+        except Exception as e:
+            logger.error(f"Failed to fetch events from database: {e}")
+            events = []
+
+    # Generate features
+    generator = EventFeatureGenerator()
+    return generator.generate_event_features(dates, events)

services/training/app/ml/hybrid_trainer.py

@@ -0,0 +1,447 @@
+"""
+Hybrid Prophet + XGBoost Trainer
+Combines Prophet's seasonality modeling with XGBoost's pattern learning
+"""
+
+import pandas as pd
+import numpy as np
+from typing import Dict, List, Any, Optional, Tuple
+import structlog
+from datetime import datetime
+import joblib
+from sklearn.metrics import mean_absolute_error, mean_squared_error, mean_absolute_percentage_error
+from sklearn.model_selection import TimeSeriesSplit
+import warnings
+warnings.filterwarnings('ignore')
+
+# Import XGBoost
+try:
+    import xgboost as xgb
+except ImportError:
+    raise ImportError("XGBoost not installed. Run: pip install xgboost")
+
+from app.ml.prophet_manager import BakeryProphetManager
+from app.ml.enhanced_features import AdvancedFeatureEngineer
+
+logger = structlog.get_logger()
+
+
+class HybridProphetXGBoost:
+    """
+    Hybrid forecasting model combining Prophet and XGBoost.
+
+    Approach:
+    1. Train Prophet on historical data (captures trend, seasonality, holidays)
+    2. Calculate residuals (actual - prophet_prediction)
+    3. Train XGBoost on residuals using enhanced features
+    4. Final prediction = prophet_prediction + xgboost_residual_prediction
+
+    Benefits:
+    - Prophet handles seasonality, holidays, trends
+    - XGBoost captures complex patterns Prophet misses
+    - Maintains Prophet's interpretability
+    - Improves accuracy by 10-25% over Prophet alone
+    """
+
+    def __init__(self, database_manager=None):
+        self.prophet_manager = BakeryProphetManager(database_manager)
+        self.feature_engineer = AdvancedFeatureEngineer()
+        self.xgb_model = None
+        self.feature_columns = []
+        self.prophet_model_data = None
+
+    async def train_hybrid_model(
+        self,
+        tenant_id: str,
+        inventory_product_id: str,
+        df: pd.DataFrame,
+        job_id: str,
+        validation_split: float = 0.2
+    ) -> Dict[str, Any]:
+        """
+        Train hybrid Prophet + XGBoost model.
+
+        Args:
+            tenant_id: Tenant identifier
+            inventory_product_id: Product identifier
+            df: Training data (must have 'ds', 'y' and regressor columns)
+            job_id: Training job identifier
+            validation_split: Fraction of data for validation
+
+        Returns:
+            Dictionary with model metadata and performance metrics
+        """
+        logger.info(
+            "Starting hybrid Prophet + XGBoost training",
+            tenant_id=tenant_id,
+            inventory_product_id=inventory_product_id,
+            data_points=len(df)
+        )
+
+        # Step 1: Train Prophet model (base forecaster)
+        logger.info("Step 1: Training Prophet base model")
+        prophet_result = await self.prophet_manager.train_bakery_model(
+            tenant_id=tenant_id,
+            inventory_product_id=inventory_product_id,
+            df=df.copy(),
+            job_id=job_id
+        )
+
+        self.prophet_model_data = prophet_result
+
+        # Step 2: Create enhanced features for XGBoost
+        logger.info("Step 2: Engineering enhanced features for XGBoost")
+        df_enhanced = self._prepare_xgboost_features(df)
+
+        # Step 3: Split into train/validation
+        split_idx = int(len(df_enhanced) * (1 - validation_split))
+        train_df = df_enhanced.iloc[:split_idx].copy()
+        val_df = df_enhanced.iloc[split_idx:].copy()
+
+        logger.info(
+            "Data split",
+            train_samples=len(train_df),
+            val_samples=len(val_df)
+        )
+
+        # Step 4: Get Prophet predictions on training data
+        logger.info("Step 3: Generating Prophet predictions for residual calculation")
+        train_prophet_pred = self._get_prophet_predictions(prophet_result, train_df)
+        val_prophet_pred = self._get_prophet_predictions(prophet_result, val_df)
+
+        # Step 5: Calculate residuals (actual - prophet_prediction)
+        train_residuals = train_df['y'].values - train_prophet_pred
+        val_residuals = val_df['y'].values - val_prophet_pred
+
+        logger.info(
+            "Residuals calculated",
+            train_residual_mean=float(np.mean(train_residuals)),
+            train_residual_std=float(np.std(train_residuals))
+        )
+
+        # Step 6: Prepare feature matrix for XGBoost
+        X_train = train_df[self.feature_columns].values
+        X_val = val_df[self.feature_columns].values
+
+        # Step 7: Train XGBoost on residuals
+        logger.info("Step 4: Training XGBoost on residuals")
+        self.xgb_model = self._train_xgboost(
+            X_train, train_residuals,
+            X_val, val_residuals
+        )
+
+        # Step 8: Evaluate hybrid model
+        logger.info("Step 5: Evaluating hybrid model performance")
+        metrics = self._evaluate_hybrid_model(
+            train_df, val_df,
+            train_prophet_pred, val_prophet_pred,
+            prophet_result
+        )
+
+        # Step 9: Save hybrid model
+        model_data = self._package_hybrid_model(
+            prophet_result, metrics, tenant_id, inventory_product_id
+        )
+
+        logger.info(
+            "Hybrid model training complete",
+            prophet_mape=metrics['prophet_val_mape'],
+            hybrid_mape=metrics['hybrid_val_mape'],
+            improvement_pct=metrics['improvement_percentage']
+        )
+
+        return model_data
+
+    def _prepare_xgboost_features(self, df: pd.DataFrame) -> pd.DataFrame:
+        """
+        Prepare enhanced features for XGBoost.
+
+        Args:
+            df: Base dataframe with 'ds', 'y' and regressor columns
+
+        Returns:
+            DataFrame with all enhanced features
+        """
+        # Rename 'ds' to 'date' for feature engineering
+        df_prep = df.copy()
+        if 'ds' in df_prep.columns:
+            df_prep['date'] = df_prep['ds']
+
+        # Ensure 'quantity' column for feature engineering
+        if 'y' in df_prep.columns:
+            df_prep['quantity'] = df_prep['y']
+
+        # Create all enhanced features
+        df_enhanced = self.feature_engineer.create_all_features(
+            df_prep,
+            date_column='date',
+            include_lags=True,
+            include_rolling=True,
+            include_interactions=True,
+            include_cyclical=True
+        )
+
+        # Fill NA values (from lagged features at beginning)
+        df_enhanced = self.feature_engineer.fill_na_values(df_enhanced)
+
+        # Get feature column list (excluding target and date columns)
+        self.feature_columns = [
+            col for col in self.feature_engineer.get_feature_columns()
+            if col in df_enhanced.columns
+        ]
+
+        # Also include original regressor columns if present
+        regressor_cols = [
+            col for col in df.columns
+            if col not in ['ds', 'y', 'date', 'quantity'] and col in df_enhanced.columns
+        ]
+
+        self.feature_columns.extend(regressor_cols)
+        self.feature_columns = list(set(self.feature_columns))  # Remove duplicates
+
+        logger.info(f"Prepared {len(self.feature_columns)} features for XGBoost")
+
+        return df_enhanced
+
+    def _get_prophet_predictions(
+        self,
+        prophet_result: Dict[str, Any],
+        df: pd.DataFrame
+    ) -> np.ndarray:
+        """
+        Get Prophet predictions for given dataframe.
+
+        Args:
+            prophet_result: Prophet model result from training
+            df: DataFrame with 'ds' column
+
+        Returns:
+            Array of predictions
+        """
+        # Get the Prophet model from result
+        prophet_model = prophet_result.get('model')
+
+        if prophet_model is None:
+            raise ValueError("Prophet model not found in result")
+
+        # Prepare dataframe for prediction
+        pred_df = df[['ds']].copy()
+
+        # Add regressors if present
+        regressor_cols = [col for col in df.columns if col not in ['ds', 'y', 'date', 'quantity']]
+        for col in regressor_cols:
+            if col in df.columns:
+                pred_df[col] = df[col]
+
+        # Get predictions
+        forecast = prophet_model.predict(pred_df)
+
+        return forecast['yhat'].values
+
+    def _train_xgboost(
+        self,
+        X_train: np.ndarray,
+        y_train: np.ndarray,
+        X_val: np.ndarray,
+        y_val: np.ndarray
+    ) -> xgb.XGBRegressor:
+        """
+        Train XGBoost model on residuals.
+
+        Args:
+            X_train: Training features
+            y_train: Training residuals
+            X_val: Validation features
+            y_val: Validation residuals
+
+        Returns:
+            Trained XGBoost model
+        """
+        # XGBoost parameters optimized for residual learning
+        params = {
+            'n_estimators': 100,
+            'max_depth': 3,  # Shallow trees to prevent overfitting
+            'learning_rate': 0.1,
+            'subsample': 0.8,
+            'colsample_bytree': 0.8,
+            'min_child_weight': 3,
+            'reg_alpha': 0.1,  # L1 regularization
+            'reg_lambda': 1.0,  # L2 regularization
+            'objective': 'reg:squarederror',
+            'random_state': 42,
+            'n_jobs': -1
+        }
+
+        # Initialize model
+        model = xgb.XGBRegressor(**params)
+
+        # Train with early stopping
+        model.fit(
+            X_train, y_train,
+            eval_set=[(X_val, y_val)],
+            early_stopping_rounds=10,
+            verbose=False
+        )
+
+        logger.info(
+            "XGBoost training complete",
+            best_iteration=model.best_iteration if hasattr(model, 'best_iteration') else None
+        )
+
+        return model
+
+    def _evaluate_hybrid_model(
+        self,
+        train_df: pd.DataFrame,
+        val_df: pd.DataFrame,
+        train_prophet_pred: np.ndarray,
+        val_prophet_pred: np.ndarray,
+        prophet_result: Dict[str, Any]
+    ) -> Dict[str, float]:
+        """
+        Evaluate hybrid model vs Prophet-only on validation set.
+
+        Args:
+            train_df: Training data
+            val_df: Validation data
+            train_prophet_pred: Prophet predictions on training set
+            val_prophet_pred: Prophet predictions on validation set
+            prophet_result: Prophet training result
+
+        Returns:
+            Dictionary of metrics
+        """
+        # Get actual values
+        train_actual = train_df['y'].values
+        val_actual = val_df['y'].values
+
+        # Get XGBoost predictions on residuals
+        X_train = train_df[self.feature_columns].values
+        X_val = val_df[self.feature_columns].values
+
+        train_xgb_pred = self.xgb_model.predict(X_train)
+        val_xgb_pred = self.xgb_model.predict(X_val)
+
+        # Hybrid predictions = Prophet + XGBoost residual correction
+        train_hybrid_pred = train_prophet_pred + train_xgb_pred
+        val_hybrid_pred = val_prophet_pred + val_xgb_pred
+
+        # Calculate metrics for Prophet-only
+        prophet_train_mae = mean_absolute_error(train_actual, train_prophet_pred)
+        prophet_val_mae = mean_absolute_error(val_actual, val_prophet_pred)
+        prophet_train_mape = mean_absolute_percentage_error(train_actual, train_prophet_pred) * 100
+        prophet_val_mape = mean_absolute_percentage_error(val_actual, val_prophet_pred) * 100
+
+        # Calculate metrics for Hybrid
+        hybrid_train_mae = mean_absolute_error(train_actual, train_hybrid_pred)
+        hybrid_val_mae = mean_absolute_error(val_actual, val_hybrid_pred)
+        hybrid_train_mape = mean_absolute_percentage_error(train_actual, train_hybrid_pred) * 100
+        hybrid_val_mape = mean_absolute_percentage_error(val_actual, val_hybrid_pred) * 100
+
+        # Calculate improvement
+        mae_improvement = ((prophet_val_mae - hybrid_val_mae) / prophet_val_mae) * 100
+        mape_improvement = ((prophet_val_mape - hybrid_val_mape) / prophet_val_mape) * 100
+
+        metrics = {
+            'prophet_train_mae': float(prophet_train_mae),
+            'prophet_val_mae': float(prophet_val_mae),
+            'prophet_train_mape': float(prophet_train_mape),
+            'prophet_val_mape': float(prophet_val_mape),
+            'hybrid_train_mae': float(hybrid_train_mae),
+            'hybrid_val_mae': float(hybrid_val_mae),
+            'hybrid_train_mape': float(hybrid_train_mape),
+            'hybrid_val_mape': float(hybrid_val_mape),
+            'mae_improvement_pct': float(mae_improvement),
+            'mape_improvement_pct': float(mape_improvement),
+            'improvement_percentage': float(mape_improvement)  # Primary metric
+        }
+
+        return metrics
+
+    def _package_hybrid_model(
+        self,
+        prophet_result: Dict[str, Any],
+        metrics: Dict[str, float],
+        tenant_id: str,
+        inventory_product_id: str
+    ) -> Dict[str, Any]:
+        """
+        Package hybrid model for storage.
+
+        Args:
+            prophet_result: Prophet model result
+            metrics: Hybrid model metrics
+            tenant_id: Tenant ID
+            inventory_product_id: Product ID
+
+        Returns:
+            Model package dictionary
+        """
+        return {
+            'model_type': 'hybrid_prophet_xgboost',
+            'prophet_model': prophet_result.get('model'),
+            'xgboost_model': self.xgb_model,
+            'feature_columns': self.feature_columns,
+            'prophet_metrics': {
+                'train_mae': metrics['prophet_train_mae'],
+                'val_mae': metrics['prophet_val_mae'],
+                'train_mape': metrics['prophet_train_mape'],
+                'val_mape': metrics['prophet_val_mape']
+            },
+            'hybrid_metrics': {
+                'train_mae': metrics['hybrid_train_mae'],
+                'val_mae': metrics['hybrid_val_mae'],
+                'train_mape': metrics['hybrid_train_mape'],
+                'val_mape': metrics['hybrid_val_mape']
+            },
+            'improvement_metrics': {
+                'mae_improvement_pct': metrics['mae_improvement_pct'],
+                'mape_improvement_pct': metrics['mape_improvement_pct']
+            },
+            'tenant_id': tenant_id,
+            'inventory_product_id': inventory_product_id,
+            'trained_at': datetime.utcnow().isoformat()
+        }
+
+    async def predict(
+        self,
+        future_df: pd.DataFrame,
+        model_data: Dict[str, Any]
+    ) -> pd.DataFrame:
+        """
+        Make predictions using hybrid model.
+
+        Args:
+            future_df: DataFrame with future dates and regressors
+            model_data: Loaded hybrid model data
+
+        Returns:
+            DataFrame with predictions
+        """
+        # Step 1: Get Prophet predictions
+        prophet_model = model_data['prophet_model']
+        prophet_forecast = prophet_model.predict(future_df)
+
+        # Step 2: Prepare features for XGBoost
+        future_enhanced = self._prepare_xgboost_features(future_df)
+
+        # Step 3: Get XGBoost predictions
+        xgb_model = model_data['xgboost_model']
+        feature_columns = model_data['feature_columns']
+        X_future = future_enhanced[feature_columns].values
+        xgb_pred = xgb_model.predict(X_future)
+
+        # Step 4: Combine predictions
+        hybrid_pred = prophet_forecast['yhat'].values + xgb_pred
+
+        # Step 5: Create result dataframe
+        result = pd.DataFrame({
+            'ds': future_df['ds'],
+            'prophet_yhat': prophet_forecast['yhat'],
+            'xgb_adjustment': xgb_pred,
+            'yhat': hybrid_pred,
+            'yhat_lower': prophet_forecast['yhat_lower'] + xgb_pred,
+            'yhat_upper': prophet_forecast['yhat_upper'] + xgb_pred
+        })
+
+        return result

services/training/app/ml/model_selector.py

@@ -0,0 +1,242 @@
+"""
+Model Selection System
+Determines whether to use Prophet-only or Hybrid Prophet+XGBoost models
+"""
+
+import pandas as pd
+import numpy as np
+from typing import Dict, Any, Optional
+import structlog
+
+logger = structlog.get_logger()
+
+
+class ModelSelector:
+    """
+    Intelligent model selection based on data characteristics.
+
+    Decision Criteria:
+    - Data size: Hybrid needs more data (min 90 days)
+    - Complexity: High variance benefits from XGBoost
+    - Seasonality strength: Weak seasonality benefits from XGBoost
+    - Historical performance: Compare models on validation set
+    """
+
+    # Thresholds for model selection
+    MIN_DATA_POINTS_HYBRID = 90  # Minimum data points for hybrid
+    HIGH_VARIANCE_THRESHOLD = 0.5  # CV > 0.5 suggests complex patterns
+    LOW_SEASONALITY_THRESHOLD = 0.3  # Weak seasonal patterns
+    HYBRID_IMPROVEMENT_THRESHOLD = 0.05  # 5% MAPE improvement to justify hybrid
+
+    def __init__(self):
+        pass
+
+    def select_model_type(
+        self,
+        df: pd.DataFrame,
+        product_category: str = "unknown",
+        force_prophet: bool = False,
+        force_hybrid: bool = False
+    ) -> str:
+        """
+        Select best model type based on data characteristics.
+
+        Args:
+            df: Training data with 'y' column
+            product_category: Product category (bread, pastries, etc.)
+            force_prophet: Force Prophet-only model
+            force_hybrid: Force hybrid model
+
+        Returns:
+            "prophet" or "hybrid"
+        """
+        # Honor forced selections
+        if force_prophet:
+            logger.info("Prophet-only model forced by configuration")
+            return "prophet"
+
+        if force_hybrid:
+            logger.info("Hybrid model forced by configuration")
+            return "hybrid"
+
+        # Check minimum data requirements
+        if len(df) < self.MIN_DATA_POINTS_HYBRID:
+            logger.info(
+                "Insufficient data for hybrid model, using Prophet",
+                data_points=len(df),
+                min_required=self.MIN_DATA_POINTS_HYBRID
+            )
+            return "prophet"
+
+        # Calculate data characteristics
+        characteristics = self._analyze_data_characteristics(df)
+
+        # Decision logic
+        score_hybrid = 0
+        score_prophet = 0
+
+        # Factor 1: Data complexity (variance)
+        if characteristics['coefficient_of_variation'] > self.HIGH_VARIANCE_THRESHOLD:
+            score_hybrid += 2
+            logger.debug("High variance detected, favoring hybrid", cv=characteristics['coefficient_of_variation'])
+        else:
+            score_prophet += 1
+
+        # Factor 2: Seasonality strength
+        if characteristics['seasonality_strength'] < self.LOW_SEASONALITY_THRESHOLD:
+            score_hybrid += 2
+            logger.debug("Weak seasonality detected, favoring hybrid", strength=characteristics['seasonality_strength'])
+        else:
+            score_prophet += 1
+
+        # Factor 3: Data size (more data = better for hybrid)
+        if len(df) > 180:
+            score_hybrid += 1
+        elif len(df) < 120:
+            score_prophet += 1
+
+        # Factor 4: Product category considerations
+        if product_category in ['seasonal', 'cakes']:
+            # Event-driven products benefit from XGBoost pattern learning
+            score_hybrid += 1
+        elif product_category in ['bread', 'savory']:
+            # Stable products work well with Prophet
+            score_prophet += 1
+
+        # Factor 5: Zero ratio (sparse data)
+        if characteristics['zero_ratio'] > 0.3:
+            # High zero ratio suggests difficult forecasting, hybrid might help
+            score_hybrid += 1
+
+        # Make decision
+        selected_model = "hybrid" if score_hybrid > score_prophet else "prophet"
+
+        logger.info(
+            "Model selection complete",
+            selected_model=selected_model,
+            score_hybrid=score_hybrid,
+            score_prophet=score_prophet,
+            data_points=len(df),
+            cv=characteristics['coefficient_of_variation'],
+            seasonality=characteristics['seasonality_strength'],
+            category=product_category
+        )
+
+        return selected_model
+
+    def _analyze_data_characteristics(self, df: pd.DataFrame) -> Dict[str, float]:
+        """
+        Analyze time series characteristics.
+
+        Args:
+            df: DataFrame with 'y' column (sales data)
+
+        Returns:
+            Dictionary with data characteristics
+        """
+        y = df['y'].values
+
+        # Coefficient of variation
+        cv = np.std(y) / np.mean(y) if np.mean(y) > 0 else 0
+
+        # Zero ratio
+        zero_ratio = (y == 0).sum() / len(y)
+
+        # Seasonality strength (simple proxy using rolling std)
+        if len(df) >= 14:
+            rolling_mean = pd.Series(y).rolling(window=7, center=True).mean()
+            seasonality_strength = rolling_mean.std() / (np.std(y) + 1e-6) if np.std(y) > 0 else 0
+        else:
+            seasonality_strength = 0.5  # Default
+
+        # Trend strength
+        if len(df) >= 30:
+            from scipy import stats
+            x = np.arange(len(y))
+            slope, _, r_value, _, _ = stats.linregress(x, y)
+            trend_strength = abs(r_value)
+        else:
+            trend_strength = 0
+
+        return {
+            'coefficient_of_variation': float(cv),
+            'zero_ratio': float(zero_ratio),
+            'seasonality_strength': float(seasonality_strength),
+            'trend_strength': float(trend_strength),
+            'mean': float(np.mean(y)),
+            'std': float(np.std(y))
+        }
+
+    def compare_models(
+        self,
+        prophet_metrics: Dict[str, float],
+        hybrid_metrics: Dict[str, float]
+    ) -> str:
+        """
+        Compare Prophet and Hybrid model performance.
+
+        Args:
+            prophet_metrics: Prophet model metrics (with 'mape' key)
+            hybrid_metrics: Hybrid model metrics (with 'mape' key)
+
+        Returns:
+            "prophet" or "hybrid" based on better performance
+        """
+        prophet_mape = prophet_metrics.get('mape', float('inf'))
+        hybrid_mape = hybrid_metrics.get('mape', float('inf'))
+
+        # Calculate improvement
+        if prophet_mape > 0:
+            improvement = (prophet_mape - hybrid_mape) / prophet_mape
+        else:
+            improvement = 0
+
+        # Hybrid must improve by at least threshold to justify complexity
+        if improvement >= self.HYBRID_IMPROVEMENT_THRESHOLD:
+            logger.info(
+                "Hybrid model selected based on performance",
+                prophet_mape=prophet_mape,
+                hybrid_mape=hybrid_mape,
+                improvement=f"{improvement*100:.1f}%"
+            )
+            return "hybrid"
+        else:
+            logger.info(
+                "Prophet model selected (hybrid improvement insufficient)",
+                prophet_mape=prophet_mape,
+                hybrid_mape=hybrid_mape,
+                improvement=f"{improvement*100:.1f}%"
+            )
+            return "prophet"
+
+
+def should_use_hybrid_model(
+    df: pd.DataFrame,
+    product_category: str = "unknown",
+    tenant_settings: Dict[str, Any] = None
+) -> bool:
+    """
+    Convenience function to determine if hybrid model should be used.
+
+    Args:
+        df: Training data
+        product_category: Product category
+        tenant_settings: Optional tenant-specific settings
+
+    Returns:
+        True if hybrid model should be used, False otherwise
+    """
+    selector = ModelSelector()
+
+    # Check tenant settings
+    force_prophet = tenant_settings.get('force_prophet_only', False) if tenant_settings else False
+    force_hybrid = tenant_settings.get('force_hybrid', False) if tenant_settings else False
+
+    selected = selector.select_model_type(
+        df=df,
+        product_category=product_category,
+        force_prophet=force_prophet,
+        force_hybrid=force_hybrid
+    )
+
+    return selected == "hybrid"

services/training/app/ml/product_categorizer.py

@@ -0,0 +1,361 @@
+"""
+Product Categorization System
+Classifies bakery products into categories for category-specific forecasting
+"""
+
+import pandas as pd
+import numpy as np
+from typing import Dict, List, Optional, Tuple
+from enum import Enum
+import structlog
+
+logger = structlog.get_logger()
+
+
+class ProductCategory(str, Enum):
+    """Product categories for bakery items"""
+    BREAD = "bread"
+    PASTRIES = "pastries"
+    CAKES = "cakes"
+    DRINKS = "drinks"
+    SEASONAL = "seasonal"
+    SAVORY = "savory"
+    UNKNOWN = "unknown"
+
+
+class ProductCategorizer:
+    """
+    Automatic product categorization based on product name and sales patterns.
+
+    Categories have different characteristics:
+    - BREAD: Daily staple, high volume, consistent demand, short shelf life (1 day)
+    - PASTRIES: Morning peak, weekend boost, medium shelf life (2-3 days)
+    - CAKES: Event-driven, weekends, advance orders, longer shelf life (3-5 days)
+    - DRINKS: Weather-dependent, hot/cold seasonal patterns
+    - SEASONAL: Holiday-specific (roscón, panettone, etc.)
+    - SAVORY: Lunch peak, weekday focus
+    """
+
+    def __init__(self):
+        # Keywords for automatic classification
+        self.category_keywords = {
+            ProductCategory.BREAD: [
+                'pan', 'baguette', 'hogaza', 'chapata', 'integral', 'centeno',
+                'bread', 'loaf', 'barra', 'molde', 'candeal'
+            ],
+            ProductCategory.PASTRIES: [
+                'croissant', 'napolitana', 'palmera', 'ensaimada', 'magdalena',
+                'bollo', 'brioche', 'suizo', 'caracola', 'donut', 'berlina'
+            ],
+            ProductCategory.CAKES: [
+                'tarta', 'pastel', 'bizcocho', 'cake', 'torta', 'milhojas',
+                'saint honoré', 'selva negra', 'tres leches'
+            ],
+            ProductCategory.DRINKS: [
+                'café', 'coffee', 'té', 'tea', 'zumo', 'juice', 'batido',
+                'smoothie', 'refresco', 'agua', 'water'
+            ],
+            ProductCategory.SEASONAL: [
+                'roscón', 'panettone', 'turrón', 'polvorón', 'mona de pascua',
+                'huevo de pascua', 'buñuelo', 'torrija'
+            ],
+            ProductCategory.SAVORY: [
+                'empanada', 'quiche', 'pizza', 'focaccia', 'salado', 'bocadillo',
+                'sandwich', 'croqueta', 'hojaldre salado'
+            ]
+        }
+
+    def categorize_product(
+        self,
+        product_name: str,
+        product_id: str = None,
+        sales_data: pd.DataFrame = None
+    ) -> ProductCategory:
+        """
+        Categorize a product based on name and optional sales patterns.
+
+        Args:
+            product_name: Product name
+            product_id: Optional product ID
+            sales_data: Optional historical sales data for pattern analysis
+
+        Returns:
+            ProductCategory enum
+        """
+        # First try keyword matching
+        category = self._categorize_by_keywords(product_name)
+
+        if category != ProductCategory.UNKNOWN:
+            logger.info(f"Product categorized by keywords",
+                       product=product_name,
+                       category=category.value)
+            return category
+
+        # If no keyword match and we have sales data, analyze patterns
+        if sales_data is not None and len(sales_data) > 30:
+            category = self._categorize_by_sales_pattern(product_name, sales_data)
+            logger.info(f"Product categorized by sales pattern",
+                       product=product_name,
+                       category=category.value)
+            return category
+
+        logger.warning(f"Could not categorize product, using UNKNOWN",
+                      product=product_name)
+        return ProductCategory.UNKNOWN
+
+    def _categorize_by_keywords(self, product_name: str) -> ProductCategory:
+        """Categorize by matching keywords in product name"""
+        product_name_lower = product_name.lower()
+
+        # Check each category's keywords
+        for category, keywords in self.category_keywords.items():
+            for keyword in keywords:
+                if keyword in product_name_lower:
+                    return category
+
+        return ProductCategory.UNKNOWN
+
+    def _categorize_by_sales_pattern(
+        self,
+        product_name: str,
+        sales_data: pd.DataFrame
+    ) -> ProductCategory:
+        """
+        Categorize by analyzing sales patterns.
+
+        Patterns:
+        - BREAD: Consistent daily sales, low variance
+        - PASTRIES: Weekend boost, morning peak
+        - CAKES: Weekend spike, event correlation
+        - DRINKS: Temperature correlation
+        - SEASONAL: Concentrated in specific months
+        - SAVORY: Weekday focus, lunch peak
+        """
+        try:
+            # Ensure we have required columns
+            if 'date' not in sales_data.columns or 'quantity' not in sales_data.columns:
+                return ProductCategory.UNKNOWN
+
+            sales_data = sales_data.copy()
+            sales_data['date'] = pd.to_datetime(sales_data['date'])
+            sales_data['day_of_week'] = sales_data['date'].dt.dayofweek
+            sales_data['month'] = sales_data['date'].dt.month
+            sales_data['is_weekend'] = sales_data['day_of_week'].isin([5, 6])
+
+            # Calculate pattern metrics
+            weekend_avg = sales_data[sales_data['is_weekend']]['quantity'].mean()
+            weekday_avg = sales_data[~sales_data['is_weekend']]['quantity'].mean()
+            overall_avg = sales_data['quantity'].mean()
+            cv = sales_data['quantity'].std() / overall_avg if overall_avg > 0 else 0
+
+            # Weekend ratio
+            weekend_ratio = weekend_avg / weekday_avg if weekday_avg > 0 else 1.0
+
+            # Seasonal concentration (Gini coefficient for months)
+            monthly_sales = sales_data.groupby('month')['quantity'].sum()
+            seasonal_concentration = self._gini_coefficient(monthly_sales.values)
+
+            # Decision rules based on patterns
+            if seasonal_concentration > 0.6:
+                # High concentration in specific months = seasonal
+                return ProductCategory.SEASONAL
+
+            elif cv < 0.3 and weekend_ratio < 1.2:
+                # Low variance, consistent daily = bread
+                return ProductCategory.BREAD
+
+            elif weekend_ratio > 1.5:
+                # Strong weekend boost = cakes
+                return ProductCategory.CAKES
+
+            elif weekend_ratio > 1.2:
+                # Moderate weekend boost = pastries
+                return ProductCategory.PASTRIES
+
+            elif weekend_ratio < 0.9:
+                # Weekday focus = savory
+                return ProductCategory.SAVORY
+
+            else:
+                return ProductCategory.UNKNOWN
+
+        except Exception as e:
+            logger.error(f"Error analyzing sales pattern: {e}")
+            return ProductCategory.UNKNOWN
+
+    def _gini_coefficient(self, values: np.ndarray) -> float:
+        """Calculate Gini coefficient for concentration measurement"""
+        if len(values) == 0:
+            return 0.0
+
+        sorted_values = np.sort(values)
+        n = len(values)
+        cumsum = np.cumsum(sorted_values)
+
+        # Gini coefficient formula
+        return (2 * np.sum((np.arange(1, n + 1) * sorted_values))) / (n * cumsum[-1]) - (n + 1) / n
+
+    def get_category_characteristics(self, category: ProductCategory) -> Dict[str, any]:
+        """
+        Get forecasting characteristics for a category.
+
+        Returns hyperparameters and settings specific to the category.
+        """
+        characteristics = {
+            ProductCategory.BREAD: {
+                "shelf_life_days": 1,
+                "demand_stability": "high",
+                "seasonality_strength": "low",
+                "weekend_factor": 0.95,  # Slightly lower on weekends
+                "holiday_factor": 0.7,   # Much lower on holidays
+                "weather_sensitivity": "low",
+                "prophet_params": {
+                    "seasonality_mode": "additive",
+                    "yearly_seasonality": False,
+                    "weekly_seasonality": True,
+                    "daily_seasonality": False,
+                    "changepoint_prior_scale": 0.01,  # Very stable
+                    "seasonality_prior_scale": 5.0
+                }
+            },
+            ProductCategory.PASTRIES: {
+                "shelf_life_days": 2,
+                "demand_stability": "medium",
+                "seasonality_strength": "medium",
+                "weekend_factor": 1.3,   # Boost on weekends
+                "holiday_factor": 1.1,   # Slight boost on holidays
+                "weather_sensitivity": "medium",
+                "prophet_params": {
+                    "seasonality_mode": "multiplicative",
+                    "yearly_seasonality": True,
+                    "weekly_seasonality": True,
+                    "daily_seasonality": False,
+                    "changepoint_prior_scale": 0.05,
+                    "seasonality_prior_scale": 10.0
+                }
+            },
+            ProductCategory.CAKES: {
+                "shelf_life_days": 4,
+                "demand_stability": "low",
+                "seasonality_strength": "high",
+                "weekend_factor": 2.0,   # Large weekend boost
+                "holiday_factor": 1.5,   # Holiday boost
+                "weather_sensitivity": "low",
+                "prophet_params": {
+                    "seasonality_mode": "multiplicative",
+                    "yearly_seasonality": True,
+                    "weekly_seasonality": True,
+                    "daily_seasonality": False,
+                    "changepoint_prior_scale": 0.1,  # More flexible
+                    "seasonality_prior_scale": 15.0
+                }
+            },
+            ProductCategory.DRINKS: {
+                "shelf_life_days": 1,
+                "demand_stability": "medium",
+                "seasonality_strength": "high",
+                "weekend_factor": 1.1,
+                "holiday_factor": 1.2,
+                "weather_sensitivity": "very_high",
+                "prophet_params": {
+                    "seasonality_mode": "multiplicative",
+                    "yearly_seasonality": True,
+                    "weekly_seasonality": True,
+                    "daily_seasonality": False,
+                    "changepoint_prior_scale": 0.08,
+                    "seasonality_prior_scale": 12.0
+                }
+            },
+            ProductCategory.SEASONAL: {
+                "shelf_life_days": 7,
+                "demand_stability": "very_low",
+                "seasonality_strength": "very_high",
+                "weekend_factor": 1.2,
+                "holiday_factor": 3.0,   # Massive holiday boost
+                "weather_sensitivity": "low",
+                "prophet_params": {
+                    "seasonality_mode": "multiplicative",
+                    "yearly_seasonality": True,
+                    "weekly_seasonality": False,
+                    "daily_seasonality": False,
+                    "changepoint_prior_scale": 0.2,  # Very flexible
+                    "seasonality_prior_scale": 20.0
+                }
+            },
+            ProductCategory.SAVORY: {
+                "shelf_life_days": 1,
+                "demand_stability": "medium",
+                "seasonality_strength": "low",
+                "weekend_factor": 0.8,   # Lower on weekends
+                "holiday_factor": 0.6,   # Much lower on holidays
+                "weather_sensitivity": "medium",
+                "prophet_params": {
+                    "seasonality_mode": "additive",
+                    "yearly_seasonality": False,
+                    "weekly_seasonality": True,
+                    "daily_seasonality": False,
+                    "changepoint_prior_scale": 0.03,
+                    "seasonality_prior_scale": 7.0
+                }
+            },
+            ProductCategory.UNKNOWN: {
+                "shelf_life_days": 2,
+                "demand_stability": "medium",
+                "seasonality_strength": "medium",
+                "weekend_factor": 1.0,
+                "holiday_factor": 1.0,
+                "weather_sensitivity": "medium",
+                "prophet_params": {
+                    "seasonality_mode": "multiplicative",
+                    "yearly_seasonality": True,
+                    "weekly_seasonality": True,
+                    "daily_seasonality": False,
+                    "changepoint_prior_scale": 0.05,
+                    "seasonality_prior_scale": 10.0
+                }
+            }
+        }
+
+        return characteristics.get(category, characteristics[ProductCategory.UNKNOWN])
+
+    def batch_categorize(
+        self,
+        products: List[Dict[str, any]],
+        sales_data: pd.DataFrame = None
+    ) -> Dict[str, ProductCategory]:
+        """
+        Categorize multiple products at once.
+
+        Args:
+            products: List of dicts with 'id' and 'name' keys
+            sales_data: Optional sales data with 'inventory_product_id' column
+
+        Returns:
+            Dict mapping product_id to category
+        """
+        results = {}
+
+        for product in products:
+            product_id = product.get('id')
+            product_name = product.get('name', '')
+
+            # Filter sales data for this product if available
+            product_sales = None
+            if sales_data is not None and 'inventory_product_id' in sales_data.columns:
+                product_sales = sales_data[
+                    sales_data['inventory_product_id'] == product_id
+                ].copy()
+
+            category = self.categorize_product(
+                product_name=product_name,
+                product_id=product_id,
+                sales_data=product_sales
+            )
+
+            results[product_id] = category
+
+        logger.info(f"Batch categorization complete",
+                   total_products=len(products),
+                   categories=dict(pd.Series(list(results.values())).value_counts()))
+
+        return results

services/training/app/ml/prophet_manager.py

@@ -19,6 +19,8 @@ import json
 from pathlib import Path
 import math
 import warnings
+import shutil
+import errno
 warnings.filterwarnings('ignore')
 
 from sqlalchemy.ext.asyncio import AsyncSession
@@ -39,6 +41,38 @@ from app.utils.distributed_lock import get_training_lock, LockAcquisitionError
 
 logger = logging.getLogger(__name__)
 
+def check_disk_space(path='/tmp', min_free_gb=1.0):
+    """
+    Check if there's enough disk space available.
+
+    Args:
+        path: Path to check disk space for
+        min_free_gb: Minimum required free space in GB
+
+    Returns:
+        tuple: (bool: has_space, float: free_gb, float: total_gb, float: used_percent)
+    """
+    try:
+        stat = shutil.disk_usage(path)
+        total_gb = stat.total / (1024**3)
+        free_gb = stat.free / (1024**3)
+        used_gb = stat.used / (1024**3)
+        used_percent = (stat.used / stat.total) * 100
+
+        has_space = free_gb >= min_free_gb
+
+        logger.info(f"Disk space check for {path}: "
+                   f"total={total_gb:.2f}GB, free={free_gb:.2f}GB, "
+                   f"used={used_gb:.2f}GB ({used_percent:.1f}%)")
+
+        if used_percent > 85:
+            logger.warning(f"Disk usage is high: {used_percent:.1f}% - this may cause issues")
+
+        return has_space, free_gb, total_gb, used_percent
+    except Exception as e:
+        logger.error(f"Failed to check disk space: {e}")
+        return True, 0, 0, 0  # Assume OK if we can't check
+
 class BakeryProphetManager:
     """
     Simplified Prophet Manager with built-in hyperparameter optimization.
@@ -58,10 +92,27 @@ class BakeryProphetManager:
                                 tenant_id: str,
                                 inventory_product_id: str,
                                 df: pd.DataFrame,
-                                job_id: str) -> Dict[str, Any]:
+                                job_id: str,
+                                product_category: 'ProductCategory' = None,
+                                category_hyperparameters: Dict[str, Any] = None) -> Dict[str, Any]:
         """
         Train a Prophet model with automatic hyperparameter optimization and distributed locking.
+
+        Args:
+            tenant_id: Tenant identifier
+            inventory_product_id: Product identifier
+            df: Training data DataFrame
+            job_id: Training job identifier
+            product_category: Optional product category for category-specific settings
+            category_hyperparameters: Optional category-specific Prophet hyperparameters
         """
+        # Check disk space before starting training
+        has_space, free_gb, total_gb, used_percent = check_disk_space('/tmp', min_free_gb=0.5)
+        if not has_space:
+            error_msg = f"Insufficient disk space: {free_gb:.2f}GB free ({used_percent:.1f}% used). Need at least 0.5GB free."
+            logger.error(error_msg)
+            raise RuntimeError(error_msg)
+
         # Acquire distributed lock to prevent concurrent training of same product
         lock = get_training_lock(tenant_id, inventory_product_id, use_advisory=True)
 
@@ -79,9 +130,33 @@ class BakeryProphetManager:
                     # Get regressor columns
                     regressor_columns = self._extract_regressor_columns(prophet_data)
 
-                    # Automatically optimize hyperparameters
-                    logger.info(f"Optimizing hyperparameters for {inventory_product_id}...")
-                    best_params = await self._optimize_hyperparameters(prophet_data, inventory_product_id, regressor_columns)
+                    # Use category-specific hyperparameters if provided, otherwise optimize
+                    if category_hyperparameters:
+                        logger.info(f"Using category-specific hyperparameters for {inventory_product_id} (category: {product_category.value if product_category else 'unknown'})")
+                        best_params = category_hyperparameters.copy()
+                        use_optimized = False  # Not optimized, but category-specific
+                    else:
+                        # Automatically optimize hyperparameters
+                        logger.info(f"Optimizing hyperparameters for {inventory_product_id}...")
+                        try:
+                            best_params = await self._optimize_hyperparameters(prophet_data, inventory_product_id, regressor_columns)
+                            use_optimized = True
+                        except Exception as opt_error:
+                            logger.warning(f"Hyperparameter optimization failed for {inventory_product_id}: {opt_error}")
+                            logger.warning("Falling back to default Prophet parameters")
+                            # Use conservative default parameters
+                            best_params = {
+                                'changepoint_prior_scale': 0.05,
+                                'seasonality_prior_scale': 10.0,
+                                'holidays_prior_scale': 10.0,
+                                'changepoint_range': 0.8,
+                                'seasonality_mode': 'additive',
+                                'daily_seasonality': False,
+                                'weekly_seasonality': True,
+                                'yearly_seasonality': len(prophet_data) > 365,
+                                'uncertainty_samples': 0  # Disable uncertainty sampling to avoid cmdstan
+                            }
+                            use_optimized = False
 
                     # Create optimized Prophet model
                     model = self._create_optimized_prophet_model(best_params, regressor_columns)
@@ -91,8 +166,38 @@ class BakeryProphetManager:
                         if regressor in prophet_data.columns:
                             model.add_regressor(regressor)
 
-                    # Fit the model
-                    model.fit(prophet_data)
+                    # Set environment variable for cmdstan tmp directory
+                    import os
+                    tmpdir = os.environ.get('TMPDIR', '/tmp/cmdstan')
+                    os.makedirs(tmpdir, mode=0o777, exist_ok=True)
+                    os.environ['TMPDIR'] = tmpdir
+
+                    # Verify tmp directory is writable
+                    test_file = os.path.join(tmpdir, f'test_write_{inventory_product_id}.tmp')
+                    try:
+                        with open(test_file, 'w') as f:
+                            f.write('test')
+                        os.remove(test_file)
+                        logger.debug(f"Verified {tmpdir} is writable")
+                    except Exception as e:
+                        logger.error(f"TMPDIR {tmpdir} is not writable: {e}")
+                        raise RuntimeError(f"Cannot write to {tmpdir}: {e}")
+
+                    # Fit the model with enhanced error handling
+                    try:
+                        logger.info(f"Starting Prophet model fit for {inventory_product_id}")
+                        model.fit(prophet_data)
+                        logger.info(f"Prophet model fit completed successfully for {inventory_product_id}")
+                    except Exception as fit_error:
+                        error_details = {
+                            'error_type': type(fit_error).__name__,
+                            'error_message': str(fit_error),
+                            'errno': getattr(fit_error, 'errno', None),
+                            'tmpdir': tmpdir,
+                            'disk_space': check_disk_space(tmpdir, 0)
+                        }
+                        logger.error(f"Prophet model fit failed for {inventory_product_id}: {error_details}")
+                        raise RuntimeError(f"Prophet training failed: {error_details['error_message']}") from fit_error
 
                     # Calculate enhanced training metrics first
                     training_metrics = await self._calculate_training_metrics(model, prophet_data, best_params)
@@ -104,18 +209,39 @@ class BakeryProphetManager:
                     )
 
                     # Return same format as before, but with optimization info
+                    # Ensure hyperparameters are JSON-serializable
+                    def _serialize_hyperparameters(params):
+                        """Helper to ensure hyperparameters are JSON serializable"""
+                        if not params:
+                            return {}
+                        safe_params = {}
+                        for k, v in params.items():
+                            try:
+                                if isinstance(v, (int, float, str, bool, type(None))):
+                                    safe_params[k] = v
+                                elif hasattr(v, 'item'):  # numpy scalars
+                                    safe_params[k] = v.item()
+                                elif isinstance(v, (list, tuple)):
+                                    safe_params[k] = [x.item() if hasattr(x, 'item') else x for x in v]
+                                else:
+                                    safe_params[k] = float(v) if isinstance(v, (np.integer, np.floating)) else str(v)
+                            except:
+                                safe_params[k] = str(v)  # fallback to string conversion
+                        return safe_params
+
                     model_info = {
                         "model_id": model_id,
                         "model_path": model_path,
                         "type": "prophet_optimized",
                         "training_samples": len(prophet_data),
                         "features": regressor_columns,
-                        "hyperparameters": best_params,
+                        "hyperparameters": _serialize_hyperparameters(best_params),
                         "training_metrics": training_metrics,
+                        "product_category": product_category.value if product_category else "unknown",
                         "trained_at": datetime.now().isoformat(),
                         "data_period": {
-                            "start_date": prophet_data['ds'].min().isoformat(),
-                            "end_date": prophet_data['ds'].max().isoformat(),
+                            "start_date": pd.Timestamp(prophet_data['ds'].min()).isoformat(),
+                            "end_date": pd.Timestamp(prophet_data['ds'].max()).isoformat(),
                             "total_days": len(prophet_data)
                         }
                     }
@@ -238,7 +364,7 @@ class BakeryProphetManager:
                     'daily_seasonality': trial.suggest_categorical('daily_seasonality', [True, False]),
                     'weekly_seasonality': True,  # Always keep weekly
                     'yearly_seasonality': trial.suggest_categorical('yearly_seasonality', [True, False]),
-                    'uncertainty_samples': trial.suggest_int('uncertainty_samples', uncertainty_range[0], uncertainty_range[1])  # ✅ FIX: Adaptive uncertainty sampling
+                    'uncertainty_samples': int(trial.suggest_int('uncertainty_samples', int(uncertainty_range[0]), int(uncertainty_range[1])))  # ✅ FIX: Explicit int casting for all values
                 }
                 
                 # Simple 2-fold cross-validation for speed
@@ -254,17 +380,32 @@ class BakeryProphetManager:
                     
                     try:
                         # Create and train model with adaptive uncertainty sampling
-                        uncertainty_samples = params.get('uncertainty_samples', 200)  # ✅ FIX: Use adaptive uncertainty samples
-                        model = Prophet(**{k: v for k, v in params.items() if k != 'uncertainty_samples'}, 
+                        uncertainty_samples = int(params.get('uncertainty_samples', 200))  # ✅ FIX: Explicit int casting to prevent type errors
+
+                        # Set environment variable for cmdstan tmp directory
+                        import os
+                        tmpdir = os.environ.get('TMPDIR', '/tmp/cmdstan')
+                        os.makedirs(tmpdir, mode=0o777, exist_ok=True)
+                        os.environ['TMPDIR'] = tmpdir
+
+                        model = Prophet(**{k: v for k, v in params.items() if k != 'uncertainty_samples'},
                                       interval_width=0.8, uncertainty_samples=uncertainty_samples)
-                        
+
                         for regressor in regressor_columns:
                             if regressor in train_data.columns:
                                 model.add_regressor(regressor)
-                        
+
                         with warnings.catch_warnings():
                             warnings.simplefilter("ignore")
-                            model.fit(train_data)
+                            try:
+                                model.fit(train_data)
+                            except OSError as e:
+                                # Log errno for "Operation not permitted" errors
+                                if e.errno == errno.EPERM:
+                                    logger.error(f"Permission denied during Prophet fit (errno={e.errno}): {e}")
+                                    logger.error(f"TMPDIR: {tmpdir}, exists: {os.path.exists(tmpdir)}, "
+                                               f"writable: {os.access(tmpdir, os.W_OK)}")
+                                raise
                         
                         # Predict on validation set
                         future_df = model.make_future_dataframe(periods=0)
@@ -317,9 +458,9 @@ class BakeryProphetManager:
         
         logger.info(f"Optimization completed for {inventory_product_id}. Best score: {best_score:.2f}%. "
                    f"Parameters: {best_params}")
-        
-        # ✅ FIX: Log uncertainty sampling configuration for debugging confidence intervals
-        uncertainty_samples = best_params.get('uncertainty_samples', 500)
+
+        # ✅ FIX: Log uncertainty sampling configuration for debugging confidence intervals with explicit int casting
+        uncertainty_samples = int(best_params.get('uncertainty_samples', 500))
         logger.info(f"Prophet model will use {uncertainty_samples} uncertainty samples for {inventory_product_id} "
                    f"(category: {product_category}, zero_ratio: {zero_ratio:.2f})")
         
@@ -363,25 +504,43 @@ class BakeryProphetManager:
     def _create_optimized_prophet_model(self, optimized_params: Dict[str, Any], regressor_columns: List[str]) -> Prophet:
         """Create Prophet model with optimized parameters and adaptive uncertainty sampling"""
         holidays = self._get_spanish_holidays()
-        
-        # Determine uncertainty samples based on data characteristics
-        uncertainty_samples = optimized_params.get('uncertainty_samples', 500)
-        
-        model = Prophet(
-            holidays=holidays if not holidays.empty else None,
-            daily_seasonality=optimized_params.get('daily_seasonality', True),
-            weekly_seasonality=optimized_params.get('weekly_seasonality', True),
-            yearly_seasonality=optimized_params.get('yearly_seasonality', True),
-            seasonality_mode=optimized_params.get('seasonality_mode', 'additive'),
-            changepoint_prior_scale=optimized_params.get('changepoint_prior_scale', 0.05),
-            seasonality_prior_scale=optimized_params.get('seasonality_prior_scale', 10.0),
-            holidays_prior_scale=optimized_params.get('holidays_prior_scale', 10.0),
-            changepoint_range=optimized_params.get('changepoint_range', 0.8),
-            interval_width=0.8,
-            mcmc_samples=0,
-            uncertainty_samples=uncertainty_samples
-        )
-        
+
+        # Determine uncertainty samples based on data characteristics with explicit int casting
+        uncertainty_samples = int(optimized_params.get('uncertainty_samples', 500)) if optimized_params.get('uncertainty_samples') is not None else 500
+
+        # If uncertainty_samples is 0, we're in fallback mode (no cmdstan)
+        if uncertainty_samples == 0:
+            logger.info("Creating Prophet model without uncertainty sampling (fallback mode)")
+            model = Prophet(
+                holidays=holidays if not holidays.empty else None,
+                daily_seasonality=optimized_params.get('daily_seasonality', True),
+                weekly_seasonality=optimized_params.get('weekly_seasonality', True),
+                yearly_seasonality=optimized_params.get('yearly_seasonality', True),
+                seasonality_mode=optimized_params.get('seasonality_mode', 'additive'),
+                changepoint_prior_scale=float(optimized_params.get('changepoint_prior_scale', 0.05)),
+                seasonality_prior_scale=float(optimized_params.get('seasonality_prior_scale', 10.0)),
+                holidays_prior_scale=float(optimized_params.get('holidays_prior_scale', 10.0)),
+                changepoint_range=float(optimized_params.get('changepoint_range', 0.8)),
+                interval_width=0.8,
+                mcmc_samples=0,
+                uncertainty_samples=1  # Minimum value to avoid errors
+            )
+        else:
+            model = Prophet(
+                holidays=holidays if not holidays.empty else None,
+                daily_seasonality=optimized_params.get('daily_seasonality', True),
+                weekly_seasonality=optimized_params.get('weekly_seasonality', True),
+                yearly_seasonality=optimized_params.get('yearly_seasonality', True),
+                seasonality_mode=optimized_params.get('seasonality_mode', 'additive'),
+                changepoint_prior_scale=float(optimized_params.get('changepoint_prior_scale', 0.05)),
+                seasonality_prior_scale=float(optimized_params.get('seasonality_prior_scale', 10.0)),
+                holidays_prior_scale=float(optimized_params.get('holidays_prior_scale', 10.0)),
+                changepoint_range=float(optimized_params.get('changepoint_range', 0.8)),
+                interval_width=0.8,
+                mcmc_samples=0,
+                uncertainty_samples=uncertainty_samples
+            )
+
         return model
     
     # All the existing methods remain the same, just with enhanced metrics
@@ -539,8 +698,8 @@ class BakeryProphetManager:
             "regressor_columns": regressor_columns,
             "training_samples": len(training_data),
             "data_period": {
-                "start_date": training_data['ds'].min().isoformat(),
-                "end_date": training_data['ds'].max().isoformat()
+                "start_date": pd.Timestamp(training_data['ds'].min()).isoformat(),
+                "end_date": pd.Timestamp(training_data['ds'].max()).isoformat()
             },
             "optimized": True,
             "optimized_parameters": optimized_params or {},
@@ -566,6 +725,25 @@ class BakeryProphetManager:
                 # Deactivate previous models for this product
                 await self._deactivate_previous_models_with_session(db_session, tenant_id, inventory_product_id)
                 
+                # Helper to ensure hyperparameters are JSON serializable
+                def _serialize_hyperparameters(params):
+                    if not params:
+                        return {}
+                    safe_params = {}
+                    for k, v in params.items():
+                        try:
+                            if isinstance(v, (int, float, str, bool, type(None))):
+                                safe_params[k] = v
+                            elif hasattr(v, 'item'):  # numpy scalars
+                                safe_params[k] = v.item()
+                            elif isinstance(v, (list, tuple)):
+                                safe_params[k] = [x.item() if hasattr(x, 'item') else x for x in v]
+                            else:
+                                safe_params[k] = float(v) if isinstance(v, (np.integer, np.floating)) else str(v)
+                        except:
+                            safe_params[k] = str(v)  # fallback to string conversion
+                    return safe_params
+
                 # Create new database record
                 db_model = TrainedModel(
                     id=model_id,
@@ -575,22 +753,22 @@ class BakeryProphetManager:
                     job_id=model_id.split('_')[0],  # Extract job_id from model_id
                     model_path=str(model_path),
                     metadata_path=str(metadata_path),
-                    hyperparameters=optimized_params or {},
-                    features_used=regressor_columns,
+                    hyperparameters=_serialize_hyperparameters(optimized_params or {}),
+                    features_used=[str(f) for f in regressor_columns] if regressor_columns else [],
                     is_active=True,
                     is_production=True,  # New models are production-ready
-                    training_start_date=training_data['ds'].min().to_pydatetime().replace(tzinfo=None) if training_data['ds'].min().tz is None else training_data['ds'].min().to_pydatetime(),
-                    training_end_date=training_data['ds'].max().to_pydatetime().replace(tzinfo=None) if training_data['ds'].max().tz is None else training_data['ds'].max().to_pydatetime(),
+                    training_start_date=pd.Timestamp(training_data['ds'].min()).to_pydatetime().replace(tzinfo=None),
+                    training_end_date=pd.Timestamp(training_data['ds'].max()).to_pydatetime().replace(tzinfo=None),
                     training_samples=len(training_data)
                 )
                 
                 # Add training metrics if available
                 if training_metrics:
-                    db_model.mape = training_metrics.get('mape')
-                    db_model.mae = training_metrics.get('mae')
-                    db_model.rmse = training_metrics.get('rmse')
-                    db_model.r2_score = training_metrics.get('r2')
-                    db_model.data_quality_score = training_metrics.get('data_quality_score')
+                    db_model.mape = float(training_metrics.get('mape')) if training_metrics.get('mape') is not None else None
+                    db_model.mae = float(training_metrics.get('mae')) if training_metrics.get('mae') is not None else None
+                    db_model.rmse = float(training_metrics.get('rmse')) if training_metrics.get('rmse') is not None else None
+                    db_model.r2_score = float(training_metrics.get('r2')) if training_metrics.get('r2') is not None else None
+                    db_model.data_quality_score = float(training_metrics.get('data_quality_score')) if training_metrics.get('data_quality_score') is not None else None
                 
                 db_session.add(db_model)
                 await db_session.commit()
@@ -698,7 +876,7 @@ class BakeryProphetManager:
         # Ensure y values are non-negative
         prophet_data['y'] = prophet_data['y'].clip(lower=0)
 
-        logger.info(f"Prepared Prophet data: {len(prophet_data)} rows, date range: {prophet_data['ds'].min()} to {prophet_data['ds'].max()}")
+        logger.info(f"Prepared Prophet data: {len(prophet_data)} rows, date range: {pd.Timestamp(prophet_data['ds'].min())} to {pd.Timestamp(prophet_data['ds'].max())}")
 
         return prophet_data
     
@@ -714,12 +892,69 @@ class BakeryProphetManager:
         logger.info(f"Identified regressor columns: {regressor_columns}")
         return regressor_columns
     
-    def _get_spanish_holidays(self) -> pd.DataFrame:
-        """Get Spanish holidays (unchanged)"""
+    def _get_spanish_holidays(self, region: str = None) -> pd.DataFrame:
+        """
+        Get Spanish holidays dynamically using holidays library.
+        Supports national and regional holidays, including dynamic Easter calculation.
+
+        Args:
+            region: Region code (e.g., 'MD' for Madrid, 'PV' for Basque Country)
+
+        Returns:
+            DataFrame with holiday dates and names
+        """
+        try:
+            import holidays
+
+            holidays_list = []
+            years = range(2020, 2035)  # Extended range for better coverage
+
+            # Get Spanish holidays for each year
+            for year in years:
+                # National holidays
+                spain_holidays = holidays.Spain(years=year, prov=region)
+
+                for date, name in spain_holidays.items():
+                    holidays_list.append({
+                        'holiday': self._normalize_holiday_name(name),
+                        'ds': pd.Timestamp(date),
+                        'lower_window': 0,
+                        'upper_window': 0  # Can be adjusted for multi-day holidays
+                    })
+
+            if holidays_list:
+                holidays_df = pd.DataFrame(holidays_list)
+                # Remove duplicates (some holidays may repeat)
+                holidays_df = holidays_df.drop_duplicates(subset=['ds', 'holiday'])
+                holidays_df = holidays_df.sort_values('ds').reset_index(drop=True)
+
+                logger.info(f"Loaded {len(holidays_df)} Spanish holidays dynamically",
+                           region=region or 'National',
+                           years=f"{min(years)}-{max(years)}")
+
+                return holidays_df
+            else:
+                return pd.DataFrame()
+
+        except Exception as e:
+            logger.warning(f"Could not load Spanish holidays dynamically: {str(e)}")
+            # Fallback to minimal hardcoded holidays
+            return self._get_fallback_holidays()
+
+    def _normalize_holiday_name(self, name: str) -> str:
+        """Normalize holiday name to a consistent format for Prophet"""
+        # Convert to lowercase and replace spaces with underscores
+        normalized = name.lower().replace(' ', '_').replace("'", '')
+        # Remove special characters
+        normalized = ''.join(c for c in normalized if c.isalnum() or c == '_')
+        return normalized
+
+    def _get_fallback_holidays(self) -> pd.DataFrame:
+        """Fallback to basic hardcoded holidays if dynamic loading fails"""
         try:
             holidays_list = []
-            years = range(2020, 2030)
-            
+            years = range(2020, 2035)
+
             for year in years:
                 holidays_list.extend([
                     {'holiday': 'new_year', 'ds': f'{year}-01-01'},
@@ -732,14 +967,10 @@ class BakeryProphetManager:
                     {'holiday': 'immaculate_conception', 'ds': f'{year}-12-08'},
                     {'holiday': 'christmas', 'ds': f'{year}-12-25'}
                 ])
-            
-            if holidays_list:
-                holidays_df = pd.DataFrame(holidays_list)
-                holidays_df['ds'] = pd.to_datetime(holidays_df['ds'])
-                return holidays_df
-            else:
-                return pd.DataFrame()
-                
+
+            holidays_df = pd.DataFrame(holidays_list)
+            holidays_df['ds'] = pd.to_datetime(holidays_df['ds'])
+            return holidays_df
         except Exception as e:
-            logger.warning(f"Could not load Spanish holidays: {str(e)}")
+            logger.error(f"Fallback holidays failed: {e}")
             return pd.DataFrame()

services/training/app/ml/traffic_forecaster.py

@@ -0,0 +1,284 @@
+"""
+Traffic Forecasting System
+Predicts bakery foot traffic using weather and temporal features
+"""
+
+import pandas as pd
+import numpy as np
+from typing import Dict, List, Any, Optional
+from prophet import Prophet
+import structlog
+from datetime import datetime, timedelta
+
+logger = structlog.get_logger()
+
+
+class TrafficForecaster:
+    """
+    Forecast bakery foot traffic using Prophet with weather and temporal features.
+
+    Traffic patterns are influenced by:
+    - Weather: Temperature, precipitation, conditions
+    - Time: Day of week, holidays, season
+    - Special events: Local events, promotions
+    """
+
+    def __init__(self):
+        self.model = None
+        self.is_trained = False
+
+    def train(
+        self,
+        historical_traffic: pd.DataFrame,
+        weather_data: pd.DataFrame = None
+    ) -> Dict[str, Any]:
+        """
+        Train traffic forecasting model.
+
+        Args:
+            historical_traffic: DataFrame with columns ['date', 'traffic_count']
+            weather_data: Optional weather data with columns ['date', 'temperature', 'precipitation', 'condition']
+
+        Returns:
+            Training metrics
+        """
+        try:
+            logger.info("Training traffic forecasting model",
+                       data_points=len(historical_traffic))
+
+            # Prepare Prophet format
+            df = historical_traffic.copy()
+            df = df.rename(columns={'date': 'ds', 'traffic_count': 'y'})
+            df['ds'] = pd.to_datetime(df['ds'])
+            df = df.sort_values('ds')
+
+            # Merge with weather data if available
+            if weather_data is not None:
+                weather_data = weather_data.copy()
+                weather_data['date'] = pd.to_datetime(weather_data['date'])
+                df = df.merge(weather_data, left_on='ds', right_on='date', how='left')
+
+            # Create Prophet model with custom settings for traffic
+            self.model = Prophet(
+                seasonality_mode='multiplicative',
+                yearly_seasonality=True,
+                weekly_seasonality=True,
+                daily_seasonality=False,
+                changepoint_prior_scale=0.05,  # Moderate flexibility
+                seasonality_prior_scale=10.0,
+                holidays_prior_scale=10.0
+            )
+
+            # Add weather regressors if available
+            if 'temperature' in df.columns:
+                self.model.add_regressor('temperature')
+            if 'precipitation' in df.columns:
+                self.model.add_regressor('precipitation')
+            if 'is_rainy' in df.columns:
+                self.model.add_regressor('is_rainy')
+
+            # Add custom holidays for Spain
+            from app.ml.prophet_manager import BakeryProphetManager
+            spanish_holidays = self._get_spanish_holidays(
+                df['ds'].min().year,
+                df['ds'].max().year + 1
+            )
+            self.model.add_country_holidays(country_name='ES')
+
+            # Fit model
+            self.model.fit(df)
+            self.is_trained = True
+
+            # Calculate training metrics
+            predictions = self.model.predict(df)
+            metrics = self._calculate_metrics(df['y'].values, predictions['yhat'].values)
+
+            logger.info("Traffic forecasting model trained successfully",
+                       mape=metrics['mape'],
+                       rmse=metrics['rmse'])
+
+            return metrics
+
+        except Exception as e:
+            logger.error(f"Failed to train traffic forecasting model: {e}")
+            raise
+
+    def predict(
+        self,
+        future_dates: pd.DatetimeIndex,
+        weather_forecast: pd.DataFrame = None
+    ) -> pd.DataFrame:
+        """
+        Predict traffic for future dates.
+
+        Args:
+            future_dates: Dates to predict traffic for
+            weather_forecast: Optional weather forecast data
+
+        Returns:
+            DataFrame with columns ['date', 'predicted_traffic', 'yhat_lower', 'yhat_upper']
+        """
+        if not self.is_trained:
+            raise ValueError("Model not trained. Call train() first.")
+
+        try:
+            # Create future dataframe
+            future = pd.DataFrame({'ds': future_dates})
+
+            # Add weather features if available
+            if weather_forecast is not None:
+                weather_forecast = weather_forecast.copy()
+                weather_forecast['date'] = pd.to_datetime(weather_forecast['date'])
+                future = future.merge(weather_forecast, left_on='ds', right_on='date', how='left')
+
+                # Fill missing weather with defaults
+                if 'temperature' in future.columns:
+                    future['temperature'].fillna(15.0, inplace=True)
+                if 'precipitation' in future.columns:
+                    future['precipitation'].fillna(0.0, inplace=True)
+                if 'is_rainy' in future.columns:
+                    future['is_rainy'].fillna(0, inplace=True)
+
+            # Predict
+            forecast = self.model.predict(future)
+
+            # Format results
+            results = pd.DataFrame({
+                'date': forecast['ds'],
+                'predicted_traffic': forecast['yhat'].clip(lower=0),  # Traffic can't be negative
+                'yhat_lower': forecast['yhat_lower'].clip(lower=0),
+                'yhat_upper': forecast['yhat_upper'].clip(lower=0)
+            })
+
+            logger.info("Traffic predictions generated",
+                       dates=len(results),
+                       avg_traffic=results['predicted_traffic'].mean())
+
+            return results
+
+        except Exception as e:
+            logger.error(f"Failed to predict traffic: {e}")
+            raise
+
+    def _calculate_metrics(self, actual: np.ndarray, predicted: np.ndarray) -> Dict[str, float]:
+        """Calculate forecast accuracy metrics"""
+        mae = np.mean(np.abs(actual - predicted))
+        mse = np.mean((actual - predicted) ** 2)
+        rmse = np.sqrt(mse)
+
+        # MAPE (handle zeros)
+        mask = actual != 0
+        mape = np.mean(np.abs((actual[mask] - predicted[mask]) / actual[mask])) * 100 if mask.any() else 0
+
+        return {
+            'mae': float(mae),
+            'mse': float(mse),
+            'rmse': float(rmse),
+            'mape': float(mape)
+        }
+
+    def _get_spanish_holidays(self, start_year: int, end_year: int) -> pd.DataFrame:
+        """Get Spanish holidays for the date range"""
+        try:
+            import holidays
+
+            es_holidays = holidays.Spain(years=range(start_year, end_year + 1))
+
+            holiday_dates = []
+            holiday_names = []
+
+            for date, name in es_holidays.items():
+                holiday_dates.append(date)
+                holiday_names.append(name)
+
+            return pd.DataFrame({
+                'ds': pd.to_datetime(holiday_dates),
+                'holiday': holiday_names
+            })
+
+        except Exception as e:
+            logger.warning(f"Could not load Spanish holidays: {e}")
+            return pd.DataFrame(columns=['ds', 'holiday'])
+
+
+class TrafficFeatureGenerator:
+    """
+    Generate traffic-related features for demand forecasting.
+    Uses predicted traffic as a feature in product demand models.
+    """
+
+    def __init__(self, traffic_forecaster: TrafficForecaster = None):
+        self.traffic_forecaster = traffic_forecaster or TrafficForecaster()
+
+    def generate_traffic_features(
+        self,
+        dates: pd.DatetimeIndex,
+        weather_forecast: pd.DataFrame = None
+    ) -> pd.DataFrame:
+        """
+        Generate traffic features for given dates.
+
+        Args:
+            dates: Dates to generate features for
+            weather_forecast: Optional weather forecast
+
+        Returns:
+            DataFrame with traffic features
+        """
+        if not self.traffic_forecaster.is_trained:
+            logger.warning("Traffic forecaster not trained, using default traffic values")
+            return pd.DataFrame({
+                'date': dates,
+                'predicted_traffic': 100.0,  # Default baseline
+                'traffic_normalized': 1.0
+            })
+
+        # Predict traffic
+        traffic_predictions = self.traffic_forecaster.predict(dates, weather_forecast)
+
+        # Normalize traffic (0-2 range, 1 = average)
+        mean_traffic = traffic_predictions['predicted_traffic'].mean()
+        traffic_predictions['traffic_normalized'] = (
+            traffic_predictions['predicted_traffic'] / mean_traffic
+        ).clip(0, 2)
+
+        # Add traffic categories
+        traffic_predictions['traffic_category'] = pd.cut(
+            traffic_predictions['predicted_traffic'],
+            bins=[0, 50, 100, 150, np.inf],
+            labels=['low', 'medium', 'high', 'very_high']
+        )
+
+        return traffic_predictions
+
+    def add_traffic_features_to_forecast_data(
+        self,
+        forecast_data: pd.DataFrame,
+        traffic_predictions: pd.DataFrame
+    ) -> pd.DataFrame:
+        """
+        Add traffic features to forecast input data.
+
+        Args:
+            forecast_data: Existing forecast data with 'date' column
+            traffic_predictions: Traffic predictions from generate_traffic_features()
+
+        Returns:
+            Enhanced forecast data with traffic features
+        """
+        forecast_data = forecast_data.copy()
+        forecast_data['date'] = pd.to_datetime(forecast_data['date'])
+        traffic_predictions['date'] = pd.to_datetime(traffic_predictions['date'])
+
+        # Merge traffic features
+        enhanced_data = forecast_data.merge(
+            traffic_predictions[['date', 'predicted_traffic', 'traffic_normalized']],
+            on='date',
+            how='left'
+        )
+
+        # Fill missing with defaults
+        enhanced_data['predicted_traffic'].fillna(100.0, inplace=True)
+        enhanced_data['traffic_normalized'].fillna(1.0, inplace=True)
+
+        return enhanced_data

services/training/app/ml/trainer.py

@@ -14,6 +14,9 @@ import asyncio
 
 from app.ml.data_processor import EnhancedBakeryDataProcessor
 from app.ml.prophet_manager import BakeryProphetManager
+from app.ml.product_categorizer import ProductCategorizer, ProductCategory
+from app.ml.model_selector import ModelSelector
+from app.ml.hybrid_trainer import HybridProphetXGBoost
 from app.services.training_orchestrator import TrainingDataSet
 from app.core.config import settings
 
@@ -49,6 +52,9 @@ class EnhancedBakeryMLTrainer:
         self.database_manager = database_manager or create_database_manager(settings.DATABASE_URL, "training-service")
         self.enhanced_data_processor = EnhancedBakeryDataProcessor(self.database_manager)
         self.prophet_manager = BakeryProphetManager(database_manager=self.database_manager)
+        self.hybrid_trainer = HybridProphetXGBoost(database_manager=self.database_manager)
+        self.model_selector = ModelSelector()
+        self.product_categorizer = ProductCategorizer()
         
     async def _get_repositories(self, session):
         """Initialize repositories with session"""
@@ -169,6 +175,16 @@ class EnhancedBakeryMLTrainer:
                     sales_df, weather_df, traffic_df, products, tenant_id, job_id
                 )
 
+                # Categorize all products for category-specific forecasting
+                logger.info("Categorizing products for optimized forecasting")
+                product_categories = await self._categorize_all_products(
+                    sales_df, processed_data
+                )
+                logger.info("Product categorization complete",
+                           total_products=len(product_categories),
+                           categories_breakdown={cat.value: sum(1 for c in product_categories.values() if c == cat)
+                                               for cat in set(product_categories.values())})
+
                 # Event 2: Data Analysis (20%)
                 # Recalculate time remaining based on elapsed time
                 elapsed_seconds = (datetime.now(timezone.utc) - repos['training_log']._get_start_time(job_id) if hasattr(repos['training_log'], '_get_start_time') else 0) or 0
@@ -202,7 +218,7 @@ class EnhancedBakeryMLTrainer:
                 )
 
                 training_results = await self._train_all_models_enhanced(
-                    tenant_id, processed_data, job_id, repos, progress_tracker
+                    tenant_id, processed_data, job_id, repos, progress_tracker, product_categories
                 )
                 
                 # Calculate overall training summary with enhanced metrics
@@ -269,6 +285,149 @@ class EnhancedBakeryMLTrainer:
 
             raise
     
+    async def train_single_product_model(self,
+                                       tenant_id: str,
+                                       inventory_product_id: str,
+                                       training_data: pd.DataFrame,
+                                       job_id: str = None) -> Dict[str, Any]:
+        """
+        Train a model for a single product using repository pattern.
+        
+        Args:
+            tenant_id: Tenant identifier
+            inventory_product_id: Specific inventory product to train
+            training_data: Prepared training DataFrame for the product
+            job_id: Training job identifier (optional)
+            
+        Returns:
+            Dictionary with model training results
+        """
+        if not job_id:
+            job_id = f"single_product_{tenant_id}_{inventory_product_id}_{uuid.uuid4().hex[:8]}"
+            
+        logger.info("Starting single product model training",
+                   job_id=job_id,
+                   tenant_id=tenant_id,
+                   inventory_product_id=inventory_product_id,
+                   data_points=len(training_data))
+        
+        try:
+            # Get database session and repositories
+            async with self.database_manager.get_session() as db_session:
+                repos = await self._get_repositories(db_session)
+                
+                # Validate input data
+                if training_data.empty or len(training_data) < settings.MIN_TRAINING_DATA_DAYS:
+                    raise ValueError(f"Insufficient training data: need at least {settings.MIN_TRAINING_DATA_DAYS} data points, got {len(training_data)}")
+                
+                # Validate required columns
+                required_columns = ['ds', 'y']
+                missing_cols = [col for col in required_columns if col not in training_data.columns]
+                if missing_cols:
+                    raise ValueError(f"Missing required columns in training data: {missing_cols}")
+                
+                # Create a simple progress tracker for single product
+                from app.services.progress_tracker import ParallelProductProgressTracker
+                progress_tracker = ParallelProductProgressTracker(
+                    job_id=job_id, 
+                    tenant_id=tenant_id, 
+                    total_products=1
+                )
+                
+                # Ensure training data has proper data types before training
+                if 'ds' in training_data.columns:
+                    training_data['ds'] = pd.to_datetime(training_data['ds'])
+                if 'y' in training_data.columns:
+                    training_data['y'] = pd.to_numeric(training_data['y'], errors='coerce')
+                
+                # Remove any rows with NaN values
+                training_data = training_data.dropna()
+                
+                # Train the model using the existing _train_single_product method
+                product_id, result = await self._train_single_product(
+                    tenant_id=tenant_id,
+                    inventory_product_id=inventory_product_id,
+                    product_data=training_data,
+                    job_id=job_id,
+                    repos=repos,
+                    progress_tracker=progress_tracker
+                )
+                
+                logger.info("Single product training completed",
+                           job_id=job_id,
+                           inventory_product_id=inventory_product_id,
+                           result_status=result.get('status'))
+                
+                # Get training metrics and filter out non-numeric values
+                raw_metrics = result.get('model_info', {}).get('training_metrics', {})
+                # Filter metrics to only include numeric values (per Pydantic schema requirement)
+                filtered_metrics = {}
+                for key, value in raw_metrics.items():
+                    if key == 'product_category':
+                        # Skip product_category as it's a string value, not a numeric metric
+                        continue
+                    try:
+                        # Try to convert to float for validation
+                        filtered_metrics[key] = float(value) if value is not None else 0.0
+                    except (ValueError, TypeError):
+                        # Skip non-numeric values
+                        continue
+                
+                # Return appropriate result format
+                return {
+                    "job_id": job_id,
+                    "tenant_id": tenant_id,
+                    "inventory_product_id": inventory_product_id,
+                    "status": result.get('status', 'success'),
+                    "model_id": str(result.get('model_record_id', '')) if result.get('model_record_id') else None,
+                    "training_metrics": filtered_metrics,
+                    "training_time": result.get('training_time_seconds', 0),
+                    "data_points": result.get('data_points', 0),
+                    "message": f"Single product model training {'completed' if result.get('status') != 'error' else 'failed'}"
+                }
+                
+        except Exception as e:
+            logger.error("Single product model training failed",
+                        job_id=job_id,
+                        inventory_product_id=inventory_product_id,
+                        error=str(e))
+            raise
+    
+    def _serialize_scalers(self, scalers: Dict[str, Any]) -> Dict[str, Any]:
+        """
+        Serialize scaler objects to basic Python types that can be stored in database.
+        This prevents issues with storing complex sklearn objects in JSON fields.
+        """
+        if not scalers:
+            return {}
+        
+        serialized = {}
+        for key, value in scalers.items():
+            try:
+                # Convert numpy scalars to Python native types
+                if hasattr(value, 'item'):  # numpy scalars
+                    serialized[key] = value.item()
+                elif isinstance(value, (np.integer, np.floating)):
+                    serialized[key] = value.item()  # Convert numpy types to Python types
+                elif isinstance(value, (int, float, str, bool, type(None))):
+                    serialized[key] = value  # Already basic type
+                elif isinstance(value, (list, tuple)):
+                    # Convert list/tuple elements to basic types
+                    serialized[key] = [v.item() if hasattr(v, 'item') else v for v in value]
+                else:
+                    # For complex objects, try to convert to string representation
+                    # or store as float if it's numeric
+                    try:
+                        serialized[key] = float(value)
+                    except (ValueError, TypeError):
+                        # If all else fails, convert to string
+                        serialized[key] = str(value)
+            except Exception:
+                # If serialization fails, set to None to prevent database errors
+                serialized[key] = None
+        
+        return serialized
+    
     async def _process_all_products_enhanced(self,
                                            sales_df: pd.DataFrame,
                                            weather_df: pd.DataFrame,
@@ -321,12 +480,15 @@ class EnhancedBakeryMLTrainer:
                                    product_data: pd.DataFrame,
                                    job_id: str,
                                    repos: Dict,
-                                   progress_tracker: ParallelProductProgressTracker) -> tuple[str, Dict[str, Any]]:
+                                   progress_tracker: ParallelProductProgressTracker,
+                                   product_category: ProductCategory = ProductCategory.UNKNOWN) -> tuple[str, Dict[str, Any]]:
         """Train a single product model - used for parallel execution with progress aggregation"""
         product_start_time = time.time()
 
         try:
-            logger.info("Training model", inventory_product_id=inventory_product_id)
+            logger.info("Training model",
+                       inventory_product_id=inventory_product_id,
+                       category=product_category.value)
 
             # Check if we have enough data
             if len(product_data) < settings.MIN_TRAINING_DATA_DAYS:
@@ -343,14 +505,58 @@ class EnhancedBakeryMLTrainer:
                               min_required=settings.MIN_TRAINING_DATA_DAYS)
                 return inventory_product_id, result
 
-            # Train the model using Prophet manager
-            model_info = await self.prophet_manager.train_bakery_model(
-                tenant_id=tenant_id,
-                inventory_product_id=inventory_product_id,
+            # Get category-specific hyperparameters
+            category_characteristics = self.product_categorizer.get_category_characteristics(product_category)
+
+            # Determine which model type to use (Prophet vs Hybrid)
+            model_type = self.model_selector.select_model_type(
                 df=product_data,
-                job_id=job_id
+                product_category=product_category.value
             )
 
+            logger.info("Model type selected",
+                       inventory_product_id=inventory_product_id,
+                       model_type=model_type,
+                       category=product_category.value)
+
+            # Train the selected model
+            if model_type == "hybrid":
+                # Train hybrid Prophet + XGBoost model
+                model_info = await self.hybrid_trainer.train_hybrid_model(
+                    tenant_id=tenant_id,
+                    inventory_product_id=inventory_product_id,
+                    df=product_data,
+                    job_id=job_id
+                )
+                model_info['model_type'] = 'hybrid_prophet_xgboost'
+            else:
+                # Train Prophet-only model with category-specific settings
+                model_info = await self.prophet_manager.train_bakery_model(
+                    tenant_id=tenant_id,
+                    inventory_product_id=inventory_product_id,
+                    df=product_data,
+                    job_id=job_id,
+                    product_category=product_category,
+                    category_hyperparameters=category_characteristics.get('prophet_params', {})
+                )
+                model_info['model_type'] = 'prophet_optimized'
+
+            # Filter training metrics to exclude non-numeric values (e.g., product_category)
+            if 'training_metrics' in model_info and model_info['training_metrics']:
+                raw_metrics = model_info['training_metrics']
+                filtered_metrics = {}
+                for key, value in raw_metrics.items():
+                    if key == 'product_category':
+                        # Skip product_category as it's a string value, not a numeric metric
+                        continue
+                    try:
+                        # Try to convert to float for validation
+                        filtered_metrics[key] = float(value) if value is not None else 0.0
+                    except (ValueError, TypeError):
+                        # Skip non-numeric values
+                        continue
+                model_info['training_metrics'] = filtered_metrics
+
             # Store model record using repository
             model_record = await self._create_model_record(
                 repos, tenant_id, inventory_product_id, model_info, job_id, product_data
@@ -366,7 +572,7 @@ class EnhancedBakeryMLTrainer:
             result = {
                 'status': 'success',
                 'model_info': model_info,
-                'model_record_id': model_record.id if model_record else None,
+                'model_record_id': str(model_record.id) if model_record else None,
                 'data_points': len(product_data),
                 'training_time_seconds': time.time() - product_start_time,
                 'trained_at': datetime.now().isoformat()
@@ -403,7 +609,8 @@ class EnhancedBakeryMLTrainer:
                                        processed_data: Dict[str, pd.DataFrame],
                                        job_id: str,
                                        repos: Dict,
-                                       progress_tracker: ParallelProductProgressTracker) -> Dict[str, Any]:
+                                       progress_tracker: ParallelProductProgressTracker,
+                                       product_categories: Dict[str, ProductCategory] = None) -> Dict[str, Any]:
         """Train models with throttled parallel execution and progress tracking"""
         total_products = len(processed_data)
         logger.info(f"Starting throttled parallel training for {total_products} products")
@@ -416,7 +623,8 @@ class EnhancedBakeryMLTrainer:
                 product_data=product_data,
                 job_id=job_id,
                 repos=repos,
-                progress_tracker=progress_tracker
+                progress_tracker=progress_tracker,
+                product_category=product_categories.get(inventory_product_id, ProductCategory.UNKNOWN) if product_categories else ProductCategory.UNKNOWN
             )
             for inventory_product_id, product_data in processed_data.items()
         ]
@@ -478,6 +686,29 @@ class EnhancedBakeryMLTrainer:
                                  processed_data: pd.DataFrame):
         """Create model record using repository"""
         try:
+            # Extract training period from the processed data
+            training_start_date = None
+            training_end_date = None
+            if 'ds' in processed_data.columns and not processed_data.empty:
+                # Ensure ds column is datetime64 before extracting dates (prevents object dtype issues)
+                ds_datetime = pd.to_datetime(processed_data['ds'])
+
+                # Get min/max as pandas Timestamps (guaranteed to work correctly)
+                min_ts = ds_datetime.min()
+                max_ts = ds_datetime.max()
+
+                # Convert to python datetime with timezone removal
+                if pd.notna(min_ts):
+                    training_start_date = pd.Timestamp(min_ts).to_pydatetime().replace(tzinfo=None)
+                if pd.notna(max_ts):
+                    training_end_date = pd.Timestamp(max_ts).to_pydatetime().replace(tzinfo=None)
+            
+            # Ensure features are clean string list
+            try:
+                features_used = [str(col) for col in processed_data.columns]
+            except Exception:
+                features_used = []
+            
             model_data = {
                 "tenant_id": tenant_id,
                 "inventory_product_id": inventory_product_id,
@@ -485,17 +716,20 @@ class EnhancedBakeryMLTrainer:
                 "model_type": "enhanced_prophet",
                 "model_path": model_info.get("model_path"),
                 "metadata_path": model_info.get("metadata_path"),
-                "mape": model_info.get("training_metrics", {}).get("mape"),
-                "mae": model_info.get("training_metrics", {}).get("mae"),
-                "rmse": model_info.get("training_metrics", {}).get("rmse"),
-                "r2_score": model_info.get("training_metrics", {}).get("r2"),
-                "training_samples": len(processed_data),
-                "hyperparameters": model_info.get("hyperparameters"),
-                "features_used": list(processed_data.columns),
-                "normalization_params": self.enhanced_data_processor.get_scalers(),  # Include scalers for prediction consistency
+                "mape": float(model_info.get("training_metrics", {}).get("mape", 0)) if model_info.get("training_metrics", {}).get("mape") is not None else 0,
+                "mae": float(model_info.get("training_metrics", {}).get("mae", 0)) if model_info.get("training_metrics", {}).get("mae") is not None else 0,
+                "rmse": float(model_info.get("training_metrics", {}).get("rmse", 0)) if model_info.get("training_metrics", {}).get("rmse") is not None else 0,
+                "r2_score": float(model_info.get("training_metrics", {}).get("r2", 0)) if model_info.get("training_metrics", {}).get("r2") is not None else 0,
+                "training_samples": int(len(processed_data)),
+                "hyperparameters": self._serialize_scalers(model_info.get("hyperparameters", {})),
+                "features_used": [str(f) for f in features_used] if features_used else [],
+                "normalization_params": self._serialize_scalers(self.enhanced_data_processor.get_scalers()) or {},  # Include scalers for prediction consistency
+                "product_category": model_info.get("product_category", "unknown"),  # Store product category
                 "is_active": True,
                 "is_production": True,
-                "data_quality_score": model_info.get("data_quality_score", 100.0)
+                "data_quality_score": float(model_info.get("data_quality_score", 100.0)) if model_info.get("data_quality_score") is not None else 100.0,
+                "training_start_date": training_start_date,
+                "training_end_date": training_end_date
             }
             
             model_record = await repos['model'].create_model(model_data)
@@ -533,13 +767,13 @@ class EnhancedBakeryMLTrainer:
                 "model_id": str(model_id),
                 "tenant_id": tenant_id,
                 "inventory_product_id": inventory_product_id,
-                "mae": metrics.get("mae"),
-                "mse": metrics.get("mse"),
-                "rmse": metrics.get("rmse"),
-                "mape": metrics.get("mape"),
-                "r2_score": metrics.get("r2"),
-                "accuracy_percentage": 100 - metrics.get("mape", 0) if metrics.get("mape") else None,
-                "evaluation_samples": metrics.get("data_points", 0)
+                "mae": float(metrics.get("mae")) if metrics.get("mae") is not None else None,
+                "mse": float(metrics.get("mse")) if metrics.get("mse") is not None else None,
+                "rmse": float(metrics.get("rmse")) if metrics.get("rmse") is not None else None,
+                "mape": float(metrics.get("mape")) if metrics.get("mape") is not None else None,
+                "r2_score": float(metrics.get("r2")) if metrics.get("r2") is not None else None,
+                "accuracy_percentage": float(100 - metrics.get("mape", 0)) if metrics.get("mape") is not None else None,
+                "evaluation_samples": int(metrics.get("data_points", 0)) if metrics.get("data_points") is not None else 0
             }
             
             await repos['performance'].create_performance_metric(metric_data)
@@ -672,7 +906,59 @@ class EnhancedBakeryMLTrainer:
                 sales_df['quantity'] = pd.to_numeric(sales_df['quantity'], errors='coerce')
             except Exception:
                 raise ValueError("Quantity column must be numeric")
-    
+
+    async def _categorize_all_products(
+        self,
+        sales_df: pd.DataFrame,
+        processed_data: Dict[str, pd.DataFrame]
+    ) -> Dict[str, ProductCategory]:
+        """
+        Categorize all products for category-specific forecasting.
+
+        Args:
+            sales_df: Raw sales data with product names
+            processed_data: Processed data by product ID
+
+        Returns:
+            Dict mapping inventory_product_id to ProductCategory
+        """
+        product_categories = {}
+
+        for inventory_product_id in processed_data.keys():
+            try:
+                # Get product name from sales data (if available)
+                product_sales = sales_df[sales_df['inventory_product_id'] == inventory_product_id]
+
+                # Extract product name (try multiple possible column names)
+                product_name = "unknown"
+                for name_col in ['product_name', 'name', 'item_name']:
+                    if name_col in product_sales.columns and not product_sales[name_col].empty:
+                        product_name = product_sales[name_col].iloc[0]
+                        break
+
+                # Prepare sales data for pattern analysis
+                sales_for_analysis = product_sales[['date', 'quantity']].copy() if 'date' in product_sales.columns else None
+
+                # Categorize product
+                category = self.product_categorizer.categorize_product(
+                    product_name=str(product_name),
+                    product_id=inventory_product_id,
+                    sales_data=sales_for_analysis
+                )
+
+                product_categories[inventory_product_id] = category
+
+                logger.debug("Product categorized",
+                           inventory_product_id=inventory_product_id,
+                           product_name=product_name,
+                           category=category.value)
+
+            except Exception as e:
+                logger.warning(f"Failed to categorize product {inventory_product_id}: {e}")
+                product_categories[inventory_product_id] = ProductCategory.UNKNOWN
+
+        return product_categories
+
     async def evaluate_model_performance_enhanced(self,
                                                 tenant_id: str,
                                                 inventory_product_id: str,

services/training/app/models/__init__.py

@@ -18,6 +18,7 @@ from .training import (
     ModelPerformanceMetric,
     TrainingJobQueue,
     ModelArtifact,
+    TrainingPerformanceMetrics,
 )
 
 # List all models for easier access
@@ -27,5 +28,6 @@ __all__ = [
     "ModelPerformanceMetric",
     "TrainingJobQueue",
     "ModelArtifact",
+    "TrainingPerformanceMetrics",
     "AuditLog",
 ]

services/training/app/models/training.py

@@ -150,7 +150,8 @@ class TrainedModel(Base):
     hyperparameters = Column(JSON)  # Store optimized parameters
     features_used = Column(JSON)    # List of regressor columns
     normalization_params = Column(JSON)  # Store feature normalization parameters for consistent predictions
-    
+    product_category = Column(String, nullable=True)  # Product category for category-specific forecasting
+
     # Model status
     is_active = Column(Boolean, default=True)
     is_production = Column(Boolean, default=False)
@@ -185,6 +186,7 @@ class TrainedModel(Base):
             "training_samples": self.training_samples,
             "hyperparameters": self.hyperparameters,
             "features_used": self.features_used,
+            "product_category": self.product_category,
             "is_active": self.is_active,
             "is_production": self.is_production,
             "created_at": self.created_at.isoformat() if self.created_at else None,

services/training/app/schemas/training.py

@@ -5,7 +5,7 @@ Includes all request/response schemas used by the API endpoints
 """
 
 from pydantic import BaseModel, Field, validator
-from typing import List, Optional, Dict, Any, Union
+from typing import List, Optional, Dict, Any, Union, Tuple
 from datetime import datetime
 from enum import Enum
 from uuid import UUID
@@ -37,6 +37,9 @@ class SingleProductTrainingRequest(BaseModel):
     daily_seasonality: bool = Field(True, description="Enable daily seasonality")
     weekly_seasonality: bool = Field(True, description="Enable weekly seasonality")
     yearly_seasonality: bool = Field(True, description="Enable yearly seasonality")
+    
+    # Location parameters
+    bakery_location: Optional[Tuple[float, float]] = Field(None, description="Bakery coordinates (latitude, longitude)")
 
 class DateRangeInfo(BaseModel):
     """Schema for date range information"""

services/training/app/services/training_orchestrator.py

@@ -170,6 +170,7 @@ class TrainingDataOrchestrator:
             logger.error(f"Training data preparation failed: {str(e)}")
             raise ValueError(f"Failed to prepare training data: {str(e)}")
     
+    @staticmethod
     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.
@@ -246,12 +247,14 @@ class TrainingDataOrchestrator:
                 if 'date' in record:
                     record_date = record['date']
                     
-                    # ✅ FIX: Proper timezone handling for date parsing
+                    # ✅ FIX: Proper timezone handling for date parsing - FIXED THE TRUNCATION ISSUE
                     if isinstance(record_date, str):
+                        # Parse complete ISO datetime string with timezone info intact
+                        # DO NOT truncate to date part only - this was causing the filtering issue
                         if 'T' in record_date:
                             record_date = record_date.replace('Z', '+00:00')
-                        # Parse with timezone info intact
-                        parsed_date = datetime.fromisoformat(record_date.split('T')[0])
+                        # Parse with FULL datetime info, not just date part
+                        parsed_date = datetime.fromisoformat(record_date)
                         # Ensure timezone-aware
                         if parsed_date.tzinfo is None:
                             parsed_date = parsed_date.replace(tzinfo=timezone.utc)
@@ -260,8 +263,8 @@ class TrainingDataOrchestrator:
                         # Ensure timezone-aware
                         if record_date.tzinfo is None:
                             record_date = record_date.replace(tzinfo=timezone.utc)
-                        # Normalize to start of day
-                        record_date = record_date.replace(hour=0, minute=0, second=0, microsecond=0)
+                        # DO NOT normalize to start of day - keep actual datetime for proper filtering
+                        # Only normalize if needed for daily aggregation, but preserve original for filtering
                     
                     # ✅ FIX: Ensure aligned_range dates are also timezone-aware for comparison
                     aligned_start = aligned_range.start
@@ -885,4 +888,4 @@ class TrainingDataOrchestrator:
                     1 if len(dataset.traffic_data) > 0 else 0
                 ])
             }
-        }
+        }

services/training/app/services/training_service.py

@@ -468,6 +468,7 @@ class EnhancedTrainingService:
         """
         try:
             from app.models.training import TrainingPerformanceMetrics
+            from shared.database.repository import BaseRepository
 
             # Extract timing and success data
             models_trained = training_results.get("models_trained", {})
@@ -508,10 +509,13 @@ class EnhancedTrainingService:
                 "completed_at": datetime.now(timezone.utc)
             }
 
+            # Create a temporary repository for the TrainingPerformanceMetrics model
+            # Use the session from one of the initialized repositories to ensure it's available
+            session = self.model_repo.session  # This should be the same session used by all repositories
+            metrics_repo = BaseRepository(TrainingPerformanceMetrics, session)
+            
             # Use repository to create record
-            performance_metrics = TrainingPerformanceMetrics(**metric_data)
-            self.session.add(performance_metrics)
-            await self.session.commit()
+            await metrics_repo.create(metric_data)
 
             logger.info("Saved training performance metrics for future estimations",
                        tenant_id=tenant_id,
@@ -777,17 +781,154 @@ class EnhancedTrainingService:
                        inventory_product_id=inventory_product_id,
                        job_id=job_id)
             
-            # This would use the data client to fetch data for the specific product
-            # and then use the enhanced training pipeline
-            # For now, return a success response
+            # Create initial training log
+            await self._update_job_status_repository(
+                job_id=job_id,
+                status="running",
+                progress=0,
+                current_step="Fetching training data",
+                tenant_id=tenant_id
+            )
             
+            # Prepare training data for all products to get weather/traffic data
+            # then filter down to the specific product
+            training_dataset = await self.orchestrator.prepare_training_data(
+                tenant_id=tenant_id,
+                bakery_location=bakery_location,
+                job_id=job_id + "_temp"
+            )
+            
+            # Filter sales data to the specific product
+            sales_df = pd.DataFrame(training_dataset.sales_data)
+            product_sales_df = sales_df[sales_df['inventory_product_id'] == inventory_product_id]
+            
+            if product_sales_df.empty:
+                raise ValueError(f"No sales data available for product {inventory_product_id}")
+            
+            # Prepare the data in Prophet format (ds and y columns)
+            # Ensure proper column names and types for Prophet
+            product_data = product_sales_df.copy()
+            product_data = product_data.rename(columns={
+                'sale_date': 'ds',   # Common sales date column
+                'sale_datetime': 'ds',  # Alternative date column
+                'date': 'ds',        # Alternative date column
+                'quantity': 'y',     # Quantity sold
+                'total_amount': 'y', # Alternative for sales data
+                'sales_amount': 'y', # Alternative for sales data
+                'sale_amount': 'y'   # Alternative for sales data
+            })
+            
+            # If 'ds' and 'y' columns are not renamed properly, try to infer them
+            if 'ds' not in product_data.columns:
+                # Try to find date-like columns
+                date_cols = [col for col in product_data.columns if 'date' in col.lower() or 'time' in col.lower()]
+                if date_cols:
+                    product_data = product_data.rename(columns={date_cols[0]: 'ds'})
+            
+            if 'y' not in product_data.columns:
+                # Try to find sales/quantity-like columns
+                sales_cols = [col for col in product_data.columns if 
+                            any(word in col.lower() for word in ['amount', 'quantity', 'sales', 'total', 'count', 'value'])]
+                if sales_cols:
+                    product_data = product_data.rename(columns={sales_cols[0]: 'y'})
+            
+            # Ensure required columns exist
+            if 'ds' not in product_data.columns or 'y' not in product_data.columns:
+                raise ValueError(f"Sales data must contain 'date' and 'quantity/sales' columns. Available columns: {list(product_data.columns)}")
+            
+            # Convert the date column to datetime if it's not already
+            product_data['ds'] = pd.to_datetime(product_data['ds'])
+            
+            # Convert to numeric ensuring no pandas/numpy objects remain
+            product_data['y'] = pd.to_numeric(product_data['y'], errors='coerce')
+            
+            # Sort by date to ensure proper chronological order
+            product_data = product_data.sort_values('ds').reset_index(drop=True)
+            
+            # Drop any rows with NaN values
+            product_data = product_data.dropna(subset=['ds', 'y'])
+            
+            # Ensure the data is in the right format for Prophet
+            product_data = product_data[['ds', 'y']].copy()
+
+            # Convert to pandas datetime and float types (keep as pandas Series for proper min/max operations)
+            product_data['ds'] = pd.to_datetime(product_data['ds'])
+            product_data['y'] = product_data['y'].astype(float)
+
+            # DEBUG: Log data types to diagnose dict comparison error
+            logger.info(f"DEBUG: product_data dtypes after conversion: ds={product_data['ds'].dtype}, y={product_data['y'].dtype}")
+            logger.info(f"DEBUG: product_data['ds'] sample values: {product_data['ds'].head(3).tolist()}")
+            logger.info(f"DEBUG: Attempting to get min/max...")
+            try:
+                min_val = product_data['ds'].min()
+                max_val = product_data['ds'].max()
+                logger.info(f"DEBUG: min_val type={type(min_val)}, value={min_val}")
+                logger.info(f"DEBUG: max_val type={type(max_val)}, value={max_val}")
+            except Exception as debug_e:
+                logger.error(f"DEBUG: Failed to get min/max: {debug_e}")
+                import traceback
+                logger.error(f"DEBUG: Traceback: {traceback.format_exc()}")
+
+            logger.info("Prepared training data for single product",
+                       inventory_product_id=inventory_product_id,
+                       data_points=len(product_data),
+                       date_range=f"{product_data['ds'].min()} to {product_data['ds'].max()}")
+            
+            # Update progress
+            await self._update_job_status_repository(
+                job_id=job_id,
+                status="running",
+                progress=30,
+                current_step="Training model",
+                tenant_id=tenant_id
+            )
+            
+            # Train the model using the trainer
+            # Extract datetime values with proper pandas Timestamp wrapper for type safety
+            try:
+                training_start = pd.Timestamp(product_data['ds'].min()).to_pydatetime()
+                training_end = pd.Timestamp(product_data['ds'].max()).to_pydatetime()
+            except Exception as e:
+                import traceback
+                logger.error(f"Failed to extract training dates: {e}")
+                logger.error(f"Traceback: {traceback.format_exc()}")
+                logger.error(f"product_data['ds'] dtype: {product_data['ds'].dtype}")
+                logger.error(f"product_data['ds'] first 5 values: {product_data['ds'].head().tolist()}")
+                raise
+
+            # Run the actual training
+            try:
+                model_info = await self.trainer.train_single_product_model(
+                    tenant_id=tenant_id,
+                    inventory_product_id=inventory_product_id,
+                    training_data=product_data,
+                    job_id=job_id
+                )
+            except Exception as e:
+                import traceback
+                logger.error(f"Training failed with error: {e}")
+                logger.error(f"Full traceback: {traceback.format_exc()}")
+                raise
+            
+            # Update progress
+            await self._update_job_status_repository(
+                job_id=job_id,
+                status="running",
+                progress=80,
+                current_step="Saving model",
+                tenant_id=tenant_id
+            )
+            
+            # The model should already be saved by train_single_product_model
+            # Return appropriate response
             return {
                 "job_id": job_id,
                 "tenant_id": tenant_id,
                 "inventory_product_id": inventory_product_id,
                 "status": "completed",
                 "message": "Enhanced single product training completed successfully",
-                "created_at": datetime.now(),
+                "created_at": datetime.now(timezone.utc),
+                "estimated_duration_minutes": 15,  # Default estimate for single product
                 "training_results": {
                     "total_products": 1,
                     "successful_trainings": 1,
@@ -795,21 +936,37 @@ class EnhancedTrainingService:
                     "products": [{
                         "inventory_product_id": inventory_product_id,
                         "status": "completed",
-                        "model_id": f"model_{inventory_product_id}_{job_id[:8]}",
-                        "data_points": 100,
-                        "metrics": {"mape": 15.5, "mae": 2.3, "rmse": 3.1, "r2_score": 0.85}
+                        "model_id": str(model_info.get('model_id', f"model_{inventory_product_id}_{job_id[:8]}")) if model_info.get('model_id') else None,
+                        "data_points": len(product_data) if product_data is not None else 0,
+                        # Filter metrics to ensure only numeric values are included 
+                        "metrics": {
+                            k: float(v) if not isinstance(v, (int, float)) else v
+                            for k, v in model_info.get('training_metrics', {"mape": 0.0, "mae": 0.0, "rmse": 0.0, "r2_score": 0.0}).items()
+                            if k != 'product_category' and v is not None
+                        }
                     }],
-                    "overall_training_time_seconds": 45.2
+                    "overall_training_time_seconds": model_info.get('training_time', 45.2)
                 },
                 "enhanced_features": True,
                 "repository_integration": True,
-                "completed_at": datetime.now().isoformat()
+                "completed_at": datetime.now(timezone.utc).isoformat()
             }
             
         except Exception as e:
             logger.error("Enhanced single product training failed",
                         inventory_product_id=inventory_product_id,
                         error=str(e))
+            
+            # Update status to failed
+            await self._update_job_status_repository(
+                job_id=job_id,
+                status="failed",
+                progress=0,
+                current_step="Training failed",
+                error_message=str(e),
+                tenant_id=tenant_id
+            )
+            
             raise
 
     def _create_detailed_training_response(self, final_result: Dict[str, Any]) -> Dict[str, Any]:
@@ -842,6 +999,7 @@ class EnhancedTrainingService:
                 "status": final_result["status"],
                 "message": f"Training {final_result['status']} successfully",
                 "created_at": datetime.now(),
+                "estimated_duration_minutes": final_result.get("estimated_duration_minutes", 15),
                 "training_results": {
                     "total_products": len(products),
                     "successful_trainings": len([p for p in products if p["status"] == "completed"]),

services/training/migrations/versions/20251015_1229_26a665cd5348_initial_schema_20251015_1229.py

@@ -1,226 +0,0 @@
-"""initial_schema_20251015_1229
-
-Revision ID: 26a665cd5348
-Revises: 
-Create Date: 2025-10-15 12:29:01.717552+02:00
-
-"""
-from typing import Sequence, Union
-
-from alembic import op
-import sqlalchemy as sa
-from sqlalchemy.dialects import postgresql
-
-# revision identifiers, used by Alembic.
-revision: str = '26a665cd5348'
-down_revision: Union[str, None] = None
-branch_labels: Union[str, Sequence[str], None] = None
-depends_on: Union[str, Sequence[str], None] = None
-
-
-def upgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.create_table('audit_logs',
-    sa.Column('id', sa.UUID(), nullable=False),
-    sa.Column('tenant_id', sa.UUID(), nullable=False),
-    sa.Column('user_id', sa.UUID(), nullable=False),
-    sa.Column('action', sa.String(length=100), nullable=False),
-    sa.Column('resource_type', sa.String(length=100), nullable=False),
-    sa.Column('resource_id', sa.String(length=255), nullable=True),
-    sa.Column('severity', sa.String(length=20), nullable=False),
-    sa.Column('service_name', sa.String(length=100), nullable=False),
-    sa.Column('description', sa.Text(), nullable=True),
-    sa.Column('changes', postgresql.JSON(astext_type=sa.Text()), nullable=True),
-    sa.Column('audit_metadata', postgresql.JSON(astext_type=sa.Text()), nullable=True),
-    sa.Column('ip_address', sa.String(length=45), nullable=True),
-    sa.Column('user_agent', sa.Text(), nullable=True),
-    sa.Column('endpoint', sa.String(length=255), nullable=True),
-    sa.Column('method', sa.String(length=10), nullable=True),
-    sa.Column('created_at', sa.DateTime(timezone=True), nullable=False),
-    sa.PrimaryKeyConstraint('id')
-    )
-    op.create_index('idx_audit_resource_type_action', 'audit_logs', ['resource_type', 'action'], unique=False)
-    op.create_index('idx_audit_service_created', 'audit_logs', ['service_name', 'created_at'], unique=False)
-    op.create_index('idx_audit_severity_created', 'audit_logs', ['severity', 'created_at'], unique=False)
-    op.create_index('idx_audit_tenant_created', 'audit_logs', ['tenant_id', 'created_at'], unique=False)
-    op.create_index('idx_audit_user_created', 'audit_logs', ['user_id', 'created_at'], unique=False)
-    op.create_index(op.f('ix_audit_logs_action'), 'audit_logs', ['action'], unique=False)
-    op.create_index(op.f('ix_audit_logs_created_at'), 'audit_logs', ['created_at'], unique=False)
-    op.create_index(op.f('ix_audit_logs_resource_id'), 'audit_logs', ['resource_id'], unique=False)
-    op.create_index(op.f('ix_audit_logs_resource_type'), 'audit_logs', ['resource_type'], unique=False)
-    op.create_index(op.f('ix_audit_logs_service_name'), 'audit_logs', ['service_name'], unique=False)
-    op.create_index(op.f('ix_audit_logs_severity'), 'audit_logs', ['severity'], unique=False)
-    op.create_index(op.f('ix_audit_logs_tenant_id'), 'audit_logs', ['tenant_id'], unique=False)
-    op.create_index(op.f('ix_audit_logs_user_id'), 'audit_logs', ['user_id'], unique=False)
-    op.create_table('model_artifacts',
-    sa.Column('id', sa.Integer(), nullable=False),
-    sa.Column('model_id', sa.String(length=255), nullable=False),
-    sa.Column('tenant_id', sa.UUID(), nullable=False),
-    sa.Column('artifact_type', sa.String(length=50), nullable=False),
-    sa.Column('file_path', sa.String(length=1000), nullable=False),
-    sa.Column('file_size_bytes', sa.Integer(), nullable=True),
-    sa.Column('checksum', sa.String(length=255), nullable=True),
-    sa.Column('storage_location', sa.String(length=100), nullable=False),
-    sa.Column('compression', sa.String(length=50), nullable=True),
-    sa.Column('created_at', sa.DateTime(timezone=True), nullable=True),
-    sa.Column('expires_at', sa.DateTime(timezone=True), nullable=True),
-    sa.PrimaryKeyConstraint('id')
-    )
-    op.create_index(op.f('ix_model_artifacts_id'), 'model_artifacts', ['id'], unique=False)
-    op.create_index(op.f('ix_model_artifacts_model_id'), 'model_artifacts', ['model_id'], unique=False)
-    op.create_index(op.f('ix_model_artifacts_tenant_id'), 'model_artifacts', ['tenant_id'], unique=False)
-    op.create_table('model_performance_metrics',
-    sa.Column('id', sa.Integer(), nullable=False),
-    sa.Column('model_id', sa.String(length=255), nullable=False),
-    sa.Column('tenant_id', sa.UUID(), nullable=False),
-    sa.Column('inventory_product_id', sa.UUID(), nullable=False),
-    sa.Column('mae', sa.Float(), nullable=True),
-    sa.Column('mse', sa.Float(), nullable=True),
-    sa.Column('rmse', sa.Float(), nullable=True),
-    sa.Column('mape', sa.Float(), nullable=True),
-    sa.Column('r2_score', sa.Float(), nullable=True),
-    sa.Column('accuracy_percentage', sa.Float(), nullable=True),
-    sa.Column('prediction_confidence', sa.Float(), nullable=True),
-    sa.Column('evaluation_period_start', sa.DateTime(), nullable=True),
-    sa.Column('evaluation_period_end', sa.DateTime(), nullable=True),
-    sa.Column('evaluation_samples', sa.Integer(), nullable=True),
-    sa.Column('measured_at', sa.DateTime(timezone=True), nullable=True),
-    sa.Column('created_at', sa.DateTime(timezone=True), nullable=True),
-    sa.PrimaryKeyConstraint('id')
-    )
-    op.create_index(op.f('ix_model_performance_metrics_id'), 'model_performance_metrics', ['id'], unique=False)
-    op.create_index(op.f('ix_model_performance_metrics_inventory_product_id'), 'model_performance_metrics', ['inventory_product_id'], unique=False)
-    op.create_index(op.f('ix_model_performance_metrics_model_id'), 'model_performance_metrics', ['model_id'], unique=False)
-    op.create_index(op.f('ix_model_performance_metrics_tenant_id'), 'model_performance_metrics', ['tenant_id'], unique=False)
-    op.create_table('model_training_logs',
-    sa.Column('id', sa.Integer(), nullable=False),
-    sa.Column('job_id', sa.String(length=255), nullable=False),
-    sa.Column('tenant_id', sa.UUID(), nullable=False),
-    sa.Column('status', sa.String(length=50), nullable=False),
-    sa.Column('progress', sa.Integer(), nullable=True),
-    sa.Column('current_step', sa.String(length=500), nullable=True),
-    sa.Column('start_time', sa.DateTime(timezone=True), nullable=True),
-    sa.Column('end_time', sa.DateTime(timezone=True), nullable=True),
-    sa.Column('config', sa.JSON(), nullable=True),
-    sa.Column('results', sa.JSON(), nullable=True),
-    sa.Column('error_message', sa.Text(), nullable=True),
-    sa.Column('created_at', sa.DateTime(timezone=True), nullable=True),
-    sa.Column('updated_at', sa.DateTime(timezone=True), nullable=True),
-    sa.PrimaryKeyConstraint('id')
-    )
-    op.create_index(op.f('ix_model_training_logs_id'), 'model_training_logs', ['id'], unique=False)
-    op.create_index(op.f('ix_model_training_logs_job_id'), 'model_training_logs', ['job_id'], unique=True)
-    op.create_index(op.f('ix_model_training_logs_tenant_id'), 'model_training_logs', ['tenant_id'], unique=False)
-    op.create_table('trained_models',
-    sa.Column('id', sa.UUID(), nullable=False),
-    sa.Column('tenant_id', sa.UUID(), nullable=False),
-    sa.Column('inventory_product_id', sa.UUID(), nullable=False),
-    sa.Column('model_type', sa.String(), nullable=True),
-    sa.Column('model_version', sa.String(), nullable=True),
-    sa.Column('job_id', sa.String(), nullable=False),
-    sa.Column('model_path', sa.String(), nullable=False),
-    sa.Column('metadata_path', sa.String(), nullable=True),
-    sa.Column('mape', sa.Float(), nullable=True),
-    sa.Column('mae', sa.Float(), nullable=True),
-    sa.Column('rmse', sa.Float(), nullable=True),
-    sa.Column('r2_score', sa.Float(), nullable=True),
-    sa.Column('training_samples', sa.Integer(), nullable=True),
-    sa.Column('hyperparameters', sa.JSON(), nullable=True),
-    sa.Column('features_used', sa.JSON(), nullable=True),
-    sa.Column('normalization_params', sa.JSON(), nullable=True),
-    sa.Column('is_active', sa.Boolean(), nullable=True),
-    sa.Column('is_production', sa.Boolean(), nullable=True),
-    sa.Column('created_at', sa.DateTime(timezone=True), nullable=True),
-    sa.Column('updated_at', sa.DateTime(timezone=True), nullable=True),
-    sa.Column('last_used_at', sa.DateTime(timezone=True), nullable=True),
-    sa.Column('training_start_date', sa.DateTime(timezone=True), nullable=True),
-    sa.Column('training_end_date', sa.DateTime(timezone=True), nullable=True),
-    sa.Column('data_quality_score', sa.Float(), nullable=True),
-    sa.Column('notes', sa.Text(), nullable=True),
-    sa.Column('created_by', sa.String(), nullable=True),
-    sa.PrimaryKeyConstraint('id')
-    )
-    op.create_index(op.f('ix_trained_models_inventory_product_id'), 'trained_models', ['inventory_product_id'], unique=False)
-    op.create_index(op.f('ix_trained_models_tenant_id'), 'trained_models', ['tenant_id'], unique=False)
-    op.create_table('training_job_queue',
-    sa.Column('id', sa.Integer(), nullable=False),
-    sa.Column('job_id', sa.String(length=255), nullable=False),
-    sa.Column('tenant_id', sa.UUID(), nullable=False),
-    sa.Column('job_type', sa.String(length=50), nullable=False),
-    sa.Column('priority', sa.Integer(), nullable=True),
-    sa.Column('config', sa.JSON(), nullable=True),
-    sa.Column('scheduled_at', sa.DateTime(), nullable=True),
-    sa.Column('started_at', sa.DateTime(), nullable=True),
-    sa.Column('estimated_duration_minutes', sa.Integer(), nullable=True),
-    sa.Column('status', sa.String(length=50), nullable=False),
-    sa.Column('retry_count', sa.Integer(), nullable=True),
-    sa.Column('max_retries', sa.Integer(), nullable=True),
-    sa.Column('created_at', sa.DateTime(timezone=True), nullable=True),
-    sa.Column('updated_at', sa.DateTime(timezone=True), nullable=True),
-    sa.Column('cancelled_by', sa.String(), nullable=True),
-    sa.PrimaryKeyConstraint('id')
-    )
-    op.create_index(op.f('ix_training_job_queue_id'), 'training_job_queue', ['id'], unique=False)
-    op.create_index(op.f('ix_training_job_queue_job_id'), 'training_job_queue', ['job_id'], unique=True)
-    op.create_index(op.f('ix_training_job_queue_tenant_id'), 'training_job_queue', ['tenant_id'], unique=False)
-    op.create_table('training_performance_metrics',
-    sa.Column('id', sa.UUID(), nullable=False),
-    sa.Column('tenant_id', sa.UUID(), nullable=False),
-    sa.Column('job_id', sa.String(length=255), nullable=False),
-    sa.Column('total_products', sa.Integer(), nullable=False),
-    sa.Column('successful_products', sa.Integer(), nullable=False),
-    sa.Column('failed_products', sa.Integer(), nullable=False),
-    sa.Column('total_duration_seconds', sa.Float(), nullable=False),
-    sa.Column('avg_time_per_product', sa.Float(), nullable=False),
-    sa.Column('data_analysis_time_seconds', sa.Float(), nullable=True),
-    sa.Column('training_time_seconds', sa.Float(), nullable=True),
-    sa.Column('finalization_time_seconds', sa.Float(), nullable=True),
-    sa.Column('completed_at', sa.DateTime(timezone=True), nullable=False),
-    sa.Column('created_at', sa.DateTime(timezone=True), nullable=True),
-    sa.PrimaryKeyConstraint('id')
-    )
-    op.create_index(op.f('ix_training_performance_metrics_job_id'), 'training_performance_metrics', ['job_id'], unique=False)
-    op.create_index(op.f('ix_training_performance_metrics_tenant_id'), 'training_performance_metrics', ['tenant_id'], unique=False)
-    # ### end Alembic commands ###
-
-
-def downgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.drop_index(op.f('ix_training_performance_metrics_tenant_id'), table_name='training_performance_metrics')
-    op.drop_index(op.f('ix_training_performance_metrics_job_id'), table_name='training_performance_metrics')
-    op.drop_table('training_performance_metrics')
-    op.drop_index(op.f('ix_training_job_queue_tenant_id'), table_name='training_job_queue')
-    op.drop_index(op.f('ix_training_job_queue_job_id'), table_name='training_job_queue')
-    op.drop_index(op.f('ix_training_job_queue_id'), table_name='training_job_queue')
-    op.drop_table('training_job_queue')
-    op.drop_index(op.f('ix_trained_models_tenant_id'), table_name='trained_models')
-    op.drop_index(op.f('ix_trained_models_inventory_product_id'), table_name='trained_models')
-    op.drop_table('trained_models')
-    op.drop_index(op.f('ix_model_training_logs_tenant_id'), table_name='model_training_logs')
-    op.drop_index(op.f('ix_model_training_logs_job_id'), table_name='model_training_logs')
-    op.drop_index(op.f('ix_model_training_logs_id'), table_name='model_training_logs')
-    op.drop_table('model_training_logs')
-    op.drop_index(op.f('ix_model_performance_metrics_tenant_id'), table_name='model_performance_metrics')
-    op.drop_index(op.f('ix_model_performance_metrics_model_id'), table_name='model_performance_metrics')
-    op.drop_index(op.f('ix_model_performance_metrics_inventory_product_id'), table_name='model_performance_metrics')
-    op.drop_index(op.f('ix_model_performance_metrics_id'), table_name='model_performance_metrics')
-    op.drop_table('model_performance_metrics')
-    op.drop_index(op.f('ix_model_artifacts_tenant_id'), table_name='model_artifacts')
-    op.drop_index(op.f('ix_model_artifacts_model_id'), table_name='model_artifacts')
-    op.drop_index(op.f('ix_model_artifacts_id'), table_name='model_artifacts')
-    op.drop_table('model_artifacts')
-    op.drop_index(op.f('ix_audit_logs_user_id'), table_name='audit_logs')
-    op.drop_index(op.f('ix_audit_logs_tenant_id'), table_name='audit_logs')
-    op.drop_index(op.f('ix_audit_logs_severity'), table_name='audit_logs')
-    op.drop_index(op.f('ix_audit_logs_service_name'), table_name='audit_logs')
-    op.drop_index(op.f('ix_audit_logs_resource_type'), table_name='audit_logs')
-    op.drop_index(op.f('ix_audit_logs_resource_id'), table_name='audit_logs')
-    op.drop_index(op.f('ix_audit_logs_created_at'), table_name='audit_logs')
-    op.drop_index(op.f('ix_audit_logs_action'), table_name='audit_logs')
-    op.drop_index('idx_audit_user_created', table_name='audit_logs')
-    op.drop_index('idx_audit_tenant_created', table_name='audit_logs')
-    op.drop_index('idx_audit_severity_created', table_name='audit_logs')
-    op.drop_index('idx_audit_service_created', table_name='audit_logs')
-    op.drop_index('idx_audit_resource_type_action', table_name='audit_logs')
-    op.drop_table('audit_logs')
-    # ### end Alembic commands ###

services/training/migrations/versions/26a665cd5348_initial_schema.py

@@ -0,0 +1,250 @@
+"""Initial schema with all training tables and columns
+
+Revision ID: 26a665cd5348
+Revises: 
+Create Date: 2025-10-15 12:29:01.717552+02:00
+
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.dialects import postgresql
+
+
+# revision identifiers, used by Alembic.
+revision: str = '26a665cd5348'
+down_revision: Union[str, None] = None
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    # Create audit_logs table
+    op.create_table('audit_logs',
+        sa.Column('id', sa.UUID(), nullable=False),
+        sa.Column('tenant_id', sa.UUID(), nullable=False),
+        sa.Column('user_id', sa.UUID(), nullable=False),
+        sa.Column('action', sa.String(length=100), nullable=False),
+        sa.Column('resource_type', sa.String(length=100), nullable=False),
+        sa.Column('resource_id', sa.String(length=255), nullable=True),
+        sa.Column('severity', sa.String(length=20), nullable=False),
+        sa.Column('service_name', sa.String(length=100), nullable=False),
+        sa.Column('description', sa.Text(), nullable=True),
+        sa.Column('changes', postgresql.JSON(astext_type=sa.Text()), nullable=True),
+        sa.Column('audit_metadata', postgresql.JSON(astext_type=sa.Text()), nullable=True),
+        sa.Column('ip_address', sa.String(length=45), nullable=True),
+        sa.Column('user_agent', sa.Text(), nullable=True),
+        sa.Column('endpoint', sa.String(length=255), nullable=True),
+        sa.Column('method', sa.String(length=10), nullable=True),
+        sa.Column('created_at', sa.DateTime(timezone=True), nullable=False),
+        sa.PrimaryKeyConstraint('id')
+    )
+    op.create_index('idx_audit_resource_type_action', 'audit_logs', ['resource_type', 'action'], unique=False)
+    op.create_index('idx_audit_service_created', 'audit_logs', ['service_name', 'created_at'], unique=False)
+    op.create_index('idx_audit_severity_created', 'audit_logs', ['severity', 'created_at'], unique=False)
+    op.create_index('idx_audit_tenant_created', 'audit_logs', ['tenant_id', 'created_at'], unique=False)
+    op.create_index('idx_audit_user_created', 'audit_logs', ['user_id', 'created_at'], unique=False)
+    op.create_index(op.f('ix_audit_logs_action'), 'audit_logs', ['action'], unique=False)
+    op.create_index(op.f('ix_audit_logs_created_at'), 'audit_logs', ['created_at'], unique=False)
+    op.create_index(op.f('ix_audit_logs_resource_id'), 'audit_logs', ['resource_id'], unique=False)
+    op.create_index(op.f('ix_audit_logs_resource_type'), 'audit_logs', ['resource_type'], unique=False)
+    op.create_index(op.f('ix_audit_logs_service_name'), 'audit_logs', ['service_name'], unique=False)
+    op.create_index(op.f('ix_audit_logs_severity'), 'audit_logs', ['severity'], unique=False)
+    op.create_index(op.f('ix_audit_logs_tenant_id'), 'audit_logs', ['tenant_id'], unique=False)
+    op.create_index(op.f('ix_audit_logs_user_id'), 'audit_logs', ['user_id'], unique=False)
+
+    # Create trained_models table
+    op.create_table('trained_models',
+        sa.Column('id', sa.UUID(), nullable=False),
+        sa.Column('tenant_id', sa.UUID(), nullable=False),
+        sa.Column('inventory_product_id', sa.UUID(), nullable=False),
+        sa.Column('model_type', sa.String(), nullable=True),
+        sa.Column('model_version', sa.String(), nullable=True),
+        sa.Column('job_id', sa.String(), nullable=False),
+        sa.Column('model_path', sa.String(), nullable=False),
+        sa.Column('metadata_path', sa.String(), nullable=True),
+        sa.Column('mape', sa.Float(), nullable=True),
+        sa.Column('mae', sa.Float(), nullable=True),
+        sa.Column('rmse', sa.Float(), nullable=True),
+        sa.Column('r2_score', sa.Float(), nullable=True),
+        sa.Column('training_samples', sa.Integer(), nullable=True),
+        sa.Column('hyperparameters', sa.JSON(), nullable=True),
+        sa.Column('features_used', sa.JSON(), nullable=True),
+        sa.Column('normalization_params', sa.JSON(), nullable=True),
+        sa.Column('is_active', sa.Boolean(), nullable=True),
+        sa.Column('is_production', sa.Boolean(), nullable=True),
+        sa.Column('created_at', sa.DateTime(timezone=True), nullable=True),
+        sa.Column('updated_at', sa.DateTime(timezone=True), nullable=True),
+        sa.Column('last_used_at', sa.DateTime(timezone=True), nullable=True),
+        sa.Column('training_start_date', sa.DateTime(timezone=True), nullable=True),
+        sa.Column('training_end_date', sa.DateTime(timezone=True), nullable=True),
+        sa.Column('data_quality_score', sa.Float(), nullable=True),
+        sa.Column('notes', sa.Text(), nullable=True),
+        sa.Column('created_by', sa.String(), nullable=True),
+        sa.Column('product_category', sa.String(), nullable=True),
+        sa.PrimaryKeyConstraint('id')
+    )
+    op.create_index(op.f('ix_trained_models_inventory_product_id'), 'trained_models', ['inventory_product_id'], unique=False)
+    op.create_index(op.f('ix_trained_models_tenant_id'), 'trained_models', ['tenant_id'], unique=False)
+
+    # Create model_training_logs table
+    op.create_table('model_training_logs',
+        sa.Column('id', sa.Integer(), nullable=False),
+        sa.Column('job_id', sa.String(length=255), nullable=False),
+        sa.Column('tenant_id', sa.UUID(), nullable=False),
+        sa.Column('status', sa.String(length=50), nullable=False),
+        sa.Column('progress', sa.Integer(), nullable=True),
+        sa.Column('current_step', sa.String(length=500), nullable=True),
+        sa.Column('start_time', sa.DateTime(timezone=True), nullable=True),
+        sa.Column('end_time', sa.DateTime(timezone=True), nullable=True),
+        sa.Column('config', sa.JSON(), nullable=True),
+        sa.Column('results', sa.JSON(), nullable=True),
+        sa.Column('error_message', sa.Text(), nullable=True),
+        sa.Column('created_at', sa.DateTime(timezone=True), nullable=True),
+        sa.Column('updated_at', sa.DateTime(timezone=True), nullable=True),
+        sa.PrimaryKeyConstraint('id')
+    )
+    op.create_index(op.f('ix_model_training_logs_id'), 'model_training_logs', ['id'], unique=False)
+    op.create_index(op.f('ix_model_training_logs_job_id'), 'model_training_logs', ['job_id'], unique=True)
+    op.create_index(op.f('ix_model_training_logs_tenant_id'), 'model_training_logs', ['tenant_id'], unique=False)
+
+    # Create model_performance_metrics table
+    op.create_table('model_performance_metrics',
+        sa.Column('id', sa.Integer(), nullable=False),
+        sa.Column('model_id', sa.String(length=255), nullable=False),
+        sa.Column('tenant_id', sa.UUID(), nullable=False),
+        sa.Column('inventory_product_id', sa.UUID(), nullable=False),
+        sa.Column('mae', sa.Float(), nullable=True),
+        sa.Column('mse', sa.Float(), nullable=True),
+        sa.Column('rmse', sa.Float(), nullable=True),
+        sa.Column('mape', sa.Float(), nullable=True),
+        sa.Column('r2_score', sa.Float(), nullable=True),
+        sa.Column('accuracy_percentage', sa.Float(), nullable=True),
+        sa.Column('prediction_confidence', sa.Float(), nullable=True),
+        sa.Column('evaluation_period_start', sa.DateTime(), nullable=True),
+        sa.Column('evaluation_period_end', sa.DateTime(), nullable=True),
+        sa.Column('evaluation_samples', sa.Integer(), nullable=True),
+        sa.Column('measured_at', sa.DateTime(timezone=True), nullable=True),
+        sa.Column('created_at', sa.DateTime(timezone=True), nullable=True),
+        sa.PrimaryKeyConstraint('id')
+    )
+    op.create_index(op.f('ix_model_performance_metrics_id'), 'model_performance_metrics', ['id'], unique=False)
+    op.create_index(op.f('ix_model_performance_metrics_inventory_product_id'), 'model_performance_metrics', ['inventory_product_id'], unique=False)
+    op.create_index(op.f('ix_model_performance_metrics_model_id'), 'model_performance_metrics', ['model_id'], unique=False)
+    op.create_index(op.f('ix_model_performance_metrics_tenant_id'), 'model_performance_metrics', ['tenant_id'], unique=False)
+
+    # Create training_job_queue table
+    op.create_table('training_job_queue',
+        sa.Column('id', sa.Integer(), nullable=False),
+        sa.Column('job_id', sa.String(length=255), nullable=False),
+        sa.Column('tenant_id', sa.UUID(), nullable=False),
+        sa.Column('job_type', sa.String(length=50), nullable=False),
+        sa.Column('priority', sa.Integer(), nullable=True),
+        sa.Column('config', sa.JSON(), nullable=True),
+        sa.Column('scheduled_at', sa.DateTime(), nullable=True),
+        sa.Column('started_at', sa.DateTime(), nullable=True),
+        sa.Column('estimated_duration_minutes', sa.Integer(), nullable=True),
+        sa.Column('status', sa.String(length=50), nullable=False),
+        sa.Column('retry_count', sa.Integer(), nullable=True),
+        sa.Column('max_retries', sa.Integer(), nullable=True),
+        sa.Column('created_at', sa.DateTime(timezone=True), nullable=True),
+        sa.Column('updated_at', sa.DateTime(timezone=True), nullable=True),
+        sa.Column('cancelled_by', sa.String(), nullable=True),
+        sa.PrimaryKeyConstraint('id')
+    )
+    op.create_index(op.f('ix_training_job_queue_id'), 'training_job_queue', ['id'], unique=False)
+    op.create_index(op.f('ix_training_job_queue_job_id'), 'training_job_queue', ['job_id'], unique=True)
+    op.create_index(op.f('ix_training_job_queue_tenant_id'), 'training_job_queue', ['tenant_id'], unique=False)
+
+    # Create model_artifacts table
+    op.create_table('model_artifacts',
+        sa.Column('id', sa.Integer(), nullable=False),
+        sa.Column('model_id', sa.String(length=255), nullable=False),
+        sa.Column('tenant_id', sa.UUID(), nullable=False),
+        sa.Column('artifact_type', sa.String(length=50), nullable=False),
+        sa.Column('file_path', sa.String(length=1000), nullable=False),
+        sa.Column('file_size_bytes', sa.Integer(), nullable=True),
+        sa.Column('checksum', sa.String(length=255), nullable=True),
+        sa.Column('storage_location', sa.String(length=100), nullable=False),
+        sa.Column('compression', sa.String(length=50), nullable=True),
+        sa.Column('created_at', sa.DateTime(timezone=True), nullable=True),
+        sa.Column('expires_at', sa.DateTime(timezone=True), nullable=True),
+        sa.PrimaryKeyConstraint('id')
+    )
+    op.create_index(op.f('ix_model_artifacts_id'), 'model_artifacts', ['id'], unique=False)
+    op.create_index(op.f('ix_model_artifacts_model_id'), 'model_artifacts', ['model_id'], unique=False)
+    op.create_index(op.f('ix_model_artifacts_tenant_id'), 'model_artifacts', ['tenant_id'], unique=False)
+
+    # Create training_performance_metrics table
+    op.create_table('training_performance_metrics',
+        sa.Column('id', sa.UUID(), nullable=False),
+        sa.Column('tenant_id', sa.UUID(), nullable=False),
+        sa.Column('job_id', sa.String(length=255), nullable=False),
+        sa.Column('total_products', sa.Integer(), nullable=False),
+        sa.Column('successful_products', sa.Integer(), nullable=False),
+        sa.Column('failed_products', sa.Integer(), nullable=False),
+        sa.Column('total_duration_seconds', sa.Float(), nullable=False),
+        sa.Column('avg_time_per_product', sa.Float(), nullable=False),
+        sa.Column('data_analysis_time_seconds', sa.Float(), nullable=True),
+        sa.Column('training_time_seconds', sa.Float(), nullable=True),
+        sa.Column('finalization_time_seconds', sa.Float(), nullable=True),
+        sa.Column('completed_at', sa.DateTime(timezone=True), nullable=False),
+        sa.Column('created_at', sa.DateTime(timezone=True), nullable=True),
+        sa.PrimaryKeyConstraint('id')
+    )
+    op.create_index(op.f('ix_training_performance_metrics_job_id'), 'training_performance_metrics', ['job_id'], unique=False)
+    op.create_index(op.f('ix_training_performance_metrics_tenant_id'), 'training_performance_metrics', ['tenant_id'], unique=False)
+
+
+def downgrade() -> None:
+    # Drop training_performance_metrics table
+    op.drop_index(op.f('ix_training_performance_metrics_tenant_id'), table_name='training_performance_metrics')
+    op.drop_index(op.f('ix_training_performance_metrics_job_id'), table_name='training_performance_metrics')
+    op.drop_table('training_performance_metrics')
+
+    # Drop model_artifacts table
+    op.drop_index(op.f('ix_model_artifacts_tenant_id'), table_name='model_artifacts')
+    op.drop_index(op.f('ix_model_artifacts_model_id'), table_name='model_artifacts')
+    op.drop_index(op.f('ix_model_artifacts_id'), table_name='model_artifacts')
+    op.drop_table('model_artifacts')
+
+    # Drop training_job_queue table
+    op.drop_index(op.f('ix_training_job_queue_tenant_id'), table_name='training_job_queue')
+    op.drop_index(op.f('ix_training_job_queue_job_id'), table_name='training_job_queue')
+    op.drop_index(op.f('ix_training_job_queue_id'), table_name='training_job_queue')
+    op.drop_table('training_job_queue')
+
+    # Drop model_performance_metrics table
+    op.drop_index(op.f('ix_model_performance_metrics_tenant_id'), table_name='model_performance_metrics')
+    op.drop_index(op.f('ix_model_performance_metrics_model_id'), table_name='model_performance_metrics')
+    op.drop_index(op.f('ix_model_performance_metrics_inventory_product_id'), table_name='model_performance_metrics')
+    op.drop_index(op.f('ix_model_performance_metrics_id'), table_name='model_performance_metrics')
+    op.drop_table('model_performance_metrics')
+
+    # Drop model_training_logs table
+    op.drop_index(op.f('ix_model_training_logs_tenant_id'), table_name='model_training_logs')
+    op.drop_index(op.f('ix_model_training_logs_job_id'), table_name='model_training_logs')
+    op.drop_index(op.f('ix_model_training_logs_id'), table_name='model_training_logs')
+    op.drop_table('model_training_logs')
+
+    # Drop trained_models table (with the product_category column)
+    op.drop_index(op.f('ix_trained_models_tenant_id'), table_name='trained_models')
+    op.drop_index(op.f('ix_trained_models_inventory_product_id'), table_name='trained_models')
+    op.drop_table('trained_models')
+
+    # Drop audit_logs table
+    op.drop_index(op.f('ix_audit_logs_user_id'), table_name='audit_logs')
+    op.drop_index(op.f('ix_audit_logs_tenant_id'), table_name='audit_logs')
+    op.drop_index(op.f('ix_audit_logs_severity'), table_name='audit_logs')
+    op.drop_index(op.f('ix_audit_logs_service_name'), table_name='audit_logs')
+    op.drop_index(op.f('ix_audit_logs_resource_type'), table_name='audit_logs')
+    op.drop_index(op.f('ix_audit_logs_resource_id'), table_name='audit_logs')
+    op.drop_index(op.f('ix_audit_logs_created_at'), table_name='audit_logs')
+    op.drop_index(op.f('ix_audit_logs_action'), table_name='audit_logs')
+    op.drop_index('idx_audit_user_created', table_name='audit_logs')
+    op.drop_index('idx_audit_tenant_created', table_name='audit_logs')
+    op.drop_index('idx_audit_severity_created', table_name='audit_logs')
+    op.drop_index('idx_audit_service_created', table_name='audit_logs')
+    op.drop_index('idx_audit_resource_type_action', table_name='audit_logs')
+    op.drop_table('audit_logs')

services/training/requirements.txt

@@ -12,10 +12,12 @@ psycopg2-binary==2.9.10
 
 # ML libraries
 prophet==1.2.1
+cmdstanpy==1.2.4
 scikit-learn==1.6.1
 pandas==2.2.3
 numpy==2.2.2
 joblib==1.4.2
+xgboost==2.1.3
 
 # HTTP client
 httpx==0.28.1
@@ -48,6 +50,7 @@ psutil==6.1.1
 # Utilities
 python-dateutil==2.9.0.post0
 pytz==2024.2
+holidays==0.63
 
 # Hyperparameter optimization
 optuna==4.2.0