Skip to main content
Learn how to create and configure AI Voice Agents using AgentSession.

​
Basic Agent Creation

The simplest way to create an AI Voice Agent is using AgentSession: 
from kuralit.server.agent_session import AgentSession

# Create a basic AI Voice Agent
agent = AgentSession(
    stt="deepgram/nova-2:en-US",      # Speech-to-Text
    llm="gemini/gemini-2.0-flash-001", # Large Language Model
    vad="silero/v3",                    # Voice Activity Detection
    turn_detection="multilingual/v1",   # Turn Detection
    instructions="You are a helpful AI Voice Agent assistant"
)

​
Agent Configuration

​
Required Components

  • LLM - Large Language Model (the “brain”)
  • STT - Speech-to-Text (for voice input)
  • VAD - Voice Activity Detection (detects when user speaks)
  • Turn Detection - Determines when user finishes speaking

​
Optional Components

  • Tools - Functions the agent can call
  • Instructions - Agent personality and behavior
  • Name - Agent identifier

​
Configuration Methods

agent = AgentSession(
    stt="deepgram/nova-2:en-US",
    llm="gemini/gemini-2.0-flash-001",
    vad="silero/v3",
    turn_detection="multilingual/v1",
)

​
Environment Variables

# .env file
DEEPGRAM_API_KEY=your-key
GEMINI_API_KEY=your-key
KURALIT_AGENT_INSTRUCTIONS=You are a helpful assistant.

​
Agent with Instructions

Instructions define your AI Voice Agent’s personality and behavior: 
agent = AgentSession(
    stt="deepgram/nova-2:en-US",
    llm="gemini/gemini-2.0-flash-001",
    vad="silero/v3",
    turn_detection="multilingual/v1",
    instructions="""You are a friendly customer support AI Voice Agent.
    
Your role is to:
- Help customers with their questions
- Be polite and professional
- Escalate complex issues when needed
- Provide clear, concise answers"""
)

​
Agent with Tools

Add capabilities to your agent with tools: 
from kuralit.tools import Toolkit

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

agent = AgentSession(
    stt="deepgram/nova-2:en-US",
    llm="gemini/gemini-2.0-flash-001",
    tools=[Toolkit(tools=[get_weather])],
    instructions="You are a helpful assistant with weather tools."
)

​
Complete Example

from kuralit.server.agent_session import AgentSession
from kuralit.server.websocket_server import create_app

# Create agent
agent = AgentSession(
    stt="deepgram/nova-2:en-US",
    llm="gemini/gemini-2.0-flash-001",
    vad="silero/v3",
    turn_detection="multilingual/v1",
    instructions="You are a helpful assistant.",
    name="My AI Voice Agent",
)

# Create server
app = create_app(
    api_key_validator=lambda key: key == "demo-api-key",
    agent_session=agent,
)

​
Next Steps