1. Dynamic Multi-Tenant Core
Zero hardcoded assumptions. The system dynamically reads legal entity classifications (LLC, Agency, Corporation, Government, NGO), business domains (E-Commerce, Software, Real Estate, Healthcare), brand names, and bot personas directly from database configuration. Switch corporate identities or white-label instances instantly without touching a single line of backend logic.
2. Visual Admin Control Panel
Empower business managers with complete zero-code operational control. Configure bot personas, manage real-time WhatsApp alert destinations, inspect customer leads, monitor server load capacities, adjust session inactivity timers, and manage ban lists directly from a sleek web interface without ever diving into complex node architectures.
3. Multimodal & Voice Intelligence
Seamlessly ingest raw voice notes, audio messages, and high-resolution media. Built-in audio filters and local media analyzers perform OCR, extract architectural specs and wireframes, and chunk complex PDF requirement documents into structured semantic entities for instant agent reasoning.
4. Automated Validation & WhatsApp Alerts
Autonomous customer verification without repetitive questioning. Global phone standardization (NumVerify E.164), email token authentication with bounce tracking, and live social URL verification. When an order or ticket is confirmed, instant alerts are dispatched via CallMeBot WhatsApp API directly to managerial devices.
5. Dual-Tier Vector Memory & RAG
Powered by local Ollama embeddings (nomic-embed-text) and PostgreSQL vector storage. Features a dual-tier recall pipeline: short-term session chunk lookup for ongoing conversations (≤ 30 days) and an autonomous daily memory archiver for long-term customer behavioral insights (> 30 days).
6. Security & Anti-Injection Guard
Hardened enterprise security protocol. The active Injection Guard prevents prompt injection, system prompt leakage, and jailbreak exploits. Includes an automated ban subsystem with cooldown timers, transactional logging in PostgreSQL, and strict zero-hallucination ordering gates.
End-to-End Autonomous Lifecycle
Customer interacts via WhatsApp, Messenger, or Web. The system dynamically adopts Lebanese Arabic or British English, evaluates past conversational memory, and executes human typing cadences.
Documents, audio notes, and wireframes are analyzed in real time. Contact parameters (phone, email, websites) undergo automated external verification and database persistence.
A unique cryptographic Order ID is generated and returned to the client. Simultaneously, a high-priority alert is dispatched via WhatsApp to the management team and logged into the Admin Dashboard.
Administrators monitor conversations, manage alert channels, inspect lead data, and configure bot behavior in real time from the visual control panel without touching code.