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/ai_chat.py

@@ -3,8 +3,17 @@
 
 import requests
 import json
-from config import OPENROUTER_API_KEY, HERMES_API_URL, HERMES_API_KEY
-from config import QWEN_API_URL, QWEN_API_KEY, QWEN_MODEL
+import os as _os
+try:  # config del POS; con sys.path del dashboard puede resolver el config raíz
+    from config import OPENROUTER_API_KEY, HERMES_API_URL, HERMES_API_KEY
+    from config import QWEN_API_URL, QWEN_API_KEY, QWEN_MODEL
+except ImportError:
+    OPENROUTER_API_KEY = _os.environ.get('OPENROUTER_API_KEY', '')
+    HERMES_API_URL = _os.environ.get('HERMES_API_URL', '')
+    HERMES_API_KEY = _os.environ.get('HERMES_API_KEY', '')
+    QWEN_API_URL = _os.environ.get('QWEN_API_URL', '')
+    QWEN_API_KEY = _os.environ.get('QWEN_API_KEY', '')
+    QWEN_MODEL = _os.environ.get('QWEN_MODEL', 'qwen3.6')
 
 OPENROUTER_URL = "https://openrouter.ai/api/v1/chat/completions"
 HERMES_ENABLED = bool(HERMES_API_KEY and HERMES_API_URL)
@@ -394,7 +403,7 @@ def classify_part(part_number):
     
     for url, key, model_id in backends:
         _validate_model(model_id)
-        result = _post_chat_completion(url, key, model_id, messages, max_tokens=300, temperature=0.2, timeout=15)
+        result = _post_chat_completion(url, key, model_id, messages, max_tokens=800, temperature=0.2, timeout=70)
         if result is None:
             continue
         content = result["content"]
@@ -543,7 +552,7 @@ def chat_cache_clear():
     _RESPONSE_CACHE.clear()
 
 
-def chat(user_message, conversation_history=None, inventory_context=None):
+def chat(user_message, conversation_history=None, inventory_context=None, persona=None):
     """Send a message to the AI and get a response with search suggestions.
 
     Caches responses for repeated identical questions (subject to bypass
@@ -553,6 +562,7 @@ def chat(user_message, conversation_history=None, inventory_context=None):
         user_message: The user's chat message.
         conversation_history: Previous messages in the conversation.
         inventory_context: Optional inventory summary string to inject into the system prompt.
+        persona: Optional persona intro (first line of system prompt) for tenant branding.
     """
     # Cache lookup — only when there's no conversation history (stateless)
     cache_key = None
@@ -564,6 +574,9 @@ def chat(user_message, conversation_history=None, inventory_context=None):
             return cached
 
     system_content = SYSTEM_PROMPT
+    if persona:
+        # Replace the hardcoded first line (persona de marca por tenant)
+        system_content = persona + "\n" + system_content.split("\n", 1)[1]
     if inventory_context:
         system_content = SYSTEM_PROMPT + "\n\n" + inventory_context