Add improvements 2

2026-01-12 22:15:11 +01:00
parent 230bbe6a19
commit b931a5c45e
40 changed files with 1820 additions and 887 deletions
--- a/DOCKER_MAINTENANCE.md
+++ b/DOCKER_MAINTENANCE.md
@@ -0,0 +1,120 @@
+# Docker Maintenance Guide for Local Development
+
+## The Problem
+
+When developing with Tilt and local Kubernetes (Kind), Docker accumulates:
+- **Multiple image versions** from each code change (Tilt rebuilds)
+- **Unused volumes** from previous cluster runs
+- **Build cache** that grows over time
+
+This quickly fills up disk space, causing pods to fail with "No space left on device" errors.
+
+## Quick Fix (When You Hit Disk Issues)
+
+```bash
+# Clean up all unused Docker resources
+docker system prune -a --volumes -f
+```
+
+This removes:
+- All unused images
+- All unused volumes
+- All build cache
+
+**Expected recovery**: 60-100GB
+
+## Regular Maintenance
+
+### Option 1: Use the Cleanup Script (Recommended)
+
+Run the maintenance script weekly:
+
+```bash
+./scripts/cleanup-docker.sh
+```
+
+Or run it automatically without confirmation:
+
+```bash
+./scripts/cleanup-docker.sh --auto
+```
+
+### Option 2: Manual Commands
+
+```bash
+# Remove images older than 24 hours
+docker image prune -af --filter "until=24h"
+
+# Remove unused volumes
+docker volume prune -f
+
+# Remove build cache
+docker builder prune -af
+```
+
+### Option 3: Set Up Automated Cleanup
+
+Add to your crontab (run every Sunday at 2 AM):
+
+```bash
+crontab -e
+# Add this line:
+0 2 * * 0 /Users/urtzialfaro/Documents/bakery-ia/scripts/cleanup-docker.sh --auto >> /tmp/docker-cleanup.log 2>&1
+```
+
+## Monitoring Disk Usage
+
+### Check Docker disk usage:
+```bash
+docker system df
+```
+
+### Check Kind node disk usage:
+```bash
+docker exec bakery-ia-local-control-plane df -h /var
+```
+
+### Alert thresholds:
+- **< 70%**: Healthy ✅
+- **70-85%**: Consider cleanup soon ⚠️
+- **> 85%**: Run cleanup immediately 🚨
+- **> 95%**: Critical - pods will fail ❌
+
+## Prevention Tips
+
+1. **Run cleanup weekly** to prevent accumulation
+2. **Monitor disk usage** before long dev sessions
+3. **Delete old Kind clusters** when switching projects:
+   ```bash
+   kind delete cluster --name bakery-ia-local
+   ```
+4. **Increase Docker disk allocation** in Docker Desktop settings if you frequently rebuild many services
+
+## Troubleshooting
+
+### Pods in CrashLoopBackOff after disk issues:
+
+1. Run cleanup (see Quick Fix above)
+2. Restart failed pods:
+   ```bash
+   kubectl get pods -n bakery-ia | grep -E "(CrashLoopBackOff|Error)" | awk '{print $1}' | xargs kubectl delete pod -n bakery-ia
+   ```
+
+### Cleanup didn't free enough space:
+
+If still above 90% after cleanup:
+
+```bash
+# Nuclear option - rebuild everything
+kind delete cluster --name bakery-ia-local
+docker system prune -a --volumes -f
+# Then recreate cluster with your setup scripts
+```
+
+## What Happened Today (2026-01-12)
+
+- **Issue**: Disk was 100% full (113GB/113GB), causing database pods to crash
+- **Root cause**: 122 unused Docker images + 16 unused volumes + 6GB build cache
+- **Solution**: Ran `docker system prune -a --volumes -f`
+- **Result**: Freed 89GB, disk now at 22% usage (24GB/113GB)
+- **All services recovered successfully**
--- a/Lista_de_Mejoras_Propuestas.md
+++ b/Lista_de_Mejoras_Propuestas.md
@@ -1,185 +0,0 @@
-# Lista de Mejoras Propuestas
-
-## 1. Nueva Sección: Seguridad y Ciberseguridad (Añadir después de 5.2)
-
-### 5.3. Arquitectura de Seguridad y Cumplimiento Normativo Europeo
-
-**Autenticación y Autorización:**
- JWT con rotación de tokens cada 15 minutos
- Control de acceso basado en roles (RBAC)
- Rate limiting (300 req/min por cliente)
- Autenticación multifactor (MFA) planificada para Q1 2026
-
-**Protección de Datos:**
- Cifrado AES-256 en reposo
- HTTPS obligatorio con certificados Let's Encrypt auto-renovables
- Aislamiento multi-tenant a nivel de base de datos
- Prevención SQL injection mediante Pydantic schemas
- Protección XSS y CORS
-
-**Monitorización y Trazabilidad:**
- OpenTelemetry para trazabilidad distribuida end-to-end
- SigNoz como plataforma unificada de observabilidad
- Logs centralizados con correlación de trazas
- Auditoría completa de accesos y cambios
- Alertas en tiempo real (email y Slack)
-
-**Cumplimiento RGPD:**
- Privacy by design
- Derecho al olvido implementado
- Exportación de datos en CSV/JSON
- Gestión de consentimientos con historial
- Anonimización de datos analíticos
-
-**Infraestructura Segura:**
- Kubernetes con políticas de seguridad de pods
- Actualizaciones automáticas de seguridad
- Backups cifrados diarios
- Plan de recuperación ante desastres (DR)
- PostgreSQL 17 con configuraciones hardened
-
-## 2. Actualizar Sección 4.4 (Competencia) - Añadir Ventaja de Seguridad
-
-**Ventaja Competitiva en Seguridad:**
- Arquitectura "Security-First" vs. soluciones legacy vulnerables
- Cumplimiento RGPD desde el diseño (competidores retrofitting)
- Observabilidad completa (OpenTelemetry + SigNoz) vs. cajas negras
- Certificaciones de seguridad planificadas (ISO 27001, ENS)
- Alineación con NIS2 Directive (obligatoria 2024 para cadena alimentaria)
-
-## 3. Actualizar Sección 8.1 (Financiación) - Nuevas Líneas de Ayuda
-
-### Añadir subsección: "Financiación Europea en Ciberseguridad 2026-2027"
-
-**Programas Europeos Identificados (2026-2027):**
-
-1. **Digital Europe Programme - Ciberseguridad (€390M totales)**
-   - **UPTAKE Program**: €15M para SMEs en cumplimiento normativo
-   - **AI-Based Cybersecurity**: €15M para sistemas IA de seguridad
-   - Cofinanciación: hasta 75% de costes de proyecto
-   - Proyectos: €3-5M por proyecto, duración 36 meses
-   - **Elegibilidad**: Bakery-IA califica como SME tecnológica
-   - **Solicitud estimada**: €200.000 (Q1 2026)
-
-2. **INCIBE EMPRENDE Program (España Digital 2026)**
-   - Presupuesto: €191M (2023-2026)
-   - 34 entidades colaboradoras en España
-   - Ideación, incubación y aceleración en ciberseguridad
-   - Fondos Next Generation-EU
-   - **Solicitud estimada**: €50.000 (Q2 2026)
-
-3. **ENISA Emprendedoras Digitales**
-   - Préstamos participativos hasta €51M movilizables
-   - Líneas específicas para emprendimiento digital
-   - Sin avales personales, condiciones flexibles
-   - **Solicitud estimada**: €75.000 (Q2 2026)
-
-**Alineación con Prioridades UE:**
- NIS2 Directive (seguridad sector alimentario)
- Cyber Resilience Act (productos digitales seguros)
- AI Act (IA transparente y auditable)
- GDPR (protección datos desde diseño)
-
-## 4. Actualizar Sección 5.2 (Arquitectura Técnica)
-
-### Añadir detalles de la documentación técnica:
-
-**Arquitectura de Microservicios (21 servicios independientes):**
- API Gateway centralizado con JWT y cache Redis (95% hit rate)
- Frontend: React 18 + TypeScript (PWA mobile-first)
- 18 bases de datos PostgreSQL 17 (patrón database-per-service)
- Redis 7.4 para caché y RabbitMQ 4.1 para eventos
- Kubernetes en VPS (escalabilidad horizontal)
-
-**Innovación Técnica Destacable:**
- **Sistema de Alertas Enriquecidas** (3 niveles: Alertas/Notificaciones/Recomendaciones)
- **Priorización Inteligente** con scoring 0-100 (4 factores ponderados)
- **Escalado Temporal** (+10 a 48h, +20 a 72h, +30 cerca deadline)
- **Encadenamiento Causal** (stock shortage → retraso producción → riesgo pedido)
- **Deduplicación** (95% reducción spam de alertas)
- **SSE + WebSocket** para actualizaciones en tiempo real
- **Prophet ML** con 20+ features (AEMET, tráfico Madrid, festivos)
-
-**Observabilidad de Clase Empresarial:**
- **SigNoz**: Trazas, métricas y logs unificados
- **OpenTelemetry**: Auto-instrumentación de 18 servicios
- **ClickHouse**: Backend de alto rendimiento para análisis
- **Alerting**: Multi-canal (email, Slack) vía AlertManager
- Monitorización: 18 DBs PostgreSQL, Redis, RabbitMQ, Kubernetes
-
-## 5. Actualizar Resumen Ejecutivo (Sección 0)
-
-### Añadir bullet:
-
- **Seguridad y Cumplimiento:** Arquitectura Security-First con cumplimiento RGPD, observabilidad completa (OpenTelemetry/SigNoz), y alineación con normativas europeas (NIS2, Cyber Resilience Act). Elegible para €390M del Digital Europe Programme en ciberseguridad.
-
-## 6. Actualizar Sección 9 (Decálogo) - Añadir Oportunidades de Financiación
-
-**6. Alineación Estratégica con Prioridades Europeas 2026-2027:**
-   - **€390M disponibles** en Digital Europe Programme para ciberseguridad
-   - **€191M** del programa INCIBE EMPRENDE (España Digital 2026)
-   - Bakery-IA califica para **3 líneas de financiación simultáneas**:
-     * UPTAKE (€200K) - Cumplimiento normativo SMEs
-     * INCIBE EMPRENDE (€50K) - Aceleración cybersecurity startups
-     * ENISA Digital (€75K) - Préstamo participativo
-   - **Total potencial**: €325.000 en financiación no dilutiva adicional
-   - Ventaja competitiva: Security-First vs. competidores legacy
-
-## 7. Nueva Tabla de Costes Operativos - Añadir Línea de Seguridad
-
-| **Seguridad y Cumplimiento** | | |
-| Certificado SSL (Let's Encrypt) | Gratuito | €0 | Renovación automática |
-| SigNoz Observability (self-hosted) | Incluido en VPS | €0 | Vs. €500+/mes en SaaS |
-| Auditoría RGPD anual | Externa | €1,200/año | Compliance obligatorio |
-| Backups cifrados (off-site) | Backblaze B2 | €5/mes | €60/año |
-
-## 8. Actualizar Roadmap (Sección 9) - Añadir Hitos de Seguridad
-
-**Q1 2026:**
- ✅ Implementación MFA (Multi-Factor Authentication)
- ✅ Solicitud Digital Europe Programme (UPTAKE)
- ✅ Auditoría RGPD externa
-
-**Q2 2026:**
- 📋 Certificación ISO 27001 (inicio proceso)
- 📋 Implementación NIS2 compliance
- 📋 Solicitud INCIBE EMPRENDE
-
-**Q3 2026:**
- 📋 Penetration testing externo
- 📋 Certificación ENS (Esquema Nacional de Seguridad)
-
-## 9. Actualizar Petición Concreta a VUE (Sección 9.2)
-
-### Añadir nuevo punto 5:
-
-**5. Conexión con Programas Europeos de Ciberseguridad:**
-   - Orientación para solicitud Digital Europe Programme (UPTAKE: €200K)
-   - Introducción a INCIBE EMPRENDE y red de 34 entidades colaboradoras
-   - Asesoramiento en preparación de propuestas técnicas para fondos EU
-   - Contacto con CDTI para programa NEOTEC (R&D+Ciberseguridad)
-
-## 10. Añadir Anexo 7 - Compliance y Certificaciones
-
-## ANEXO 7: ROADMAP DE COMPLIANCE Y CERTIFICACIONES
-
-**Normativas Aplicables:**
- ✅ RGPD (Reglamento General de Protección de Datos) - Implementado
- 📋 NIS2 Directive (Seguridad de redes y sistemas de información) - Q2 2026
- 📋 Cyber Resilience Act (Productos digitales seguros) - Q3 2026
- 📋 AI Act (Transparencia y auditoría de IA) - Q4 2026
-
-**Certificaciones Planificadas:**
- 📋 ISO 27001 (Gestión de Seguridad de la Información) - 12-18 meses
- 📋 ENS Medio (Esquema Nacional de Seguridad) - 6-9 meses
- 📋 SOC 2 Type II (para clientes Enterprise) - 18-24 meses
-
-**Inversión Estimada en Compliance (3 años):** €25,000
-**ROI Esperado:** Acceso a clientes Enterprise (+€150K ARR potencial)
-
-## Resumen de Cambios Cuantitativos:
- Nueva financiación identificada: €325.000 (vs. €18.000 original)
- Nuevos programas: 3 líneas europeas de ciberseguridad
- Secciones nuevas: 2 (Seguridad 5.3, Compliance Anexo 7)
- Actualizaciones: 8 secciones existentes mejoradas
- Ventaja competitiva: Security-First enfatizada en 4 lugares
--- a/STRIPE_TESTING_GUIDE.md
+++ b/STRIPE_TESTING_GUIDE.md
--- a/38
+++ b/38
@@ -7,6 +7,13 @@
 #   - PostgreSQL pgcrypto extension and audit logging
 #   - Organized resource dependencies and live-reload capabilities
 #   - Local registry for faster image builds and deployments
+#
+# Build Optimization:
+#   - Services only rebuild when their specific code changes (not all services)
+#   - Shared folder changes trigger rebuild of ALL services (as they all depend on it)
+#   - Uses 'only' parameter to watch only relevant files per service
+#   - Frontend only rebuilds when frontend/ code changes
+#   - Gateway only rebuilds when gateway/ or shared/ code changes
 # =============================================================================

 # =============================================================================
@@ -197,16 +204,25 @@ k8s_yaml(kustomize('infrastructure/kubernetes/overlays/dev'))
 # =============================================================================

 # Helper function for Python services with live updates
+# This function ensures services only rebuild when their specific code changes,
+# but all services rebuild when shared/ folder changes
 def build_python_service(service_name, service_path):
    docker_build(
        'bakery/' + service_name,
        context='.',
        dockerfile='./services/' + service_path + '/Dockerfile',
+        # Only watch files relevant to this specific service + shared code
+        only=[
+            './services/' + service_path,
+            './shared',
+            './scripts',
+        ],
        live_update=[
            # Fall back to full image build if Dockerfile or requirements change
            fall_back_on([
                './services/' + service_path + '/Dockerfile',
-                './services/' + service_path + '/requirements.txt'
+                './services/' + service_path + '/requirements.txt',
+                './shared/requirements-tracing.txt',
            ]),

            # Sync service code
@@ -290,10 +306,21 @@ docker_build(
    'bakery/gateway',
    context='.',
    dockerfile='./gateway/Dockerfile',
+    # Only watch gateway-specific files and shared code
+    only=[
+        './gateway',
+        './shared',
+        './scripts',
+    ],
    live_update=[
-        fall_back_on(['./gateway/Dockerfile', './gateway/requirements.txt']),
+        fall_back_on([
+            './gateway/Dockerfile',
+            './gateway/requirements.txt',
+            './shared/requirements-tracing.txt',
+        ]),
        sync('./gateway', '/app'),
        sync('./shared', '/app/shared'),
+        sync('./scripts', '/app/scripts'),
        run('kill -HUP 1', trigger=['./gateway/**/*.py', './shared/**/*.py']),
    ],
    ignore=[
@@ -680,6 +707,13 @@ Documentation:
  docs/SECURITY_IMPLEMENTATION_COMPLETE.md
  docs/DATABASE_SECURITY_ANALYSIS_REPORT.md

+Build Optimization Active:
+  ✅ Services only rebuild when their code changes
+  ✅ Shared folder changes trigger ALL services (as expected)
+  ✅ Reduces unnecessary rebuilds and disk usage
+  💡 Edit service code: only that service rebuilds
+  💡 Edit shared/ code: all services rebuild (required)
+
 Useful Commands:
  # Work on specific services only
  tilt up <service-name> <service-name>
--- a/frontend/src/components/domain/onboarding/steps/RegisterTenantStep.tsx
+++ b/frontend/src/components/domain/onboarding/steps/RegisterTenantStep.tsx
@@ -198,16 +198,27 @@ export const RegisterTenantStep: React.FC<RegisterTenantStepProps> = ({

      // Trigger POI detection in the background (non-blocking)
      // This replaces the removed POI Detection step
+      // POI detection will be cached for 90 days and reused during training
      const bakeryLocation = wizardContext.state.bakeryLocation;
      if (bakeryLocation?.latitude && bakeryLocation?.longitude && tenant.id) {
+        console.log(`🔍 Triggering background POI detection for tenant ${tenant.id}...`);
+
        // Run POI detection asynchronously without blocking the wizard flow
+        // This ensures POI data is ready before the training step
        poiContextApi.detectPOIs(
          tenant.id,
          bakeryLocation.latitude,
          bakeryLocation.longitude,
-          false // use_cache = false for initial detection
+          false // force_refresh = false, will use cache if available
        ).then((result) => {
-          console.log(`✅ POI detection completed automatically for tenant ${tenant.id}:`, result.summary);
+          const source = result.source || 'unknown';
+          console.log(`✅ POI detection completed for tenant ${tenant.id} (source: ${source})`);
+
+          if (result.poi_context) {
+            const totalPois = result.poi_context.total_pois_detected || 0;
+            const relevantCategories = result.poi_context.relevant_categories?.length || 0;
+            console.log(`📍 POI Summary: ${totalPois} POIs detected, ${relevantCategories} relevant categories`);
+          }

          // Phase 3: Handle calendar suggestion if available
          if (result.calendar_suggestion) {
@@ -229,9 +240,12 @@ export const RegisterTenantStep: React.FC<RegisterTenantStepProps> = ({
            }
          }
        }).catch((error) => {
-          console.warn('⚠️  Background POI detection failed (non-blocking):', error);
-          // This is non-critical, so we don't block the user
+          console.warn('⚠️ Background POI detection failed (non-blocking):', error);
+          console.warn('Training will continue without POI features if detection is not complete.');
+          // This is non-critical - training service will continue without POI features
        });
+      } else {
+        console.warn('⚠️ Cannot trigger POI detection: missing location data or tenant ID');
      }

      // Update the wizard context with tenant info
--- a/gateway/README.md
+++ b/gateway/README.md
@@ -352,6 +352,25 @@ headers = {
 - **Caching**: Gateway caches validated service tokens for 5 minutes
 - **No Additional HTTP Calls**: Service auth happens locally at gateway

+### Unified Header Management System
+
+The gateway uses a **centralized HeaderManager** for consistent header handling across all middleware and proxy layers.
+
+**Key Features:**
+- Standardized header names and conventions
+- Automatic header sanitization to prevent spoofing
+- Unified header injection and forwarding
+- Cross-middleware header access via `request.state.injected_headers`
+- Consistent logging and error handling
+
+**Standard Headers:**
+- `x-user-id`, `x-user-email`, `x-user-role`, `x-user-type`
+- `x-service-name`, `x-tenant-id`
+- `x-subscription-tier`, `x-subscription-status`
+- `x-is-demo`, `x-demo-session-id`, `x-demo-account-type`
+- `x-tenant-access-type`, `x-can-view-children`, `x-parent-tenant-id`
+- `x-forwarded-by`, `x-request-id`
+
 ### Context Header Injection

 When a service token is validated, the gateway injects these headers for downstream services:
--- a/gateway/app/core/header_manager.py
+++ b/gateway/app/core/header_manager.py
@@ -0,0 +1,345 @@
+"""
+Unified Header Management System for API Gateway
+Centralized header injection, forwarding, and validation
+"""
+
+import structlog
+from fastapi import Request
+from typing import Dict, Any, Optional, List
+
+logger = structlog.get_logger()
+
+
+class HeaderManager:
+    """
+    Centralized header management for consistent header handling across gateway
+    """
+    
+    # Standard header names (lowercase for consistency)
+    STANDARD_HEADERS = {
+        'user_id': 'x-user-id',
+        'user_email': 'x-user-email', 
+        'user_role': 'x-user-role',
+        'user_type': 'x-user-type',
+        'service_name': 'x-service-name',
+        'tenant_id': 'x-tenant-id',
+        'subscription_tier': 'x-subscription-tier',
+        'subscription_status': 'x-subscription-status',
+        'is_demo': 'x-is-demo',
+        'demo_session_id': 'x-demo-session-id',
+        'demo_account_type': 'x-demo-account-type',
+        'tenant_access_type': 'x-tenant-access-type',
+        'can_view_children': 'x-can-view-children',
+        'parent_tenant_id': 'x-parent-tenant-id',
+        'forwarded_by': 'x-forwarded-by',
+        'request_id': 'x-request-id'
+    }
+    
+    # Headers that should be sanitized/removed from incoming requests
+    SANITIZED_HEADERS = [
+        'x-subscription-',
+        'x-user-', 
+        'x-tenant-',
+        'x-demo-',
+        'x-forwarded-by'
+    ]
+    
+    # Headers that should be forwarded to downstream services
+    FORWARDABLE_HEADERS = [
+        'authorization',
+        'content-type',
+        'accept',
+        'accept-language',
+        'user-agent',
+        'x-internal-service'  # Required for internal service-to-service ML/alert triggers
+    ]
+    
+    def __init__(self):
+        self._initialized = False
+    
+    def initialize(self):
+        """Initialize header manager"""
+        if not self._initialized:
+            logger.info("HeaderManager initialized")
+            self._initialized = True
+    
+    def sanitize_incoming_headers(self, request: Request) -> None:
+        """
+        Remove sensitive headers from incoming request to prevent spoofing
+        """
+        if not hasattr(request.headers, '_list'):
+            return
+            
+        # Filter out headers that start with sanitized prefixes
+        sanitized_headers = [
+            (k, v) for k, v in request.headers.raw
+            if not any(k.decode().lower().startswith(prefix.lower()) 
+                      for prefix in self.SANITIZED_HEADERS)
+        ]
+        
+        request.headers.__dict__["_list"] = sanitized_headers
+        logger.debug("Sanitized incoming headers")
+    
+    def inject_context_headers(self, request: Request, user_context: Dict[str, Any], 
+                             tenant_id: Optional[str] = None) -> Dict[str, str]:
+        """
+        Inject standardized context headers into request
+        Returns dict of injected headers for reference
+        """
+        injected_headers = {}
+        
+        # Ensure headers list exists
+        if not hasattr(request.headers, '_list'):
+            request.headers.__dict__["_list"] = []
+        
+        # Store headers in request.state for cross-middleware access
+        request.state.injected_headers = {}
+        
+        # User context headers
+        if user_context.get('user_id'):
+            header_name = self.STANDARD_HEADERS['user_id']
+            header_value = str(user_context['user_id'])
+            self._add_header(request, header_name, header_value)
+            injected_headers[header_name] = header_value
+            request.state.injected_headers[header_name] = header_value
+            
+        if user_context.get('email'):
+            header_name = self.STANDARD_HEADERS['user_email']
+            header_value = str(user_context['email'])
+            self._add_header(request, header_name, header_value)
+            injected_headers[header_name] = header_value
+            request.state.injected_headers[header_name] = header_value
+            
+        if user_context.get('role'):
+            header_name = self.STANDARD_HEADERS['user_role']
+            header_value = str(user_context['role'])
+            self._add_header(request, header_name, header_value)
+            injected_headers[header_name] = header_value
+            request.state.injected_headers[header_name] = header_value
+            
+        # User type (service vs regular user)
+        if user_context.get('type'):
+            header_name = self.STANDARD_HEADERS['user_type']
+            header_value = str(user_context['type'])
+            self._add_header(request, header_name, header_value)
+            injected_headers[header_name] = header_value
+            request.state.injected_headers[header_name] = header_value
+            
+        # Service name for service tokens
+        if user_context.get('service'):
+            header_name = self.STANDARD_HEADERS['service_name']
+            header_value = str(user_context['service'])
+            self._add_header(request, header_name, header_value)
+            injected_headers[header_name] = header_value
+            request.state.injected_headers[header_name] = header_value
+            
+        # Tenant context
+        if tenant_id:
+            header_name = self.STANDARD_HEADERS['tenant_id']
+            header_value = str(tenant_id)
+            self._add_header(request, header_name, header_value)
+            injected_headers[header_name] = header_value
+            request.state.injected_headers[header_name] = header_value
+            
+        # Subscription context
+        if user_context.get('subscription_tier'):
+            header_name = self.STANDARD_HEADERS['subscription_tier']
+            header_value = str(user_context['subscription_tier'])
+            self._add_header(request, header_name, header_value)
+            injected_headers[header_name] = header_value
+            request.state.injected_headers[header_name] = header_value
+            
+        if user_context.get('subscription_status'):
+            header_name = self.STANDARD_HEADERS['subscription_status']
+            header_value = str(user_context['subscription_status'])
+            self._add_header(request, header_name, header_value)
+            injected_headers[header_name] = header_value
+            request.state.injected_headers[header_name] = header_value
+            
+        # Demo session context
+        is_demo = user_context.get('is_demo', False)
+        if is_demo:
+            header_name = self.STANDARD_HEADERS['is_demo']
+            header_value = "true"
+            self._add_header(request, header_name, header_value)
+            injected_headers[header_name] = header_value
+            request.state.injected_headers[header_name] = header_value
+            
+            if user_context.get('demo_session_id'):
+                header_name = self.STANDARD_HEADERS['demo_session_id']
+                header_value = str(user_context['demo_session_id'])
+                self._add_header(request, header_name, header_value)
+                injected_headers[header_name] = header_value
+                request.state.injected_headers[header_name] = header_value
+                
+            if user_context.get('demo_account_type'):
+                header_name = self.STANDARD_HEADERS['demo_account_type']
+                header_value = str(user_context['demo_account_type'])
+                self._add_header(request, header_name, header_value)
+                injected_headers[header_name] = header_value
+                request.state.injected_headers[header_name] = header_value
+        
+        # Hierarchical access context
+        if tenant_id:
+            tenant_access_type = getattr(request.state, 'tenant_access_type', 'direct')
+            can_view_children = getattr(request.state, 'can_view_children', False)
+            
+            header_name = self.STANDARD_HEADERS['tenant_access_type']
+            header_value = str(tenant_access_type)
+            self._add_header(request, header_name, header_value)
+            injected_headers[header_name] = header_value
+            request.state.injected_headers[header_name] = header_value
+            
+            header_name = self.STANDARD_HEADERS['can_view_children']
+            header_value = str(can_view_children).lower()
+            self._add_header(request, header_name, header_value)
+            injected_headers[header_name] = header_value
+            request.state.injected_headers[header_name] = header_value
+            
+            # Parent tenant ID if hierarchical access
+            parent_tenant_id = getattr(request.state, 'parent_tenant_id', None)
+            if parent_tenant_id:
+                header_name = self.STANDARD_HEADERS['parent_tenant_id']
+                header_value = str(parent_tenant_id)
+                self._add_header(request, header_name, header_value)
+                injected_headers[header_name] = header_value
+                request.state.injected_headers[header_name] = header_value
+        
+        # Gateway identification
+        header_name = self.STANDARD_HEADERS['forwarded_by']
+        header_value = "bakery-gateway"
+        self._add_header(request, header_name, header_value)
+        injected_headers[header_name] = header_value
+        request.state.injected_headers[header_name] = header_value
+        
+        # Request ID if available
+        request_id = getattr(request.state, 'request_id', None)
+        if request_id:
+            header_name = self.STANDARD_HEADERS['request_id']
+            header_value = str(request_id)
+            self._add_header(request, header_name, header_value)
+            injected_headers[header_name] = header_value
+            request.state.injected_headers[header_name] = header_value
+        
+        logger.info("🔧 Injected context headers", 
+                   user_id=user_context.get('user_id'),
+                   user_type=user_context.get('type', ''),
+                   service_name=user_context.get('service', ''),
+                   role=user_context.get('role', ''),
+                   tenant_id=tenant_id,
+                   is_demo=is_demo,
+                   demo_session_id=user_context.get('demo_session_id', ''),
+                   path=request.url.path)
+        
+        return injected_headers
+    
+    def _add_header(self, request: Request, header_name: str, header_value: str) -> None:
+        """
+        Safely add header to request
+        """
+        try:
+            request.headers.__dict__["_list"].append((header_name.encode(), header_value.encode()))
+        except Exception as e:
+            logger.warning(f"Failed to add header {header_name}: {e}")
+    
+    def get_forwardable_headers(self, request: Request) -> Dict[str, str]:
+        """
+        Get headers that should be forwarded to downstream services
+        Includes both original request headers and injected context headers
+        """
+        forwardable_headers = {}
+        
+        # Add forwardable original headers
+        for header_name in self.FORWARDABLE_HEADERS:
+            header_value = request.headers.get(header_name)
+            if header_value:
+                forwardable_headers[header_name] = header_value
+        
+        # Add injected context headers from request.state
+        if hasattr(request.state, 'injected_headers'):
+            for header_name, header_value in request.state.injected_headers.items():
+                forwardable_headers[header_name] = header_value
+        
+        # Add authorization header if present
+        auth_header = request.headers.get('authorization')
+        if auth_header:
+            forwardable_headers['authorization'] = auth_header
+        
+        return forwardable_headers
+    
+    def get_all_headers_for_proxy(self, request: Request, 
+                                 additional_headers: Optional[Dict[str, str]] = None) -> Dict[str, str]:
+        """
+        Get complete set of headers for proxying to downstream services
+        """
+        headers = self.get_forwardable_headers(request)
+        
+        # Add any additional headers
+        if additional_headers:
+            headers.update(additional_headers)
+        
+        # Remove host header as it will be set by httpx
+        headers.pop('host', None)
+        
+        return headers
+    
+    def validate_required_headers(self, request: Request, required_headers: List[str]) -> bool:
+        """
+        Validate that required headers are present
+        """
+        missing_headers = []
+        
+        for header_name in required_headers:
+            # Check in injected headers first
+            if hasattr(request.state, 'injected_headers'):
+                if header_name in request.state.injected_headers:
+                    continue
+            
+            # Check in request headers
+            if request.headers.get(header_name):
+                continue
+            
+            missing_headers.append(header_name)
+        
+        if missing_headers:
+            logger.warning(f"Missing required headers: {missing_headers}")
+            return False
+        
+        return True
+    
+    def get_header_value(self, request: Request, header_name: str, 
+                        default: Optional[str] = None) -> Optional[str]:
+        """
+        Get header value from either injected headers or request headers
+        """
+        # Check injected headers first
+        if hasattr(request.state, 'injected_headers'):
+            if header_name in request.state.injected_headers:
+                return request.state.injected_headers[header_name]
+        
+        # Check request headers
+        return request.headers.get(header_name, default)
+    
+    def add_header_for_middleware(self, request: Request, header_name: str, header_value: str) -> None:
+        """
+        Allow middleware to add headers to the unified header system
+        This ensures all headers are available for proxying
+        """
+        # Ensure injected_headers exists
+        if not hasattr(request.state, 'injected_headers'):
+            request.state.injected_headers = {}
+        
+        # Add header to injected_headers
+        request.state.injected_headers[header_name] = header_value
+        
+        # Also add to actual request headers for compatibility
+        try:
+            request.headers.__dict__["_list"].append((header_name.encode(), header_value.encode()))
+        except Exception as e:
+            logger.warning(f"Failed to add header {header_name} to request headers: {e}")
+        
+        logger.debug(f"Middleware added header: {header_name} = {header_value}")
+
+
+# Global instance for easy access
+header_manager = HeaderManager()
--- a/gateway/app/main.py
+++ b/gateway/app/main.py
@@ -16,6 +16,7 @@ from shared.redis_utils import initialize_redis, close_redis, get_redis_client
 from shared.service_base import StandardFastAPIService

 from app.core.config import settings
+from app.core.header_manager import header_manager
 from app.middleware.request_id import RequestIDMiddleware
 from app.middleware.auth import AuthMiddleware
 from app.middleware.logging import LoggingMiddleware
@@ -50,6 +51,10 @@ class GatewayService(StandardFastAPIService):
        """Custom startup logic for Gateway"""
        global redis_client

+        # Initialize HeaderManager
+        header_manager.initialize()
+        logger.info("HeaderManager initialized")
+
        # Initialize Redis
        try:
            await initialize_redis(settings.REDIS_URL, db=0, max_connections=50)
--- a/gateway/app/middleware/auth.py
+++ b/gateway/app/middleware/auth.py
@@ -14,6 +14,7 @@ import httpx
 import json

 from app.core.config import settings
+from app.core.header_manager import header_manager
 from shared.auth.jwt_handler import JWTHandler
 from shared.auth.tenant_access import tenant_access_manager, extract_tenant_id_from_path, is_tenant_scoped_path

@@ -60,15 +61,8 @@ class AuthMiddleware(BaseHTTPMiddleware):
        if request.method == "OPTIONS":
            return await call_next(request)

-        # SECURITY: Remove any incoming x-subscription-* headers
-        # These will be re-injected from verified JWT only
-        sanitized_headers = [
-            (k, v) for k, v in request.headers.raw
-            if not k.decode().lower().startswith('x-subscription-')
-            and not k.decode().lower().startswith('x-user-')
-            and not k.decode().lower().startswith('x-tenant-')
-        ]
-        request.headers.__dict__["_list"] = sanitized_headers
+        # SECURITY: Remove any incoming sensitive headers using HeaderManager
+        header_manager.sanitize_incoming_headers(request)

        # Skip authentication for public routes
        if self._is_public_route(request.url.path):
@@ -573,109 +567,13 @@ class AuthMiddleware(BaseHTTPMiddleware):

    async def _inject_context_headers(self, request: Request, user_context: Dict[str, Any], tenant_id: Optional[str] = None):
        """
-        Inject user and tenant context headers for downstream services
-        ENHANCED: Added logging to verify header injection
+        Inject user and tenant context headers for downstream services using unified HeaderManager
        """
-        # Enhanced logging for debugging
-        logger.info(
-            "🔧 Injecting context headers",
-            user_id=user_context.get("user_id"),
-            user_type=user_context.get("type", ""),
-            service_name=user_context.get("service", ""),
-            role=user_context.get("role", ""),
-            tenant_id=tenant_id,
-            is_demo=user_context.get("is_demo", False),
-            demo_session_id=user_context.get("demo_session_id", ""),
-            path=request.url.path
-        )
-
-        # Add user context headers
-        logger.debug(f"DEBUG: Injecting headers for user: {user_context.get('user_id')}, is_demo: {user_context.get('is_demo', False)}")
-        logger.debug(f"DEBUG: request.headers object id: {id(request.headers)}, _list id: {id(request.headers.__dict__.get('_list', []))}")
-        
-        # Store headers in request.state for cross-middleware access
-        request.state.injected_headers = {
-            "x-user-id": user_context["user_id"],
-            "x-user-email": user_context["email"],
-            "x-user-role": user_context.get("role", "user")
-        }
-        
-        request.headers.__dict__["_list"].append((
-            b"x-user-id", user_context["user_id"].encode()
-        ))
-        request.headers.__dict__["_list"].append((
-            b"x-user-email", user_context["email"].encode()
-        ))
-
-        user_role = user_context.get("role", "user")
-        request.headers.__dict__["_list"].append((
-            b"x-user-role", user_role.encode()
-        ))
-
-        user_type = user_context.get("type", "")
-        if user_type:
-            request.headers.__dict__["_list"].append((
-                b"x-user-type", user_type.encode()
-        ))
-
-        service_name = user_context.get("service", "")
-        if service_name:
-            request.headers.__dict__["_list"].append((
-                b"x-service-name", service_name.encode()
-        ))
-
-        # Add tenant context if available
-        if tenant_id:
-            request.headers.__dict__["_list"].append((
-                b"x-tenant-id", tenant_id.encode()
-        ))
-
-        # Add subscription tier if available
-        subscription_tier = user_context.get("subscription_tier", "")
-        if subscription_tier:
-            request.headers.__dict__["_list"].append((
-                b"x-subscription-tier", subscription_tier.encode()
-            ))
-
-        # Add is_demo flag for demo sessions
-        is_demo = user_context.get("is_demo", False)
-        logger.debug(f"DEBUG: is_demo value: {is_demo}, type: {type(is_demo)}")
-        if is_demo:
-            logger.info(f"🎭 Adding demo session headers",
-                       demo_session_id=user_context.get("demo_session_id", ""),
-                       demo_account_type=user_context.get("demo_account_type", ""),
-                       path=request.url.path)
-            request.headers.__dict__["_list"].append((
-                b"x-is-demo", b"true"
-            ))
-        else:
-            logger.debug(f"DEBUG: Not adding demo headers because is_demo is: {is_demo}")
-
-            # Add demo session context headers for backend services
-            demo_session_id = user_context.get("demo_session_id", "")
-            if demo_session_id:
-                request.headers.__dict__["_list"].append((
-                    b"x-demo-session-id", demo_session_id.encode()
-                ))
-
-            demo_account_type = user_context.get("demo_account_type", "")
-            if demo_account_type:
-                request.headers.__dict__["_list"].append((
-                    b"x-demo-account-type", demo_account_type.encode()
-                ))
+        # Use unified HeaderManager for consistent header injection
+        injected_headers = header_manager.inject_context_headers(request, user_context, tenant_id)
        
        # Add hierarchical access headers if tenant context exists
        if tenant_id:
-            tenant_access_type = getattr(request.state, 'tenant_access_type', 'direct')
-            can_view_children = getattr(request.state, 'can_view_children', False)
-
-            request.headers.__dict__["_list"].append((
-                b"x-tenant-access-type", tenant_access_type.encode()
-            ))
-            request.headers.__dict__["_list"].append((
-                b"x-can-view-children", str(can_view_children).encode()
-            ))
-
            # If this is hierarchical access, include parent tenant ID
            # Get parent tenant ID from the auth service if available
            try:
@@ -689,17 +587,16 @@ class AuthMiddleware(BaseHTTPMiddleware):
                        hierarchy_data = response.json()
                        parent_tenant_id = hierarchy_data.get("parent_tenant_id")
                        if parent_tenant_id:
-                            request.headers.__dict__["_list"].append((
-                                b"x-parent-tenant-id", parent_tenant_id.encode()
-                            ))
+                            # Add parent tenant ID using HeaderManager for consistency
+                            header_name = header_manager.STANDARD_HEADERS['parent_tenant_id']
+                            header_value = str(parent_tenant_id)
+                            header_manager.add_header_for_middleware(request, header_name, header_value)
+                            logger.info(f"Added parent tenant ID header: {parent_tenant_id}")
            except Exception as e:
                logger.warning(f"Failed to get parent tenant ID: {e}")
                pass
        
-        # Add gateway identification
-        request.headers.__dict__["_list"].append((
-            b"x-forwarded-by", b"bakery-gateway"
-        ))
+        return injected_headers

    async def _get_tenant_subscription_tier(self, tenant_id: str, request: Request) -> Optional[str]:
        """
--- a/gateway/app/middleware/rate_limiting.py
+++ b/gateway/app/middleware/rate_limiting.py
@@ -45,8 +45,17 @@ class APIRateLimitMiddleware(BaseHTTPMiddleware):
            return await call_next(request)

        try:
-            # Get subscription tier
-            subscription_tier = await self._get_subscription_tier(tenant_id, request)
+            # Get subscription tier from headers (added by AuthMiddleware)
+            subscription_tier = request.headers.get("x-subscription-tier")
+            
+            if not subscription_tier:
+                # Fallback: get from request state if headers not available
+                subscription_tier = getattr(request.state, "subscription_tier", None)
+            
+            if not subscription_tier:
+                # Final fallback: get from tenant service (should rarely happen)
+                subscription_tier = await self._get_subscription_tier(tenant_id, request)
+                logger.warning(f"Subscription tier not found in headers or state, fetched from tenant service: {subscription_tier}")

            # Get quota limit for tier
            quota_limit = self._get_quota_limit(subscription_tier)
--- a/gateway/app/middleware/request_id.py
+++ b/gateway/app/middleware/request_id.py
@@ -9,6 +9,8 @@ from fastapi import Request
 from starlette.middleware.base import BaseHTTPMiddleware
 from starlette.responses import Response

+from app.core.header_manager import header_manager
+
 logger = structlog.get_logger()


@@ -40,11 +42,9 @@ class RequestIDMiddleware(BaseHTTPMiddleware):
        # Bind request ID to structured logger context
        logger_ctx = logger.bind(request_id=request_id)

-        # Inject request ID header for downstream services
-        # This is done by modifying the headers that will be forwarded
-        request.headers.__dict__["_list"].append((
-            b"x-request-id", request_id.encode()
-        ))
+        # Inject request ID header for downstream services using HeaderManager
+        # Note: This runs early in middleware chain, so we use add_header_for_middleware
+        header_manager.add_header_for_middleware(request, "x-request-id", request_id)

        # Log request start
        logger_ctx.info(
--- a/gateway/app/middleware/subscription.py
+++ b/gateway/app/middleware/subscription.py
@@ -15,6 +15,7 @@ import asyncio
 from datetime import datetime, timezone

 from app.core.config import settings
+from app.core.header_manager import header_manager
 from app.utils.subscription_error_responses import create_upgrade_required_response

 logger = structlog.get_logger()
@@ -178,7 +179,10 @@ class SubscriptionMiddleware(BaseHTTPMiddleware):
            r'/api/v1/subscriptions/.*',  # Subscription management itself
            r'/api/v1/tenants/[^/]+/members.*',  # Basic tenant info
            r'/docs.*',
-            r'/openapi\.json'
+            r'/openapi\.json',
+            # Training monitoring endpoints (WebSocket and status checks)
+            r'/api/v1/tenants/[^/]+/training/jobs/.*/live.*',  # WebSocket endpoint
+            r'/api/v1/tenants/[^/]+/training/jobs/.*/status.*',  # Status polling endpoint
        ]

        # Skip OPTIONS requests (CORS preflight)
@@ -275,21 +279,11 @@ class SubscriptionMiddleware(BaseHTTPMiddleware):
                        'current_tier': current_tier
                    }

-            # Use the same authentication pattern as gateway routes for fallback
-            headers = dict(request.headers)
-            headers.pop("host", None)
+            # Use unified HeaderManager for consistent header handling
+            headers = header_manager.get_all_headers_for_proxy(request)
            
            # Extract user_id for logging (fallback path)
-            user_id = 'unknown'
-            # Add user context headers if available
-            if hasattr(request.state, 'user') and request.state.user:
-                user = request.state.user
-                user_id = str(user.get('user_id', 'unknown'))
-                headers["x-user-id"] = user_id
-                headers["x-user-email"] = str(user.get('email', ''))
-                headers["x-user-role"] = str(user.get('role', 'user'))
-                headers["x-user-full-name"] = str(user.get('full_name', ''))
-                headers["x-tenant-id"] = str(user.get('tenant_id', ''))
+            user_id = header_manager.get_header_value(request, 'x-user-id', 'unknown')

            # Call tenant service fast tier endpoint with caching (fallback for old tokens)
            timeout_config = httpx.Timeout(
--- a/gateway/app/routes/auth.py
+++ b/gateway/app/routes/auth.py
@@ -13,6 +13,7 @@ from fastapi.responses import JSONResponse
 from typing import Dict, Any

 from app.core.config import settings
+from app.core.header_manager import header_manager
 from app.core.service_discovery import ServiceDiscovery
 from shared.monitoring.metrics import MetricsCollector

@@ -136,107 +137,32 @@ class AuthProxy:
        return AUTH_SERVICE_URL

    def _prepare_headers(self, headers, request=None) -> Dict[str, str]:
-        """Prepare headers for forwarding (remove hop-by-hop headers)"""
-        # Remove hop-by-hop headers
-        hop_by_hop_headers = {
-            'connection', 'keep-alive', 'proxy-authenticate',
-            'proxy-authorization', 'te', 'trailers', 'upgrade'
-        }
-
-        # Convert headers to dict - get ALL headers including those added by middleware
-        # Middleware adds headers to _list, so we need to read from there
-        logger.debug(f"DEBUG: headers type: {type(headers)}, has _list: {hasattr(headers, '_list')}, has raw: {hasattr(headers, 'raw')}")
-        logger.debug(f"DEBUG: headers.__dict__ keys: {list(headers.__dict__.keys())}")
-        logger.debug(f"DEBUG: '_list' in headers.__dict__: {'_list' in headers.__dict__}")
-        
-        if hasattr(headers, '_list'):
-            logger.debug(f"DEBUG: Entering _list branch")
-            logger.debug(f"DEBUG: headers object id: {id(headers)}, _list id: {id(headers.__dict__.get('_list', []))}")
-            # Get headers from the _list where middleware adds them
-            all_headers_list = headers.__dict__.get('_list', [])
-            logger.debug(f"DEBUG: _list length: {len(all_headers_list)}")
-            
-            # Debug: Show first few headers in the list
-            debug_headers = []
-            for i, (k, v) in enumerate(all_headers_list):
-                if i < 5:  # Show first 5 headers for debugging
+        """Prepare headers for forwarding using unified HeaderManager"""
+        # Use unified HeaderManager to get all headers
+        if request:
+            all_headers = header_manager.get_all_headers_for_proxy(request)
+            logger.debug(f"DEBUG: Added headers from HeaderManager: {list(all_headers.keys())}")
+        else:
+            # Fallback: convert headers to dict manually
+            all_headers = {}
+            if hasattr(headers, '_list'):
+                for k, v in headers.__dict__.get('_list', []):
                    key = k.decode() if isinstance(k, bytes) else k
                    value = v.decode() if isinstance(v, bytes) else v
-                    debug_headers.append(f"{key}: {value}")
-            logger.debug(f"DEBUG: First headers in _list: {debug_headers}")
+                    all_headers[key] = value
+            elif hasattr(headers, 'raw'):
+                for k, v in headers.raw:
+                    key = k.decode() if isinstance(k, bytes) else k
+                    value = v.decode() if isinstance(v, bytes) else v
+                    all_headers[key] = value
+            else:
+                # Headers is already a dict
+                all_headers = dict(headers)
        
-            # Convert to dict for easier processing
-            all_headers = {}
-            for k, v in all_headers_list:
-                key = k.decode() if isinstance(k, bytes) else k
-                value = v.decode() if isinstance(v, bytes) else v
-                all_headers[key] = value
+        # Debug logging
+        logger.info(f"📤 Forwarding headers - x_user_id: {all_headers.get('x-user-id', 'MISSING')}, x_is_demo: {all_headers.get('x-is-demo', 'MISSING')}, x_demo_session_id: {all_headers.get('x-demo-session-id', 'MISSING')}, headers: {list(all_headers.keys())}")
        
-            # Debug: Show if x-user-id and x-is-demo are in the dict
-            logger.debug(f"DEBUG: x-user-id in all_headers: {'x-user-id' in all_headers}, x-is-demo in all_headers: {'x-is-demo' in all_headers}")
-            logger.debug(f"DEBUG: all_headers keys: {list(all_headers.keys())[:10]}...")  # Show first 10 keys
-            
-            logger.info(f"📤 Forwarding headers to auth service - x_user_id: {all_headers.get('x-user-id', 'MISSING')}, x_is_demo: {all_headers.get('x-is-demo', 'MISSING')}, x_demo_session_id: {all_headers.get('x-demo-session-id', 'MISSING')}, headers: {list(all_headers.keys())}")
-            
-            # Check if headers are missing and try to get them from request.state
-            if request and hasattr(request, 'state') and hasattr(request.state, 'injected_headers'):
-                logger.debug(f"DEBUG: Found injected_headers in request.state: {request.state.injected_headers}")
-                # Add missing headers from request.state
-                if 'x-user-id' not in all_headers and 'x-user-id' in request.state.injected_headers:
-                    all_headers['x-user-id'] = request.state.injected_headers['x-user-id']
-                    logger.debug(f"DEBUG: Added x-user-id from request.state: {all_headers['x-user-id']}")
-                if 'x-user-email' not in all_headers and 'x-user-email' in request.state.injected_headers:
-                    all_headers['x-user-email'] = request.state.injected_headers['x-user-email']
-                    logger.debug(f"DEBUG: Added x-user-email from request.state: {all_headers['x-user-email']}")
-                if 'x-user-role' not in all_headers and 'x-user-role' in request.state.injected_headers:
-                    all_headers['x-user-role'] = request.state.injected_headers['x-user-role']
-                    logger.debug(f"DEBUG: Added x-user-role from request.state: {all_headers['x-user-role']}")
-                
-                # Add is_demo flag if this is a demo session
-                if hasattr(request.state, 'is_demo_session') and request.state.is_demo_session:
-                    all_headers['x-is-demo'] = 'true'
-                    logger.debug(f"DEBUG: Added x-is-demo from request.state.is_demo_session")
-            
-            # Filter out hop-by-hop headers
-            filtered_headers = {
-                k: v for k, v in all_headers.items()
-                if k.lower() not in hop_by_hop_headers
-            }
-        elif hasattr(headers, 'raw'):
-            logger.debug(f"DEBUG: Entering raw branch")
-
-            # Filter out hop-by-hop headers
-            filtered_headers = {
-                k: v for k, v in all_headers.items()
-                if k.lower() not in hop_by_hop_headers
-            }
-        elif hasattr(headers, 'raw'):
-            # Fallback to raw headers if _list not available
-            all_headers = {
-                k.decode() if isinstance(k, bytes) else k: v.decode() if isinstance(v, bytes) else v
-                for k, v in headers.raw
-            }
-            logger.info(f"📤 Forwarding headers to auth service - x_user_id: {all_headers.get('x-user-id', 'MISSING')}, x_is_demo: {all_headers.get('x-is-demo', 'MISSING')}, x_demo_session_id: {all_headers.get('x-demo-session-id', 'MISSING')}, headers: {list(all_headers.keys())}")
-
-            filtered_headers = {
-                k.decode() if isinstance(k, bytes) else k: v.decode() if isinstance(v, bytes) else v
-                for k, v in headers.raw
-                if (k.decode() if isinstance(k, bytes) else k).lower() not in hop_by_hop_headers
-            }
-        else:
-            # Handle case where headers is already a dict
-            logger.info(f"📤 Forwarding headers to auth service - x_user_id: {headers.get('x-user-id', 'MISSING')}, x_is_demo: {headers.get('x-is-demo', 'MISSING')}, x_demo_session_id: {headers.get('x-demo-session-id', 'MISSING')}, headers: {list(headers.keys())}")
-
-            filtered_headers = {
-                k: v for k, v in headers.items()
-                if k.lower() not in hop_by_hop_headers
-            }
-
-        # Add gateway identifier
-        filtered_headers['X-Forwarded-By'] = 'bakery-gateway'
-        filtered_headers['X-Gateway-Version'] = '1.0.0'
-
-        return filtered_headers
+        return all_headers

    def _prepare_response_headers(self, headers: Dict[str, str]) -> Dict[str, str]:
        """Prepare response headers"""
--- a/gateway/app/routes/demo.py
+++ b/gateway/app/routes/demo.py
@@ -8,6 +8,7 @@ import httpx
 import structlog

 from app.core.config import settings
+from app.core.header_manager import header_manager

 logger = structlog.get_logger()

@@ -29,12 +30,8 @@ async def proxy_demo_service(path: str, request: Request):
    if request.method in ["POST", "PUT", "PATCH"]:
        body = await request.body()

-    # Forward headers (excluding host)
-    headers = {
-        key: value
-        for key, value in request.headers.items()
-        if key.lower() not in ["host", "content-length"]
-    }
+    # Use unified HeaderManager for consistent header forwarding
+    headers = header_manager.get_all_headers_for_proxy(request)

    try:
        async with httpx.AsyncClient(timeout=30.0) as client:
--- a/gateway/app/routes/geocoding.py
+++ b/gateway/app/routes/geocoding.py
@@ -5,6 +5,7 @@ from fastapi.responses import JSONResponse
 import httpx
 import structlog
 from app.core.config import settings
+from app.core.header_manager import header_manager

 logger = structlog.get_logger()
 router = APIRouter()
@@ -26,12 +27,8 @@ async def proxy_geocoding(request: Request, path: str):
        if request.method in ["POST", "PUT", "PATCH"]:
            body = await request.body()

-        # Forward headers (excluding host)
-        headers = {
-            key: value
-            for key, value in request.headers.items()
-            if key.lower() not in ["host", "content-length"]
-        }
+        # Use unified HeaderManager for consistent header forwarding
+        headers = header_manager.get_all_headers_for_proxy(request)

        # Make the proxied request
        async with httpx.AsyncClient(timeout=30.0) as client:
--- a/gateway/app/routes/poi_context.py
+++ b/gateway/app/routes/poi_context.py
@@ -8,6 +8,7 @@ from fastapi.responses import JSONResponse
 import httpx
 import structlog
 from app.core.config import settings
+from app.core.header_manager import header_manager

 logger = structlog.get_logger()
 router = APIRouter()
@@ -44,12 +45,8 @@ async def proxy_poi_context(request: Request, path: str):
        if request.method in ["POST", "PUT", "PATCH"]:
            body = await request.body()

-        # Copy headers (exclude host and content-length as they'll be set by httpx)
-        headers = {
-            key: value
-            for key, value in request.headers.items()
-            if key.lower() not in ["host", "content-length"]
-        }
+        # Use unified HeaderManager for consistent header forwarding
+        headers = header_manager.get_all_headers_for_proxy(request)

        # Make the request to the external service
        async with httpx.AsyncClient(timeout=60.0) as client:
--- a/gateway/app/routes/pos.py
+++ b/gateway/app/routes/pos.py
@@ -8,6 +8,7 @@ import httpx
 import logging

 from app.core.config import settings
+from app.core.header_manager import header_manager

 logger = logging.getLogger(__name__)
 router = APIRouter()
@@ -45,9 +46,8 @@ async def _proxy_to_pos_service(request: Request, target_path: str):
    try:
        url = f"{settings.POS_SERVICE_URL}{target_path}"

-        # Forward headers
-        headers = dict(request.headers)
-        headers.pop("host", None)
+        # Use unified HeaderManager for consistent header forwarding
+        headers = header_manager.get_all_headers_for_proxy(request)

        # Add query parameters
        params = dict(request.query_params)
--- a/gateway/app/routes/subscription.py
+++ b/gateway/app/routes/subscription.py
@@ -9,6 +9,7 @@ import logging
 from typing import Optional

 from app.core.config import settings
+from app.core.header_manager import header_manager

 logger = logging.getLogger(__name__)
 router = APIRouter()
@@ -98,29 +99,13 @@ async def _proxy_request(request: Request, target_path: str, service_url: str):
    try:
        url = f"{service_url}{target_path}"

-        # Forward headers and add user/tenant context
-        headers = dict(request.headers)
-        headers.pop("host", None)
+        # Use unified HeaderManager for consistent header forwarding
+        headers = header_manager.get_all_headers_for_proxy(request)
        
-        # Add user context headers if available
-        if hasattr(request.state, 'user') and request.state.user:
-            user = request.state.user
-            headers["x-user-id"] = str(user.get('user_id', ''))
-            headers["x-user-email"] = str(user.get('email', ''))
-            headers["x-user-role"] = str(user.get('role', 'user'))
-            headers["x-user-full-name"] = str(user.get('full_name', ''))
-            headers["x-tenant-id"] = str(user.get('tenant_id', ''))
-
-            # Add subscription context headers
-            if user.get('subscription_tier'):
-                headers["x-subscription-tier"] = str(user.get('subscription_tier', ''))
-                logger.debug(f"Forwarding subscription tier: {user.get('subscription_tier')}")
-
-            if user.get('subscription_status'):
-                headers["x-subscription-status"] = str(user.get('subscription_status', ''))
-                logger.debug(f"Forwarding subscription status: {user.get('subscription_status')}")
-
-            logger.info(f"Forwarding subscription request to {url} with user context: user_id={user.get('user_id')}, email={user.get('email')}, subscription_tier={user.get('subscription_tier', 'not_set')}")
+        # Debug logging
+        user_context = getattr(request.state, 'user', None)
+        if user_context:
+            logger.info(f"Forwarding subscription request to {url} with user context: user_id={user_context.get('user_id')}, email={user_context.get('email')}, subscription_tier={user_context.get('subscription_tier', 'not_set')}")
        else:
            logger.warning(f"No user context available when forwarding subscription request to {url}")

--- a/gateway/app/routes/tenant.py
+++ b/gateway/app/routes/tenant.py
@@ -10,6 +10,7 @@ import logging
 from typing import Optional

 from app.core.config import settings
+from app.core.header_manager import header_manager

 logger = logging.getLogger(__name__)
 router = APIRouter()
@@ -715,36 +716,18 @@ async def _proxy_request(request: Request, target_path: str, service_url: str, t
    try:
        url = f"{service_url}{target_path}"

-        # Forward headers and add user/tenant context
-        headers = dict(request.headers)
-        headers.pop("host", None)
+        # Use unified HeaderManager for consistent header forwarding
+        headers = header_manager.get_all_headers_for_proxy(request)
        
-        # Add tenant ID header if provided
+        # Add tenant ID header if provided (override if needed)
        if tenant_id:
-            headers["X-Tenant-ID"] = tenant_id
+            headers["x-tenant-id"] = tenant_id
        
-        # Add user context headers if available
-        if hasattr(request.state, 'user') and request.state.user:
-            user = request.state.user
-            headers["x-user-id"] = str(user.get('user_id', ''))
-            headers["x-user-email"] = str(user.get('email', ''))
-            headers["x-user-role"] = str(user.get('role', 'user'))
-            headers["x-user-full-name"] = str(user.get('full_name', ''))
-            headers["x-tenant-id"] = tenant_id or str(user.get('tenant_id', ''))
-
-            # Add subscription context headers
-            if user.get('subscription_tier'):
-                headers["x-subscription-tier"] = str(user.get('subscription_tier', ''))
-                logger.debug(f"Forwarding subscription tier: {user.get('subscription_tier')}")
-
-            if user.get('subscription_status'):
-                headers["x-subscription-status"] = str(user.get('subscription_status', ''))
-                logger.debug(f"Forwarding subscription status: {user.get('subscription_status')}")
-
-            # Debug logging
-            logger.info(f"Forwarding request to {url} with user context: user_id={user.get('user_id')}, email={user.get('email')}, tenant_id={tenant_id}, subscription_tier={user.get('subscription_tier', 'not_set')}")
+        # Debug logging
+        user_context = getattr(request.state, 'user', None)
+        if user_context:
+            logger.info(f"Forwarding request to {url} with user context: user_id={user_context.get('user_id')}, email={user_context.get('email')}, tenant_id={tenant_id}, subscription_tier={user_context.get('subscription_tier', 'not_set')}")
        else:
-            # Debug logging when no user context available
            logger.warning(f"No user context available when forwarding request to {url}. request.state.user: {getattr(request.state, 'user', 'NOT_SET')}")
        
        # Get request body if present
@@ -782,9 +765,10 @@ async def _proxy_request(request: Request, target_path: str, service_url: str, t

                logger.info(f"Forwarding multipart request with files={list(files.keys()) if files else None}, data={list(data.keys()) if data else None}")

-                # Remove content-type from headers - httpx will set it with new boundary
-                headers.pop("content-type", None)
-                headers.pop("content-length", None)
+                # For multipart requests, we need to get fresh headers since httpx will set content-type
+                # Get all headers again to ensure we have the complete set
+                headers = header_manager.get_all_headers_for_proxy(request)
+                # httpx will automatically set content-type for multipart, so we don't need to remove it
            else:
                # For other content types, use body as before
                body = await request.body()
--- a/gateway/app/routes/user.py
+++ b/gateway/app/routes/user.py
@@ -13,6 +13,7 @@ from typing import Dict, Any
 import json

 from app.core.config import settings
+from app.core.header_manager import header_manager
 from app.core.service_discovery import ServiceDiscovery
 from shared.monitoring.metrics import MetricsCollector

@@ -136,64 +137,28 @@ class UserProxy:
        return AUTH_SERVICE_URL
    
    def _prepare_headers(self, headers, request=None) -> Dict[str, str]:
-        """Prepare headers for forwarding (remove hop-by-hop headers)"""
-        # Remove hop-by-hop headers
-        hop_by_hop_headers = {
-            'connection', 'keep-alive', 'proxy-authenticate',
-            'proxy-authorization', 'te', 'trailers', 'upgrade'
-        }
-
-        # Convert headers to dict if it's a Headers object
-        # This ensures we get ALL headers including those added by middleware
-        if hasattr(headers, '_list'):
-            # Get headers from the _list where middleware adds them
-            all_headers_list = headers.__dict__.get('_list', [])
-            
-            # Convert to dict for easier processing
-            all_headers = {}
-            for k, v in all_headers_list:
-                key = k.decode() if isinstance(k, bytes) else k
-                value = v.decode() if isinstance(v, bytes) else v
-                all_headers[key] = value
-            
-            # Check if headers are missing and try to get them from request.state
-            if request and hasattr(request, 'state') and hasattr(request.state, 'injected_headers'):
-                # Add missing headers from request.state
-                if 'x-user-id' not in all_headers and 'x-user-id' in request.state.injected_headers:
-                    all_headers['x-user-id'] = request.state.injected_headers['x-user-id']
-                if 'x-user-email' not in all_headers and 'x-user-email' in request.state.injected_headers:
-                    all_headers['x-user-email'] = request.state.injected_headers['x-user-email']
-                if 'x-user-role' not in all_headers and 'x-user-role' in request.state.injected_headers:
-                    all_headers['x-user-role'] = request.state.injected_headers['x-user-role']
-                
-                # Add is_demo flag if this is a demo session
-                if hasattr(request.state, 'is_demo_session') and request.state.is_demo_session:
-                    all_headers['x-is-demo'] = 'true'
-            
-            # Filter out hop-by-hop headers
-            filtered_headers = {
-                k: v for k, v in all_headers.items()
-                if k.lower() not in hop_by_hop_headers
-            }
-        elif hasattr(headers, 'raw'):
-            # FastAPI/Starlette Headers object - use raw to get all headers
-            filtered_headers = {
-                k.decode() if isinstance(k, bytes) else k: v.decode() if isinstance(v, bytes) else v
-                for k, v in headers.raw
-                if (k.decode() if isinstance(k, bytes) else k).lower() not in hop_by_hop_headers
-            }
+        """Prepare headers for forwarding using unified HeaderManager"""
+        # Use unified HeaderManager to get all headers
+        if request:
+            all_headers = header_manager.get_all_headers_for_proxy(request)
        else:
-            # Already a dict
-            filtered_headers = {
-                k: v for k, v in headers.items()
-                if k.lower() not in hop_by_hop_headers
-            }
+            # Fallback: convert headers to dict manually
+            all_headers = {}
+            if hasattr(headers, '_list'):
+                for k, v in headers.__dict__.get('_list', []):
+                    key = k.decode() if isinstance(k, bytes) else k
+                    value = v.decode() if isinstance(v, bytes) else v
+                    all_headers[key] = value
+            elif hasattr(headers, 'raw'):
+                for k, v in headers.raw:
+                    key = k.decode() if isinstance(k, bytes) else k
+                    value = v.decode() if isinstance(v, bytes) else v
+                    all_headers[key] = value
+            else:
+                # Headers is already a dict
+                all_headers = dict(headers)
        
-        # Add gateway identifier
-        filtered_headers['X-Forwarded-By'] = 'bakery-gateway'
-        filtered_headers['X-Gateway-Version'] = '1.0.0'
-
-        return filtered_headers
+        return all_headers
    
    def _prepare_response_headers(self, headers: Dict[str, str]) -> Dict[str, str]:
        """Prepare response headers"""
--- a/scripts/cleanup-docker.sh
+++ b/scripts/cleanup-docker.sh
@@ -0,0 +1,82 @@
+#!/bin/bash
+# Docker Cleanup Script for Local Kubernetes Development
+# This script helps prevent disk space issues by cleaning up unused Docker resources
+
+set -e
+
+echo "🧹 Docker Cleanup Script for Bakery-IA Local Development"
+echo "========================================================="
+echo ""
+
+# Check if we should run automatically or ask for confirmation
+AUTO_MODE=${1:-""}
+
+# Show current disk usage
+echo "📊 Current Docker Disk Usage:"
+docker system df
+echo ""
+
+# Check Kind node disk usage if cluster is running
+if docker ps | grep -q "bakery-ia-local-control-plane"; then
+    echo "📊 Kind Node Disk Usage:"
+    docker exec bakery-ia-local-control-plane df -h / /var | grep -E "(Filesystem|overlay|/dev/vdb1)"
+    echo ""
+fi
+
+# Calculate reclaimable space
+RECLAIMABLE=$(docker system df | grep "Images" | awk '{print $4}')
+echo "💾 Estimated reclaimable space: $RECLAIMABLE"
+echo ""
+
+# Ask for confirmation unless in auto mode
+if [ "$AUTO_MODE" != "--auto" ]; then
+    read -p "Do you want to proceed with cleanup? (y/n) " -n 1 -r
+    echo ""
+    if [[ ! $REPLY =~ ^[Yy]$ ]]; then
+        echo "❌ Cleanup cancelled"
+        exit 0
+    fi
+fi
+
+echo "🚀 Starting cleanup..."
+echo ""
+
+# Remove unused images (keep images from last 24 hours)
+echo "1️⃣ Removing unused Docker images..."
+docker image prune -af --filter "until=24h" || true
+echo ""
+
+# Remove unused volumes
+echo "2️⃣ Removing unused Docker volumes..."
+docker volume prune -f || true
+echo ""
+
+# Remove build cache
+echo "3️⃣ Removing build cache..."
+docker builder prune -af || true
+echo ""
+
+# Show results
+echo "✅ Cleanup completed!"
+echo ""
+echo "📊 Final Docker Disk Usage:"
+docker system df
+echo ""
+
+# Check Kind node disk usage if cluster is running
+if docker ps | grep -q "bakery-ia-local-control-plane"; then
+    echo "📊 Kind Node Disk Usage After Cleanup:"
+    docker exec bakery-ia-local-control-plane df -h / /var | grep -E "(Filesystem|overlay|/dev/vdb1)"
+    echo ""
+
+    # Warn if still above 80%
+    USAGE=$(docker exec bakery-ia-local-control-plane df -h /var | tail -1 | awk '{print $5}' | sed 's/%//')
+    if [ "$USAGE" -gt 80 ]; then
+        echo "⚠️  Warning: Disk usage is still above 80%. Consider:"
+        echo "   - Deleting and recreating the Kind cluster"
+        echo "   - Increasing Docker's disk allocation"
+        echo "   - Running: docker system prune -a --volumes -f"
+    fi
+fi
+
+echo "🎉 All done!"
--- a/services/forecasting/app/clients/inventory_client.py
+++ b/services/forecasting/app/clients/inventory_client.py
@@ -1,82 +0,0 @@
-# services/forecasting/app/clients/inventory_client.py
-"""
-Simple client for inventory service integration
-Used when product names are not available locally
-"""
-
-import aiohttp
-import structlog
-from typing import Optional, Dict, Any
-import os
-
-logger = structlog.get_logger()
-
-class InventoryServiceClient:
-    """Simple client for inventory service interactions"""
-    
-    def __init__(self, base_url: str = None):
-        self.base_url = base_url or os.getenv("INVENTORY_SERVICE_URL", "http://inventory-service:8000")
-        self.timeout = aiohttp.ClientTimeout(total=5)  # 5 second timeout
-    
-    async def get_product_name(self, tenant_id: str, inventory_product_id: str) -> Optional[str]:
-        """
-        Get product name from inventory service
-        Returns None if service is unavailable or product not found
-        """
-        try:
-            async with aiohttp.ClientSession(timeout=self.timeout) as session:
-                url = f"{self.base_url}/api/v1/products/{inventory_product_id}"
-                headers = {"X-Tenant-ID": tenant_id}
-                
-                async with session.get(url, headers=headers) as response:
-                    if response.status == 200:
-                        data = await response.json()
-                        return data.get("name", f"Product-{inventory_product_id}")
-                    else:
-                        logger.debug("Product not found in inventory service", 
-                                   inventory_product_id=inventory_product_id,
-                                   status=response.status)
-                        return None
-                        
-        except Exception as e:
-            logger.debug("Failed to get product name from inventory service", 
-                        inventory_product_id=inventory_product_id,
-                        error=str(e))
-            return None
-    
-    async def get_multiple_product_names(self, tenant_id: str, product_ids: list) -> Dict[str, str]:
-        """
-        Get multiple product names efficiently
-        Returns a mapping of product_id -> product_name
-        """
-        try:
-            async with aiohttp.ClientSession(timeout=self.timeout) as session:
-                url = f"{self.base_url}/api/v1/products/batch"
-                headers = {"X-Tenant-ID": tenant_id}
-                payload = {"product_ids": product_ids}
-                
-                async with session.post(url, json=payload, headers=headers) as response:
-                    if response.status == 200:
-                        data = await response.json()
-                        return {item["id"]: item["name"] for item in data.get("products", [])}
-                    else:
-                        logger.debug("Batch product lookup failed", 
-                                   product_count=len(product_ids),
-                                   status=response.status)
-                        return {}
-                        
-        except Exception as e:
-            logger.debug("Failed to get product names from inventory service", 
-                        product_count=len(product_ids),
-                        error=str(e))
-            return {}
-
-# Global client instance
-_inventory_client = None
-
-def get_inventory_client() -> InventoryServiceClient:
-    """Get the global inventory client instance"""
-    global _inventory_client
-    if _inventory_client is None:
-        _inventory_client = InventoryServiceClient()
-    return _inventory_client
--- a/services/forecasting/app/services/forecasting_alert_service.py
+++ b/services/forecasting/app/services/forecasting_alert_service.py
@@ -12,7 +12,6 @@ from datetime import datetime, timedelta
 import structlog

 from shared.messaging import UnifiedEventPublisher
-from app.clients.inventory_client import get_inventory_client

 logger = structlog.get_logger()

--- a/services/inventory/app/api/internal_alert_trigger.py
+++ b/services/inventory/app/api/internal_alert_trigger.py
@@ -30,11 +30,11 @@ async def trigger_inventory_alerts(
    - Expiring ingredients
    - Overstock situations

-    Security: Protected by X-Internal-Service header check.
+    Security: Protected by x-internal-service header check.
    """
    try:
        # Verify internal service header
-        if not request or request.headers.get("X-Internal-Service") not in ["demo-session", "internal"]:
+        if not request or request.headers.get("x-internal-service") not in ["demo-session", "internal"]:
            logger.warning("Unauthorized internal API call", tenant_id=str(tenant_id))
            raise HTTPException(
                status_code=403,
--- a/services/inventory/app/api/ml_insights.py
+++ b/services/inventory/app/api/ml_insights.py
@@ -350,7 +350,7 @@ async def generate_safety_stock_insights_internal(
    This endpoint is called by the demo-session service after cloning data.
    It uses the same ML logic as the public endpoint but with optimized defaults.

-    Security: Protected by X-Internal-Service header check.
+    Security: Protected by x-internal-service header check.

    Args:
        tenant_id: The tenant UUID
@@ -365,7 +365,7 @@ async def generate_safety_stock_insights_internal(
        }
    """
    # Verify internal service header
-    if not request or request.headers.get("X-Internal-Service") not in ["demo-session", "internal"]:
+    if not request or request.headers.get("x-internal-service") not in ["demo-session", "internal"]:
        logger.warning("Unauthorized internal API call", tenant_id=tenant_id)
        raise HTTPException(
            status_code=403,
--- a/services/procurement/app/api/internal_delivery_tracking.py
+++ b/services/procurement/app/api/internal_delivery_tracking.py
@@ -29,7 +29,7 @@ async def trigger_delivery_tracking(
    This endpoint is called by the demo session cloning process after POs are seeded
    to generate realistic delivery alerts (arriving soon, overdue, etc.).

-    Security: Protected by X-Internal-Service header check.
+    Security: Protected by x-internal-service header check.

    Args:
        tenant_id: Tenant UUID to check deliveries for
@@ -49,7 +49,7 @@ async def trigger_delivery_tracking(
    """
    try:
        # Verify internal service header
-        if not request or request.headers.get("X-Internal-Service") not in ["demo-session", "internal"]:
+        if not request or request.headers.get("x-internal-service") not in ["demo-session", "internal"]:
            logger.warning("Unauthorized internal API call", tenant_id=str(tenant_id))
            raise HTTPException(
                status_code=403,
--- a/services/procurement/app/api/ml_insights.py
+++ b/services/procurement/app/api/ml_insights.py
@@ -566,7 +566,7 @@ async def generate_price_insights_internal(
    This endpoint is called by the demo-session service after cloning data.
    It uses the same ML logic as the public endpoint but with optimized defaults.

-    Security: Protected by X-Internal-Service header check.
+    Security: Protected by x-internal-service header check.

    Args:
        tenant_id: The tenant UUID
@@ -581,7 +581,7 @@ async def generate_price_insights_internal(
        }
    """
    # Verify internal service header
-    if not request or request.headers.get("X-Internal-Service") not in ["demo-session", "internal"]:
+    if not request or request.headers.get("x-internal-service") not in ["demo-session", "internal"]:
        logger.warning("Unauthorized internal API call", tenant_id=tenant_id)
        raise HTTPException(
            status_code=403,
--- a/services/procurement/app/core/dependencies.py
+++ b/services/procurement/app/core/dependencies.py
@@ -1,42 +1,45 @@
 """
 FastAPI Dependencies for Procurement Service
+Uses shared authentication infrastructure with UUID validation
 """

-from fastapi import Header, HTTPException, status
+from fastapi import Depends, HTTPException, status
 from uuid import UUID
 from typing import Optional
 from sqlalchemy.ext.asyncio import AsyncSession

 from .database import get_db
+from shared.auth.decorators import get_current_tenant_id_dep


 async def get_current_tenant_id(
-    x_tenant_id: Optional[str] = Header(None, alias="X-Tenant-ID")
+    tenant_id: Optional[str] = Depends(get_current_tenant_id_dep)
 ) -> UUID:
    """
-    Extract and validate tenant ID from request header.
+    Extract and validate tenant ID from request using shared infrastructure.
+    Adds UUID validation to ensure tenant ID format is correct.

    Args:
-        x_tenant_id: Tenant ID from X-Tenant-ID header
+        tenant_id: Tenant ID from shared dependency

    Returns:
        UUID: Validated tenant ID

    Raises:
-        HTTPException: If tenant ID is missing or invalid
+        HTTPException: If tenant ID is missing or invalid UUID format
    """
-    if not x_tenant_id:
+    if not tenant_id:
        raise HTTPException(
            status_code=status.HTTP_400_BAD_REQUEST,
-            detail="X-Tenant-ID header is required"
+            detail="x-tenant-id header is required"
        )

    try:
-        return UUID(x_tenant_id)
+        return UUID(tenant_id)
    except (ValueError, AttributeError):
        raise HTTPException(
            status_code=status.HTTP_400_BAD_REQUEST,
-            detail=f"Invalid tenant ID format: {x_tenant_id}"
+            detail=f"Invalid tenant ID format: {tenant_id}"
        )


--- a/services/production/app/api/internal_alert_trigger.py
+++ b/services/production/app/api/internal_alert_trigger.py
@@ -31,11 +31,11 @@ async def trigger_production_alerts(
    - Equipment maintenance alerts
    - Batch start delays

-    Security: Protected by X-Internal-Service header check.
+    Security: Protected by x-internal-service header check.
    """
    try:
        # Verify internal service header
-        if not request or request.headers.get("X-Internal-Service") not in ["demo-session", "internal"]:
+        if not request or request.headers.get("x-internal-service") not in ["demo-session", "internal"]:
            logger.warning("Unauthorized internal API call", tenant_id=str(tenant_id))
            raise HTTPException(
                status_code=403,
--- a/services/production/app/api/ml_insights.py
+++ b/services/production/app/api/ml_insights.py
@@ -331,7 +331,7 @@ async def generate_yield_insights_internal(
    This endpoint is called by the demo-session service after cloning data.
    It uses the same ML logic as the public endpoint but with optimized defaults.

-    Security: Protected by X-Internal-Service header check.
+    Security: Protected by x-internal-service header check.

    Args:
        tenant_id: The tenant UUID
@@ -346,7 +346,7 @@ async def generate_yield_insights_internal(
        }
    """
    # Verify internal service header
-    if not request or request.headers.get("X-Internal-Service") not in ["demo-session", "internal"]:
+    if not request or request.headers.get("x-internal-service") not in ["demo-session", "internal"]:
        logger.warning("Unauthorized internal API call", tenant_id=tenant_id)
        raise HTTPException(
            status_code=403,
--- a/services/tenant/app/repositories/tenant_member_repository.py
+++ b/services/tenant/app/repositories/tenant_member_repository.py
@@ -204,7 +204,7 @@ class TenantMemberRepository(TenantBaseRepository):
                        f"{auth_service_url}/api/v1/auth/users/batch",
                        json={"user_ids": user_ids},
                        timeout=10.0,
-                        headers={"X-Internal-Service": "tenant-service"}
+                        headers={"x-internal-service": "tenant-service"}
                    )

                    if response.status_code == 200:
@@ -226,7 +226,7 @@ class TenantMemberRepository(TenantBaseRepository):
                                response = await client.get(
                                    f"{auth_service_url}/api/v1/auth/users/{user_id}",
                                    timeout=5.0,
-                                    headers={"X-Internal-Service": "tenant-service"}
+                                    headers={"x-internal-service": "tenant-service"}
                                )
                                if response.status_code == 200:
                                    user_data = response.json()
@@ -243,7 +243,7 @@ class TenantMemberRepository(TenantBaseRepository):
                            response = await client.get(
                                f"{auth_service_url}/api/v1/auth/users/{user_id}",
                                timeout=5.0,
-                                headers={"X-Internal-Service": "tenant-service"}
+                                headers={"x-internal-service": "tenant-service"}
                            )
                            if response.status_code == 200:
                                user_data = response.json()
--- a/services/training/app/ml/hybrid_trainer.py
+++ b/services/training/app/ml/hybrid_trainer.py
@@ -216,17 +216,24 @@ class HybridProphetXGBoost:
        Get Prophet predictions for given dataframe.

        Args:
-            prophet_result: Prophet model result from training
+            prophet_result: Prophet model result from training (contains model_path)
            df: DataFrame with 'ds' column

        Returns:
            Array of predictions
        """
-        # Get the Prophet model from result
-        prophet_model = prophet_result.get('model')
+        # Get the model path from result instead of expecting the model object directly
+        model_path = prophet_result.get('model_path')

-        if prophet_model is None:
-            raise ValueError("Prophet model not found in result")
+        if model_path is None:
+            raise ValueError("Prophet model path not found in result")
+
+        # Load the actual Prophet model from the stored path
+        try:
+            import joblib
+            prophet_model = joblib.load(model_path)
+        except Exception as e:
+            raise ValueError(f"Failed to load Prophet model from path {model_path}: {str(e)}")

        # Prepare dataframe for prediction
        pred_df = df[['ds']].copy()
@@ -273,7 +280,8 @@ class HybridProphetXGBoost:
            'reg_lambda': 1.0,  # L2 regularization
            'objective': 'reg:squarederror',
            'random_state': 42,
-            'n_jobs': -1
+            'n_jobs': -1,
+            'early_stopping_rounds': 10
        }

        # Initialize model
@@ -285,7 +293,6 @@ class HybridProphetXGBoost:
            model.fit,
            X_train, y_train,
            eval_set=[(X_val, y_val)],
-            early_stopping_rounds=10,
            verbose=False
        )

@@ -303,109 +310,86 @@ class HybridProphetXGBoost:
        train_prophet_pred: np.ndarray,
        val_prophet_pred: np.ndarray,
        prophet_result: Dict[str, Any]
-    ) -> Dict[str, float]:
+    ) -> Dict[str, Any]:
        """
-        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
+        Evaluate the overall performance of the hybrid model using threading for metrics.
        """
-        # Get actual values
-        train_actual = train_df['y'].values
-        val_actual = val_df['y'].values
+        import asyncio
        
-        # Get XGBoost predictions on residuals
+        # Get XGBoost predictions on training and validation
        X_train = train_df[self.feature_columns].values
        X_val = val_df[self.feature_columns].values
        
-        # ✅ FIX: Run blocking predict() in thread pool to avoid blocking event loop
-        import asyncio
        train_xgb_pred = await asyncio.to_thread(self.xgb_model.predict, X_train)
        val_xgb_pred = await asyncio.to_thread(self.xgb_model.predict, X_val)
        
-        # Hybrid predictions = Prophet + XGBoost residual correction
+        # Hybrid prediction = Prophet prediction + XGBoost residual prediction
        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
+        actual_train = train_df['y'].values
+        actual_val = val_df['y'].values
        
-        # 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
+        # Basic RMSE calculation
+        train_rmse = float(np.sqrt(np.mean((actual_train - train_hybrid_pred)**2)))
+        val_rmse = float(np.sqrt(np.mean((actual_val - val_hybrid_pred)**2)))
+        
+        # MAE
+        train_mae = float(np.mean(np.abs(actual_train - train_hybrid_pred)))
+        val_mae = float(np.mean(np.abs(actual_val - val_hybrid_pred)))
+        
+        # MAPE (with safety for zero sales)
+        train_mape = float(np.mean(np.abs((actual_train - train_hybrid_pred) / np.maximum(actual_train, 1))))
+        val_mape = float(np.mean(np.abs((actual_val - val_hybrid_pred) / np.maximum(actual_val, 1))))

        # 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
+        prophet_metrics = prophet_result.get("metrics", {})
+        prophet_val_mae = prophet_metrics.get("val_mae", val_mae) # Fallback to hybrid if missing
+        prophet_val_mape = prophet_metrics.get("val_mape", val_mape)
+        
+        improvement_pct = 0.0
+        if prophet_val_mape > 0:
+            improvement_pct = ((prophet_val_mape - 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
+            "train_rmse": train_rmse,
+            "val_rmse": val_rmse,
+            "train_mae": train_mae,
+            "val_mae": val_mae,
+            "train_mape": train_mape,
+            "val_mape": val_mape,
+            "prophet_val_mape": prophet_val_mape,
+            "hybrid_val_mape": val_mape,
+            "improvement_percentage": float(improvement_pct),
+            "prophet_metrics": prophet_metrics
        }

+        logger.info(
+            "Hybrid model evaluation complete",
+            val_rmse=val_rmse,
+            val_mae=val_mae,
+            val_mape=val_mape,
+            improvement=improvement_pct
+        )
+
        return metrics

    def _package_hybrid_model(
        self,
        prophet_result: Dict[str, Any],
-        metrics: Dict[str, float],
+        metrics: Dict[str, Any],
        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'),
+            'prophet_model_path': prophet_result.get('model_path'),
            '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']
-            },
+            'metrics': metrics,
            'tenant_id': tenant_id,
            'inventory_product_id': inventory_product_id,
            'trained_at': datetime.now(timezone.utc).isoformat()
@@ -426,8 +410,18 @@ class HybridProphetXGBoost:
        Returns:
            DataFrame with predictions
        """
-        # Step 1: Get Prophet predictions
-        prophet_model = model_data['prophet_model']
+        # Step 1: Get Prophet model from path and make predictions
+        prophet_model_path = model_data.get('prophet_model_path')
+        if prophet_model_path is None:
+            raise ValueError("Prophet model path not found in model data")
+
+        # Load the Prophet model from the stored path
+        try:
+            import joblib
+            prophet_model = joblib.load(prophet_model_path)
+        except Exception as e:
+            raise ValueError(f"Failed to load Prophet model from path {prophet_model_path}: {str(e)}")
+
        # ✅ FIX: Run blocking predict() in thread pool to avoid blocking event loop
        import asyncio
        prophet_forecast = await asyncio.to_thread(prophet_model.predict, future_df)
--- a/services/training/app/ml/poi_feature_integrator.py
+++ b/services/training/app/ml/poi_feature_integrator.py
@@ -43,86 +43,79 @@ class POIFeatureIntegrator:
        force_refresh: bool = False
    ) -> Optional[Dict[str, Any]]:
        """
-        Fetch POI features for tenant location.
+        Fetch POI features for tenant location (optimized for training).

-        First checks if POI context exists, if not, triggers detection.
+        First checks if POI context exists. If not, returns None without triggering detection.
+        POI detection should be triggered during tenant registration, not during training.

        Args:
            tenant_id: Tenant UUID
            latitude: Bakery latitude
            longitude: Bakery longitude
-            force_refresh: Force re-detection
+            force_refresh: Force re-detection (only use if POI context already exists)

        Returns:
-            Dictionary with POI features or None if detection fails
+            Dictionary with POI features or None if not available
        """
        try:
            # Try to get existing POI context first
-            if not force_refresh:
-                existing_context = await self.external_client.get_poi_context(tenant_id)
-                if existing_context:
-                    poi_context = existing_context.get("poi_context", {})
-                    ml_features = poi_context.get("ml_features", {})
+            existing_context = await self.external_client.get_poi_context(tenant_id)

-                    # Check if stale
-                    is_stale = existing_context.get("is_stale", False)
-                    if not is_stale:
+            if existing_context:
+                poi_context = existing_context.get("poi_context", {})
+                ml_features = poi_context.get("ml_features", {})
+
+                # Check if stale and force_refresh is requested
+                is_stale = existing_context.get("is_stale", False)
+
+                if not is_stale or not force_refresh:
+                    logger.info(
+                        "Using existing POI context",
+                        tenant_id=tenant_id,
+                        is_stale=is_stale,
+                        feature_count=len(ml_features)
+                    )
+                    return ml_features
+                else:
+                    logger.info(
+                        "POI context is stale and force_refresh=True, refreshing",
+                        tenant_id=tenant_id
+                    )
+                    # Only refresh if explicitly requested and context exists
+                    detection_result = await self.external_client.detect_poi_for_tenant(
+                        tenant_id=tenant_id,
+                        latitude=latitude,
+                        longitude=longitude,
+                        force_refresh=True
+                    )
+
+                    if detection_result:
+                        poi_context = detection_result.get("poi_context", {})
+                        ml_features = poi_context.get("ml_features", {})
                        logger.info(
-                            "Using existing POI context",
-                            tenant_id=tenant_id
+                            "POI refresh completed",
+                            tenant_id=tenant_id,
+                            feature_count=len(ml_features)
                        )
                        return ml_features
                    else:
-                        logger.info(
-                            "POI context is stale, refreshing",
+                        logger.warning(
+                            "POI refresh failed, returning existing features",
                            tenant_id=tenant_id
                        )
-                        force_refresh = True
-                else:
-                    logger.info(
-                        "No existing POI context, will detect",
-                        tenant_id=tenant_id
-                    )
-
-            # Detect or refresh POIs
-            logger.info(
-                "Detecting POIs for tenant",
-                tenant_id=tenant_id,
-                location=(latitude, longitude)
-            )
-
-            detection_result = await self.external_client.detect_poi_for_tenant(
-                tenant_id=tenant_id,
-                latitude=latitude,
-                longitude=longitude,
-                force_refresh=force_refresh
-            )
-
-            if detection_result:
-                poi_context = detection_result.get("poi_context", {})
-                ml_features = poi_context.get("ml_features", {})
-
-                logger.info(
-                    "POI detection completed",
-                    tenant_id=tenant_id,
-                    total_pois=poi_context.get("total_pois_detected", 0),
-                    feature_count=len(ml_features)
-                )
-
-                return ml_features
+                        return ml_features
            else:
-                logger.error(
-                    "POI detection failed",
+                logger.info(
+                    "No existing POI context found - POI detection should be triggered during tenant registration",
                    tenant_id=tenant_id
                )
                return None

        except Exception as e:
-            logger.error(
-                "Unexpected error fetching POI features",
+            logger.warning(
+                "Error fetching POI features - returning None",
                tenant_id=tenant_id,
-                error=str(e),
-                exc_info=True
+                error=str(e)
            )
            return None

--- a/services/training/app/services/data_client.py
+++ b/services/training/app/services/data_client.py
@@ -29,16 +29,15 @@ class DataClient:
        self.sales_client = get_sales_client(settings, "training")
        self.external_client = get_external_client(settings, "training")

+        # ExternalServiceClient always has get_stored_traffic_data_for_training method
+        self.supports_stored_traffic_data = True
+
        # Configure timeouts for HTTP clients
        self._configure_timeouts()

        # Initialize circuit breakers for external services
        self._init_circuit_breakers()

-        # Check if the new method is available for stored traffic data
-        if hasattr(self.external_client, 'get_stored_traffic_data_for_training'):
-            self.supports_stored_traffic_data = True
-
    def _configure_timeouts(self):
        """Configure appropriate timeouts for HTTP clients"""
        timeout = httpx.Timeout(
@@ -49,14 +48,12 @@ class DataClient:
        )

        # Apply timeout to clients if they have httpx clients
+        # Note: BaseServiceClient manages its own HTTP client internally
        if hasattr(self.sales_client, 'client') and isinstance(self.sales_client.client, httpx.AsyncClient):
            self.sales_client.client.timeout = timeout

        if hasattr(self.external_client, 'client') and isinstance(self.external_client.client, httpx.AsyncClient):
            self.external_client.client.timeout = timeout
-        else:
-            self.supports_stored_traffic_data = False
-            logger.warning("Stored traffic data method not available in external client")

    def _init_circuit_breakers(self):
        """Initialize circuit breakers for external service calls"""
--- a/services/training/app/services/training_orchestrator.py
+++ b/services/training/app/services/training_orchestrator.py
@@ -404,22 +404,32 @@ class TrainingDataOrchestrator:
        tenant_id: str
    ) -> Dict[str, Any]:
        """
-        Collect POI features for bakery location.
+        Collect POI features for bakery location (non-blocking).

        POI features are static (location-based, not time-varying).
+        This method is non-blocking with a short timeout to prevent training delays.
+        If POI detection hasn't been run yet, training continues without POI features.
+
+        Returns:
+            Dictionary with POI features or empty dict if unavailable
        """
        try:
            logger.info(
-                "Collecting POI features",
+                "Collecting POI features (non-blocking)",
                tenant_id=tenant_id,
                location=(lat, lon)
            )

-            poi_features = await self.poi_feature_integrator.fetch_poi_features(
-                tenant_id=tenant_id,
-                latitude=lat,
-                longitude=lon,
-                force_refresh=False
+            # Set a short timeout to prevent blocking training
+            # POI detection should have been triggered during tenant registration
+            poi_features = await asyncio.wait_for(
+                self.poi_feature_integrator.fetch_poi_features(
+                    tenant_id=tenant_id,
+                    latitude=lat,
+                    longitude=lon,
+                    force_refresh=False
+                ),
+                timeout=15.0  # 15 second timeout - POI should be cached from registration
            )

            if poi_features:
@@ -430,18 +440,24 @@ class TrainingDataOrchestrator:
                )
            else:
                logger.warning(
-                    "No POI features collected (service may be unavailable)",
+                    "No POI features collected (service may be unavailable or not yet detected)",
                    tenant_id=tenant_id
                )

            return poi_features or {}

+        except asyncio.TimeoutError:
+            logger.warning(
+                "POI collection timeout (15s) - continuing training without POI features. "
+                "POI detection should be triggered during tenant registration for best results.",
+                tenant_id=tenant_id
+            )
+            return {}
        except Exception as e:
-            logger.error(
-                "Failed to collect POI features, continuing without them",
+            logger.warning(
+                "Failed to collect POI features (non-blocking) - continuing training without them",
                tenant_id=tenant_id,
-                error=str(e),
-                exc_info=True
+                error=str(e)
            )
            return {}
    
--- a/shared/clients/base_service_client.py
+++ b/shared/clients/base_service_client.py
@@ -71,7 +71,7 @@ class ServiceAuthenticator:
        }
        
        if tenant_id:
-            headers["X-Tenant-ID"] = str(tenant_id)
+            headers["x-tenant-id"] = str(tenant_id)
            
        return headers

--- a/shared/clients/forecast_client.py
+++ b/shared/clients/forecast_client.py
@@ -351,7 +351,7 @@ class ForecastServiceClient(BaseServiceClient):
        """
        Trigger demand forecasting insights for a tenant (internal service use only).

-        This method calls the internal endpoint which is protected by X-Internal-Service header.
+        This method calls the internal endpoint which is protected by x-internal-service header.
        Used by demo-session service after cloning to generate AI insights from seeded data.

        Args:
@@ -366,7 +366,7 @@ class ForecastServiceClient(BaseServiceClient):
                endpoint=f"forecasting/internal/ml/generate-demand-insights",
                tenant_id=tenant_id,
                data={"tenant_id": tenant_id},
-                headers={"X-Internal-Service": "demo-session"}
+                headers={"x-internal-service": "demo-session"}
            )

            if result:
--- a/shared/clients/inventory_client.py
+++ b/shared/clients/inventory_client.py
@@ -766,7 +766,7 @@ class InventoryServiceClient(BaseServiceClient):
        """
        Trigger inventory alerts for a tenant (internal service use only).

-        This method calls the internal endpoint which is protected by X-Internal-Service header.
+        This method calls the internal endpoint which is protected by x-internal-service header.
        The endpoint should trigger alerts specifically for the given tenant.

        Args:
@@ -783,7 +783,7 @@ class InventoryServiceClient(BaseServiceClient):
                endpoint="inventory/internal/alerts/trigger",
                tenant_id=tenant_id,
                data={},
-                headers={"X-Internal-Service": "demo-session"}
+                headers={"x-internal-service": "demo-session"}
            )

            if result:
@@ -819,7 +819,7 @@ class InventoryServiceClient(BaseServiceClient):
        """
        Trigger safety stock optimization insights for a tenant (internal service use only).

-        This method calls the internal endpoint which is protected by X-Internal-Service header.
+        This method calls the internal endpoint which is protected by x-internal-service header.

        Args:
            tenant_id: Tenant ID to trigger insights for
@@ -833,7 +833,7 @@ class InventoryServiceClient(BaseServiceClient):
                endpoint="inventory/internal/ml/generate-safety-stock-insights",
                tenant_id=tenant_id,
                data={"tenant_id": tenant_id},
-                headers={"X-Internal-Service": "demo-session"}
+                headers={"x-internal-service": "demo-session"}
            )

            if result:
--- a/shared/clients/procurement_client.py
+++ b/shared/clients/procurement_client.py
@@ -580,7 +580,7 @@ class ProcurementServiceClient(BaseServiceClient):
        """
        Trigger delivery tracking for a tenant (internal service use only).

-        This method calls the internal endpoint which is protected by X-Internal-Service header.
+        This method calls the internal endpoint which is protected by x-internal-service header.

        Args:
            tenant_id: Tenant ID to trigger delivery tracking for
@@ -596,7 +596,7 @@ class ProcurementServiceClient(BaseServiceClient):
                endpoint="procurement/internal/delivery-tracking/trigger",
                tenant_id=tenant_id,
                data={},
-                headers={"X-Internal-Service": "demo-session"}
+                headers={"x-internal-service": "demo-session"}
            )

            if result:
@@ -632,7 +632,7 @@ class ProcurementServiceClient(BaseServiceClient):
        """
        Trigger price forecasting insights for a tenant (internal service use only).

-        This method calls the internal endpoint which is protected by X-Internal-Service header.
+        This method calls the internal endpoint which is protected by x-internal-service header.

        Args:
            tenant_id: Tenant ID to trigger insights for
@@ -646,7 +646,7 @@ class ProcurementServiceClient(BaseServiceClient):
                endpoint="procurement/internal/ml/generate-price-insights",
                tenant_id=tenant_id,
                data={"tenant_id": tenant_id},
-                headers={"X-Internal-Service": "demo-session"}
+                headers={"x-internal-service": "demo-session"}
            )

            if result:
--- a/shared/clients/production_client.py
+++ b/shared/clients/production_client.py
@@ -630,7 +630,7 @@ class ProductionServiceClient(BaseServiceClient):
        """
        Trigger production alerts for a tenant (internal service use only).

-        This method calls the internal endpoint which is protected by X-Internal-Service header.
+        This method calls the internal endpoint which is protected by x-internal-service header.
        Includes both production alerts and equipment maintenance checks.

        Args:
@@ -647,7 +647,7 @@ class ProductionServiceClient(BaseServiceClient):
                endpoint="production/internal/alerts/trigger",
                tenant_id=tenant_id,
                data={},
-                headers={"X-Internal-Service": "demo-session"}
+                headers={"x-internal-service": "demo-session"}
            )

            if result:
@@ -683,7 +683,7 @@ class ProductionServiceClient(BaseServiceClient):
        """
        Trigger yield improvement insights for a tenant (internal service use only).

-        This method calls the internal endpoint which is protected by X-Internal-Service header.
+        This method calls the internal endpoint which is protected by x-internal-service header.

        Args:
            tenant_id: Tenant ID to trigger insights for
@@ -697,7 +697,7 @@ class ProductionServiceClient(BaseServiceClient):
                endpoint="production/internal/ml/generate-yield-insights",
                tenant_id=tenant_id,
                data={"tenant_id": tenant_id},
-                headers={"X-Internal-Service": "demo-session"}
+                headers={"x-internal-service": "demo-session"}
            )

            if result: