What are AI Voice Agents?

1. 🤖 AI Agents - The intelligence that understands and responds
2. 🛠️ Tools & Function Calling - The capabilities that let them perform actions
3. 🎤 Real-Time Voice - The interface that makes conversations natural

​

Overview

• Agents provide the conversational intelligence with context and personality
• Tools give agents capabilities through function calling (not just talking, but doing)
• Voice enables real-time spoken interaction with speech-to-text and turn detection

​

🤖 AI Agents

​

What They Do

• Understand natural language input
• Maintain conversation context across multiple turns
• Generate natural, contextual responses
• Follow instructions and personality guidelines

​

How They Work

• Instructions: Define the agent’s personality and behavior
• Context: Conversation history and session management
• Tools: Functions the agent can call to perform actions

# Creating an AI Voice Agent
from kuralit.server.agent_session import AgentSession

agent = AgentSession(
    llm="gemini/gemini-2.0-flash-001",  # The AI brain
    instructions="You are a helpful AI Voice Agent assistant",
    tools=[...],  # Tools the agent can use
    # ... voice configuration
)

​

Why It Matters

​

🛠️ Tools & Function Calling

​

What They Do

• Enable agents to perform actions (not just talk)
• Connect to APIs, databases, and external services
• Execute Python functions
• Return results that agents can use in responses

​

How They Work

# Define a tool function
def get_weather(location: str) -> str:
    """Get weather for a location."""
    # Implementation here
    return f"Weather in {location}: sunny, 22°C"

# Create a toolkit
from kuralit.tools import Toolkit

weather_tools = Toolkit(
    name="weather",
    tools=[get_weather],
    instructions="Weather tools for getting current conditions"
)

# Use with your AI Voice Agent
agent = AgentSession(
    tools=[weather_tools],
    # ...
)

​

Types of Tools

1. Custom Python Functions - Write your own functions
2. REST API Tools - Load from Postman collections
3. Toolkits - Group related tools together

​

Why It Matters

• Search the web
• Access databases
• Call APIs
• Perform calculations
• Execute any action you can code

​

🎤 Real-Time Voice

​

What It Does

• Converts speech to text in real-time
• Detects when the user is speaking
• Determines when the user has finished speaking
• Enables natural turn-taking in conversations

​

How It Works

Audio Input → VAD → STT → Turn Detection → AI Agent → Response

1. VAD (Voice Activity Detection) - Detects when speech starts and ends
2. STT (Speech-to-Text) - Converts audio to text
3. Turn Detection - Determines when the user has finished speaking
4. AI Agent - Processes the text and generates a response

# Configuring voice for your AI Voice Agent
agent = AgentSession(
    stt="deepgram/nova-2:en-US",        # Speech-to-Text
    vad="silero/v3",                     # Voice Activity Detection
    turn_detection="multilingual/v1",     # Turn Detection
    llm="gemini/gemini-2.0-flash-001",  # AI Agent
    # ...
)

​

Components

• STT Plugins: Deepgram, Google Cloud STT
• VAD Plugins: Silero (on-device, no API needed)
• Turn Detection: Multilingual turn detector

​

Why It Matters

​

How They Work Together to Create AI Voice Agents

User speaks → VAD detects speech → STT converts to text → 
Turn Detection determines end of turn → AI Agent processes → 
Agent uses Tools if needed → Agent generates response → 
Response sent back to user

​

Complete Example

# Building a complete AI Voice Agent
from kuralit.server.agent_session import AgentSession
from kuralit.tools import Toolkit

# Define tools
def get_weather(location: str) -> str:
    """Get weather for a location."""
    return f"Weather in {location}: sunny, 22°C"

# Create your complete AI Voice Agent
agent = AgentSession(
    # 🎤 Real-Time Voice
    stt="deepgram/nova-2:en-US",
    vad="silero/v3",
    turn_detection="multilingual/v1",
    
    # 🤖 AI Agent
    llm="gemini/gemini-2.0-flash-001",
    instructions="You are a helpful AI Voice Agent with weather tools",
    
    # 🛠️ Tools
    tools=[Toolkit(tools=[get_weather])]
)

# Your AI Voice Agent can now:
# - Listen to voice input
# - Understand what the user wants
# - Use tools to get information
# - Respond naturally

​

Real-World Use Cases

• Customer Support AI Voice Agents - Handle support calls with FAQ and ticket management
• Voice Assistant AI Voice Agents - Personal assistants with calendar, weather, and reminders
• Enterprise AI Voice Agents - Business applications with API integrations

​

Next Steps