Tier 1: chat quality — fuzzy aliases, reply templates, dedup, rewriter

Foco: matar repetición y adaptar respuestas. Los handlers tenían ~30 strings
hardcodeadas (3-7 lugares cada una). Aliases hacían substring exacto.

- pg_trgm + GIN indexes en product_aliases / alias_product_mappings.
  Captura plurales, diminutivos, typos sin reglas. catalogRetrieval re-busca
  el snapshot con normalized_alias cuando el query original no rinde
  (vasio→vacio→Vacío).
- reply_templates table + replyTemplates.js. 20 keys, 2-3 variantes c/u
  con DEFAULTS hardcodeados como fallback. pickVariant excluye las usadas
  en context.recent_replies (FIFO cap 8). Wired en idle/cart/cartHelpers/
  shipping/payment/waiting.
- failed_searches counter en context. count>=3 escala via humanFallback.
  Reset en cada add_to_cart exitoso.
- storeContext.js: vars derivadas de getStoreConfig (delivery_zones, hours,
  zonas) listas para inyectar en templates cuando los datos se carguen.
- replyRewriter.js: LLM call opcional (REPLY_REWRITER=1) que adapta el
  template al hilo conversacional. 1.5s timeout, fallback al template puro.
  Sólo activo en 8 slots semánticamente importantes.
- 12 unit tests para replyTemplates (rotation, recency, FIFO, vars).
  208 tests totales pasando.

Plan completo: ~/.claude/plans/ok-creo-que-tiene-humming-sutton.md

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

CLAUDE.md

@@ -27,6 +27,17 @@ npm run seed         # Seed a tenant via scripts/seed-tenant.mjs
 
 No lint command is configured.
 
+## Product goal
+
+The bot must be **conversational and intelligent**, not a menu-driven flow. Customers reach out via WhatsApp **with intent to buy** — the bot's job is to:
+
+1. **Engage in conversation** — answer questions about products, prices, availability/stock; recommend; clarify.
+2. **Take orders** — build a cart through natural dialogue (multi-product turns, quantities, units).
+3. **Collect delivery data** — address, delivery vs pickup, payment method.
+4. **Operate within store rules** — delivery zones, days/hours, pickup windows. These config tables (`delivery_zones`, store schedule in `tenant_settings`) will be populated later; the bot has to read and respect them when present.
+
+Repetitive, hardcoded responses are a known quality problem and the focus of the active improvement plan (see `~/.claude/plans/ok-creo-que-tiene-humming-sutton.md`). The system is **not yet in production** — refactors that change behavior are acceptable.
+
 ## Architecture
 
 This is a **multi-tenant WhatsApp e-commerce chatbot** powered by Express.js. Tenants are WooCommerce store operators; their customers interact via WhatsApp to browse products, build carts, and place orders. All database operations are isolated by `tenant_id`.