Viktoria is an AI-powered voice agent that handles Tier-1 customer inquiries for Dormero Hotels. Guests call in and speak naturally — the system understands their question, searches a semantic knowledge base, and responds in real-time via voice. A Control Center dashboard lets hotel staff monitor every call, track feedback, and manage the knowledge base.
The system processes calls end-to-end: speech recognition via Whisper, intent handling with GPT-4o, knowledge retrieval from ChromaDB, and voice synthesis — all in under 2 seconds per turn. The dashboard tracks call volume, average duration, resolution rates, and guest satisfaction over time.
Technical Highlights
- Real-time voice calls via ElevenLabs Conversational AI and WebRTC
- Semantic knowledge base powered by ChromaDB with automatic embedding
- Call analytics dashboard — volume trends, duration, resolution rates
- Feedback tracking system with per-call ratings
- FastAPI backend with SQLAlchemy + SQLite for call persistence
- Next.js 14 frontend with TypeScript and Recharts visualizations
Stack
- Voice: ElevenLabs Conversational AI
- Backend: Python, FastAPI, SQLAlchemy
- Frontend: Next.js 14, TypeScript, Tailwind CSS
- Knowledge: ChromaDB vector database
- Charts: Recharts