🎤 Convonet Voice AI Productivity System
Enterprise-grade voice AI platform on Google Cloud Run: FastAPI microservices (voice-gateway, agent-llm, call-center, crm-integration), multi-LLM support (Claude, Gemini, OpenAI), WebRTC / FastAPI WebSocket voice (no LiveKit on GCP — voice-gateway-service only), domain-specific agents (Productivity, Mortgage, Healthcare), SuiteCRM integration (contacts, cases, appointments, notes), and intelligent call center with transfer context.
WebRTC / FastAPI WebSocket Voice
Low-latency browser voice via FastAPI WebSocket (voice-gateway-service). Deepgram/Cartesia streaming STT, ElevenLabs/Deepgram/Cartesia TTS, and domain-specific agents (Productivity, Mortgage, Healthcare).
Team Collaboration
Multi-tenant team management with role-based access control, shared todos, and real-time collaboration features.
Domain-Specific Agents
Productivity (todos, calendar, reminders), Mortgage (applications, DTI, documents), Healthcare with SuiteCRM. Sticky context and intelligent AI-to-human transfer via Twilio/FusionPBX.
SuiteCRM Integration
Healthcare agent creates and links Contacts, Cases, Meetings (appointments), and Notes in SuiteCRM. On transfer to the call center (e.g. extension 2001), agents see full context: patient ID, case ID, appointment ID, and call summary.
38 MCP Tools
Todos, calendar, teams, reminders, mortgage tools, healthcare + SuiteCRM (patient lookup, book appointment, log case, save call summary), call transfer. Works with Claude, Gemini, and OpenAI.
Sentry Monitoring
Production-grade error tracking, performance monitoring, and automatic thread reset recovery for reliability.
Agent Monitor
Monitor LLM interactions with voice response timing (T0→buffer→STT→agent→first audio), per-tool elapsed time, and provider/domain filtering.
Tool Execution GUI
Monitor and troubleshoot tool call executions with real-time visualization and detailed analytics.
✨ Voice AI Integration
ElevenLabs, Deepgram, and Cartesia TTS with streaming support for low-latency, natural voice responses
Emotional Voice Responses
AI detects your emotional state and responds with matching voice tone - happy, calm, empathetic, or professional.
Multi-Language Support
Automatic language detection and native-accent responses in 29+ languages including Korean, Japanese, Spanish, and more.
Voice Cloning
Clone your voice in under 1 minute. Personalize the assistant's voice per user or team for a unique experience.
Voice Preferences
Customize voice settings per user: voice selection, language, emotion sensitivity, and speaking style preferences.
Real-Time Streaming
Low-latency voice generation with natural conversation flow. Responses start speaking immediately as they're generated.
Robust Fallback
Automatic fallback to Deepgram TTS if ElevenLabs is unavailable, ensuring reliability and continuous service.
🤖 Select LLM Provider
Choose your preferred AI language model for the assistant