Voice Agent Development on the VAPI Platform VAPI (Voice API) is a developer-focused infrastructure platform for building AI voice agents. Unlike no-code solutions, VAPI provides full stack control: choice of STT provider (Deepgram, AssemblyAI), LLM (GPT-4o, Claude, Llama), TTS (Eleven

Labs, Azure, OpenAI), and transport layer (WebRTC, PSTN, SIP). ### VAPI Agent ArchitecturePhone Call / WebRTC ↓ [VAPI Transport Layer] ↓ [STT: Deepgram / Whisper] ↓ [LLM: GPT-4o / Claude] ←→ [Function Calls / Tools] ↓ [TTS: ElevenLabs / Azure] ↓ Audio Response### Creating an agent via the VAPI API```python import requests from typing import Optional

class VAPIAgentBuilder: """Конструктор голосовых агентов через VAPI API"""

def __init__(self, api_key: str):
    self.api_key = api_key
    self.base_url = "https://api.vapi.ai"
    self.headers = {
        "Authorization": f"Bearer {api_key}",
        "Content-Type": "application/json"
    }

def create_assistant(self, name: str,
                      system_prompt: str,
                      model: str = "gpt-4o",
                      voice_provider: str = "elevenlabs",
                      voice_id: str = "rachel",
                      tools: Optional[list] = None) -> dict:
    """
    Создание голосового ассистента.
    tools: функции для вызова во время разговора (получение данных, запись)
    """
    assistant_config = {
        "name": name,
        "model": {
            "provider": "openai" if "gpt" in model else "anthropic",
            "model": model,
            "systemPrompt": system_prompt,
            "temperature": 0.7,
        },
        "voice": {
            "provider": voice_provider,
            "voiceId": voice_id,
            "speed": 1.0,
            "stability": 0.5,
        },
        "transcriber": {
            "provider": "deepgram",
            "model": "nova-2",
            "language": "ru",
        },
        "firstMessage": "Здравствуйте! Чем могу помочь?",
        "endCallMessage": "Спасибо за звонок. До свидания!",
        "endCallFunctionEnabled": True,
        "silenceTimeoutSeconds": 20,
        "maxDurationSeconds": 600,
    }

    if tools:
        assistant_config["model"]["tools"] = tools

    response = requests.post(
        f"{self.base_url}/assistant",
        json=assistant_config,
        headers=self.headers
    )
    return response.json()

def create_tool(self, name: str,
                 description: str,
                 parameters: dict,
                 server_url: str) -> dict:
    """
    Инструмент для агента: HTTP-вызов во время разговора.
    Типичные кейсы: проверка статуса, поиск в базе, запись заявки.
    """
    return {
        "type": "function",
        "function": {
            "name": name,
            "description": description,
            "parameters": {
                "type": "object",
                "properties": parameters,
                "required": list(parameters.keys())
            }
        },
        "server": {
            "url": server_url,
            "timeoutSeconds": 5,
        }
    }

def create_outbound_call(self, assistant_id: str,
                          phone_number: str,
                          customer_data: dict = None) -> dict:
    """Инициирование исходящего звонка с передачей контекста"""
    payload = {
        "assistantId": assistant_id,
        "customer": {
            "number": phone_number,
            "name": customer_data.get("name", "") if customer_data else "",
        },
    }

    if customer_data:
        # Передаём данные о клиенте в контекст агента
        payload["assistantOverrides"] = {
            "variableValues": customer_data
        }

    response = requests.post(
        f"{self.base_url}/call",
        json=payload,
        headers=self.headers
    )
    return response.json()

def setup_inbound_phone_number(self, phone_number: str,
                                assistant_id: str) -> dict:
    """Привязка входящего номера к ассистенту"""
    payload = {
        "number": phone_number,
        "assistantId": assistant_id,
        "fallbackDestination": {
            "type": "number",
            "number": "+1234567890"  # Fallback на живого оператора
        }
    }

    response = requests.post(
        f"{self.base_url}/phone-number",
        json=payload,
        headers=self.headers
    )
    return response.json()

### Configuring interruptions and latency VAPI allows you to fine-tune parameters that affect the naturalness of the conversation: - **`interruptionsEnabled`** — allows the user to interrupt the agent. Critical for the naturalness of the dialogue. - **`backgroundDenoisingEnabled`** — background noise filtering via Krisp. - **`numWordsToInterruptAssistant`** — how many words the user needs to interrupt the agent (1-2 recommended). - **`backchannelingEnabled`** — the agent says "uh-huh" and "understood" during pauses. ### Integration with WebRTC for web callspython

Frontend пример (TypeScript/JavaScript)

VAPI_WEB_SDK_EXAMPLE = """ import Vapi from "@vapi-ai/web";

const vapi = new Vapi("YOUR_PUBLIC_KEY");

// Начало разговора vapi.start({ assistantId: "your-assistant-id", // Или inline конфигурация ассистента });

// Слушаем события vapi.on("call-start", () => console.log("Call started")); vapi.on("call-end", () => console.log("Call ended")); vapi.on("message", (message) => { if (message.type === "transcript") { console.log(message.role, message.transcript); } if (message.type === "function-call") { // Обработка tool call на стороне клиента console.log("Tool:", message.functionCall.name); } }); """