SIP-Integration
SMartrPhone nutzt SIP (Session Initiation Protocol) für die Telefonie-Integration. Unterstützte Backends:
Asterisk → Open-Source PBX, vollständige SIP-Unterstützung
FreeSWITCH → Open-Source PBX, hochskalierbar
Twilio → Cloud-Telefonie, globale Abdeckung
Sipgate → Deutsche Cloud-Telefonie
Custom SIP → Jeder SIP-kompatible Provider
STT/TTS-Pipeline
Audio Input (SIP, 16kHz mono)
→ VAD (Voice Activity Detection)
→ Whisper STT (Streaming, 200ms Chunks)
→ NLP Intent Detection
→ Response Generation (Agent Zero)
→ Neural TTS (Coqui/Edge-TTS)
→ Audio Output (SIP, 16kHz mono)
→ CRM Sync (async)
Die Pipeline läuft mit unter 500ms End-to-End-Latenz. Whisper STT liefert über 95% Genauigkeit für alle 8 Sprachen. Neural TTS sorgt für natürliche Stimme ohne Robotik-Effekt.
API-Beispiel: Anruf-Status abrufen
POST /api/v1/agents/smartrphone/query
Content-Type: application/json
Authorization: Bearer <JWT>
{
"action": "get_call_summary",
"call_id": "CALL-2026-07-11-0042",
"include_transcript": true
}
JSON-Response
{
"status": "success",
"data": {
"call_id": "CALL-2026-07-11-0042",
"caller": { "name": "Thomas Schmidt", "phone": "+491701234567", "crm_id": "CONTACT-8942" },
"duration_seconds": 127,
"language_detected": "de",
"intent": "appointment_scheduling",
"outcome": "appointment_booked",
"appointment": { "date": "2026-07-12", "time": "14:00", "duration": 60, "calendar": "[email protected]" },
"sentiment": 0.78,
"transcript_summary": "Kunde möchte Beratung für Küchenrenovierung. Termin für morgen 14:00 Uhr gebucht. Budget: 15.000-25.000 EUR.",
"crm_synced": true
},
"audit_trail": {
"agent_id": "smartrphone-001",
"stt_engine": "whisper-large-v3",
"tts_engine": "coqui-xtts-v2",
"reasoning_steps": 6,
"model": "glm-5.2",
"tokens": { "input": 1200, "output": 400 }
}
}
CRM-Connectoren
SMartrPhone CRM Integration
├── Salesforce → REST API, Sync alle 5s
├── HubSpot → REST API, Webhooks
├── Pipedrive → REST API
├── Zoho CRM → REST API
├── Microsoft Dynamics → OData
├── Custom CRM → Webhook + REST
└── No CRM → Interne SQLite-Datenbank
Jeder Connector erstellt oder aktualisiert Kontakte, fügt Anruf-Aktivitäten hinzu und speichert Transkriptionen. Sync erfolgt asynchron, um die Gesprächsführung nicht zu verzögern.
SIP Integration
SMartrPhone uses SIP (Session Initiation Protocol) for telephony integration. Supported backends:
Asterisk → Open-source PBX, full SIP support
FreeSWITCH → Open-source PBX, highly scalable
Twilio → Cloud telephony, global coverage
Sipgate → German cloud telephony
Custom SIP → Any SIP-compatible provider
STT/TTS Pipeline
Audio Input (SIP, 16kHz mono)
→ VAD (Voice Activity Detection)
→ Whisper STT (Streaming, 200ms chunks)
→ NLP Intent Detection
→ Response Generation (Agent Zero)
→ Neural TTS (Coqui/Edge-TTS)
→ Audio Output (SIP, 16kHz mono)
→ CRM Sync (async)
The pipeline runs with under 500ms end-to-end latency. Whisper STT delivers over 95% accuracy for all 8 languages. Neural TTS ensures natural voice without robotic effect.
API Example: Query Call Status
POST /api/v1/agents/smartrphone/query
Content-Type: application/json
Authorization: Bearer <JWT>
{
"action": "get_call_summary",
"call_id": "CALL-2026-07-11-0042",
"include_transcript": true
}
JSON Response
{
"status": "success",
"data": {
"call_id": "CALL-2026-07-11-0042",
"caller": { "name": "Thomas Schmidt", "phone": "+491701234567", "crm_id": "CONTACT-8942" },
"duration_seconds": 127,
"language_detected": "de",
"intent": "appointment_scheduling",
"outcome": "appointment_booked",
"appointment": { "date": "2026-07-12", "time": "14:00", "duration": 60, "calendar": "[email protected]" },
"sentiment": 0.78,
"transcript_summary": "Customer wants consultation for kitchen renovation. Appointment booked for tomorrow 2 PM. Budget: 15,000-25,000 EUR.",
"crm_synced": true
},
"audit_trail": {
"agent_id": "smartrphone-001",
"stt_engine": "whisper-large-v3",
"tts_engine": "coqui-xtts-v2",
"reasoning_steps": 6,
"model": "glm-5.2",
"tokens": { "input": 1200, "output": 400 }
}
}
CRM Connectors
SMartrPhone CRM Integration
├── Salesforce → REST API, sync every 5s
├── HubSpot → REST API, webhooks
├── Pipedrive → REST API
├── Zoho CRM → REST API
├── Microsoft Dynamics → OData
├── Custom CRM → Webhook + REST
└── No CRM → Internal SQLite database
Each connector creates or updates contacts, adds call activities and stores transcriptions. Sync occurs asynchronously to not delay the conversation.
