AI Integration into Cloud PBX (Mango Office, Zadarma, Sipuni)

AI Integration into Cloud PBX (Mango Office, Zadarma, Sipuni) Russian cloud PBXs (Mango Office, Zadarma, UIS, Sipuni) provide APIs for retrieving call recordings and managing routing. AI integration adds auto-transcription, sentiment analysis, and smart routing. ### Integration Architectures Post-call analysis is the simplest option:

Завершённый звонок → Webhook → Скачиваем запись → STT → NLP → CRM
```**Real-time via SIP trunk** – for advanced scenarios:```
Звонок → АТС SIP-транк → Media Gateway → AI Server
```### Mango Office API Integration```python
import requests
import hashlib
import json

MANGO_API_KEY = "your_api_key"
MANGO_API_SALT = "your_salt"

def sign_request(json_string: str) -> str:
    return hashlib.sha256(
        f"{MANGO_API_KEY}{json_string}{MANGO_API_SALT}".encode()
    ).hexdigest()

def get_call_recording(call_id: str) -> bytes:
    """Получаем запись завершённого звонка"""
    payload = json.dumps({
        "recording_id": call_id,
        "action": "download"
    })
    response = requests.post(
        "https://app.mango-office.ru/vpbx/queries/recording/post_load",
        data={
            "vpbx_api_key": MANGO_API_KEY,
            "sign": sign_request(payload),
            "json": payload
        }
    )
    return response.content

async def process_completed_call(call_data: dict):
    """Webhook обработчик завершённого звонка"""
    call_id = call_data["recording"]["id"]
    recording = get_call_recording(call_id)

    # STT
    transcript = await transcribe_audio(recording)

    # Анализ
    analysis = await analyze_call(transcript)

    # Обновляем CRM
    await crm.update_call(
        call_id=call_data["entry_id"],
        transcript=transcript,
        sentiment=analysis["sentiment"],
        topics=analysis["topics"],
        summary=analysis["summary"]
    )
```### Zadarma API```python
import hmac
import hashlib

def get_zadarma_recording(call_id: str) -> str:
    """Получаем ссылку на запись"""
    params = {"call_id": call_id}
    sign_string = "&".join(f"{k}={v}" for k, v in sorted(params.items()))
    signature = hmac.new(
        ZADARMA_SECRET.encode(), sign_string.encode(), hashlib.sha1
    ).hexdigest()

    response = requests.get(
        "https://api.zadarma.com/v1/pbx/record/download/",
        params={**params, "format": "mp3"},
        headers={"Authorization": f"{ZADARMA_KEY}:{signature}"}
    )
    return response.json().get("link")
```### Smart routing via API```python
# Sipuni: изменяем правило маршрутизации на основе AI-предикции
def set_routing_rule(caller_phone: str, target_queue: str):
    requests.post(
        f"https://sipuni.com/api/routing/set",
        json={"phone": caller_phone, "target": target_queue},
        auth=(SIPUNI_USER, SIPUNI_SECRET)
    )
```Timeframe: Post-call analysis – 1–2 weeks. Real-time integration via SIP – 1.5–2 months.