Aurora — AI Field Assistant

Context

Field technicians wasted valuable time on bureaucracy: calling coordination for client data, manual paper reports, and back-office dependencies for every registration.

Solution

Aurora is a multi-agent system with 17 specialized tools (sub-workflows in n8n). It handles CRUD operations across PostgreSQL, generates signed PDF reports via Supabase Edge Functions, transcribes voice messages with Gemini, and enforces RBAC at every layer — all through a Chatwoot/WhatsApp interface.

Results

• Self-sufficient field technicians — no office dependence
• Reports generated in minutes instead of days
• Full traceability with digital signatures and automatic email delivery
• Role-based access enforced at every layer
• WhatsApp-based password recovery without admin intervention

Architecture

Chatwoot Webhook → Main Agent (GPT-4o)
  ├─ Validar Rol (PostgreSQL)
  ├─ Spelling Correction (OpenAI)
  ├─ Aurora Agent (17 tools)
  │   ├─ consultar_clientes (fuzzy search with pg_trgm)
  │   ├─ registrar_cliente / sede / equipo / cronograma
  │   ├─ actualizar_* (role-gated updates)
  │   ├─ delete_all_data_client / delete_only_* 
  │   └─ generar_informe_aire_acondicionado / obra_civil
  └─ Response Parser (smart message splitting)
      └─ Chatwoot → Technician