Mejoras de sesion: estabilizacion, IA, Nexus Yonke y marketplace B2B

- F0 blindaje: Redis/Celery, pg_hba sin trust LAN, secretos a env 600, backups, webhook auth, jobs muertos
- F1 caja: bugs POS, reembolsos al corte, apartados, CFDI, topes de descuento en cotizacion/apartado
- F2/F7 IA de catalogo: autocategorizacion, dedup de SKUs con embeddings, OCR de facturas, descripciones, claves SAT, limpieza de fitment
- F3 IA en el flujo: vision Yonke, preguntas MeLi, venta cruzada, anomalias, sinonimos, busqueda semantica (rerank/embeddings)
- F4/F8 escala y cartera: /metrics+Prometheus+Grafana, pool, score de credito, churn, metricas multi-tenant
- Nexus Yonke (M1-M6): modelo de datos 1-de-1, alta por donador, vitrina publica, MeLi usado, consignacion, garantia por dias, reportes, etiquetas
- Marketplace B2B entre yonkes (catalogo + pedidos cross-tenant)
- Seguridad: auth en /api/admin/* (35 endpoints), alertas por tenant
- Migraciones v4.27/v4.28; timers expire-demos y rollup-metrics

pos/services/crm_engine.py

@@ -7,7 +7,16 @@ Provides:
 - Customer analytics (LTV, frequency, churn risk)
 """
 
-from datetime import datetime, timedelta
+from datetime import datetime, timedelta, timezone
+
+
+def _aware(dt):
+    """Normalize a possibly-naive datetime (psycopg2 may return either) to UTC."""
+    if dt is None:
+        return None
+    if dt.tzinfo is None:
+        return dt.replace(tzinfo=timezone.utc)
+    return dt
 
 
 # ─── Customer Activities ─────────────────────────────
@@ -302,11 +311,15 @@ def get_customer_analytics(conn, customer_id):
         WHERE customer_id = %s AND status = 'completed'
     """, (customer_id,))
     ltv, total_orders, first_purchase, last_purchase, aov = cur.fetchone()
+    first_purchase = _aware(first_purchase)
+    last_purchase = _aware(last_purchase)
+
+    now = datetime.now(timezone.utc)
 
     # Days since last purchase
     days_since_last = None
     if last_purchase:
-        days_since_last = (datetime.utcnow() - last_purchase).days
+        days_since_last = (now - last_purchase).days
 
     # Purchase frequency (orders per month)
     frequency = 0.0
@@ -314,13 +327,69 @@ def get_customer_analytics(conn, customer_id):
         months = max(1, (last_purchase - first_purchase).days / 30.0)
         frequency = round(total_orders / months, 2)
 
-    # Churn risk: no purchase in 90+ days = high, 60-90 = medium, <60 = low
-    churn_risk = 'low'
-    if days_since_last is not None:
-        if days_since_last > 90:
+    # Real inter-purchase interval per customer (A12). Uses the gaps between
+    # consecutive completed sales instead of fixed 60/90-day thresholds.
+    cur.execute("""
+        SELECT created_at FROM sales
+        WHERE customer_id = %s AND status = 'completed'
+        ORDER BY created_at
+    """, (customer_id,))
+    purchase_dates = [_aware(r[0]) for r in cur.fetchall() if r[0] is not None]
+    intervals = [
+        (purchase_dates[i] - purchase_dates[i - 1]).days
+        for i in range(1, len(purchase_dates))
+    ]
+    intervals = [d for d in intervals if d >= 0]
+    avg_interval = round(sum(intervals) / len(intervals), 1) if intervals else None
+
+    # Next-purchase prediction + churn probability.
+    predicted_next_purchase_date = None
+    days_until_next_purchase = None
+    churn_probability = 0.0
+    churn_reason = ''
+    if avg_interval and last_purchase and avg_interval > 0:
+        predicted = last_purchase + timedelta(days=round(avg_interval))
+        predicted_next_purchase_date = predicted
+        days_until_next_purchase = (predicted - now).days
+        # How many of its own typical cycles the customer has been silent.
+        ratio = (days_since_last or 0) / max(float(avg_interval), 1.0)
+        if ratio > 2.0:
             churn_risk = 'high'
+            churn_probability = 0.9
+            churn_reason = (f"Lleva {days_since_last} días sin comprar, más de 2× "
+                            f"su intervalo habitual ({avg_interval:.0f} días).")
+        elif ratio > 1.5:
+            churn_risk = 'high'
+            churn_probability = 0.7
+            churn_reason = (f"Rebasó 1.5× su intervalo habitual de compra "
+                            f"({avg_interval:.0f} días).")
+        elif ratio > 1.0:
+            churn_risk = 'medium'
+            churn_probability = 0.4
+            churn_reason = f"Ya superó su intervalo habitual de compra ({avg_interval:.0f} días)."
+        else:
+            churn_risk = 'low'
+            churn_probability = 0.15
+            churn_reason = f"Dentro de su intervalo habitual de compra ({avg_interval:.0f} días)."
+    else:
+        # Fewer than two purchases: not enough history for a cadence. Fall back
+        # to the legacy fixed thresholds but report that we lacked data.
+        if days_since_last is None:
+            churn_risk = 'low'
+            churn_probability = 0.0
+            churn_reason = 'Sin compras registradas.'
+        elif days_since_last > 90:
+            churn_risk = 'high'
+            churn_probability = 0.75
+            churn_reason = f"Sin compras desde hace {days_since_last} días."
         elif days_since_last > 60:
             churn_risk = 'medium'
+            churn_probability = 0.45
+            churn_reason = f"Sin compras desde hace {days_since_last} días."
+        else:
+            churn_risk = 'low'
+            churn_probability = 0.15
+            churn_reason = f"Compra reciente (hace {days_since_last} días)."
 
     # Favorite categories (from inventory via sale_items)
     cur.execute("""
@@ -362,6 +431,15 @@ def get_customer_analytics(conn, customer_id):
         'days_since_last_purchase': days_since_last,
         'purchase_frequency_monthly': frequency,
         'churn_risk': churn_risk,
+        # A12 additions: real inter-purchase cadence + prediction.
+        'avg_days_between_purchases': avg_interval,
+        'purchase_intervals_count': len(intervals),
+        'predicted_next_purchase_date': (
+            predicted_next_purchase_date.isoformat() if predicted_next_purchase_date else None
+        ),
+        'days_until_next_purchase': days_until_next_purchase,
+        'churn_probability': churn_probability,
+        'churn_reason': churn_reason,
         'favorite_categories': categories,
         'loyalty': {
             'points_balance': loyalty_balance,