Initialize

Create an Agent

To create an agent, extend the AgentBase class and define your agent's behavior:

from signalwire_agents import AgentBase

class MyAgent(AgentBase):
    def __init__(self):
        super().__init__(
            name="my-agent",
            route="/agent",
            host="0.0.0.0",
            port=3000,
            use_pom=True  # Enable Prompt Object Model
        )
        
        # Define agent personality and behavior
        self.prompt_add_section("Personality", body="You are a helpful and friendly assistant.")
        self.prompt_add_section("Goal", body="Help users with their questions and tasks.")
        self.prompt_add_section("Instructions", bullets=[
            "Answer questions clearly and concisely",
            "If you don't know, say so",
            "Use the provided tools when appropriate"
        ])
        
        # Add a post-prompt for summary
        self.set_post_prompt("Please summarize the key points of this conversation.")

Run Agent

The SignalWire AI Agent SDK provides a run() method that automatically detects the execution environment and configures the agent appropriately. This method works across all deployment modes:

Deployment with `run()`

def main():
    agent = MyAgent()
    
    print("Starting agent server...")
    print("Note: Works in any deployment mode (server/CGI/Lambda)")
    agent.run()  # Auto-detects environment

if __name__ == "__main__":
    main()

The run() method automatically detects and configures for:

HTTP Server: When run directly, starts an HTTP server
CGI: When CGI environment variables are detected, operates in CGI mode
AWS Lambda: When Lambda environment is detected, configures for serverless execution

Deployment Modes

HTTP Server Mode

When run directly (e.g., python my_agent.py), the agent starts an HTTP server:

# Automatically starts HTTP server when run directly
agent.run()

CGI Mode

When CGI environment variables are present, operates in CGI mode with clean HTTP output:

# Same code - automatically detects CGI environment
agent.run()

AWS Lambda Mode

When AWS Lambda environment is detected, configures for serverless execution:

# Same code - automatically detects Lambda environment  
agent.run()

Environment Detection

The SDK automatically detects the execution environment:

Environment	Detection Method	Behavior
HTTP Server	Default when no serverless environment detected	Starts FastAPI server on specified host/port
CGI	`GATEWAY_INTERFACE` environment variable present	Processes single CGI request and exits
AWS Lambda	`AWS_LAMBDA_FUNCTION_NAME` environment variable	Handles Lambda event/context
Google Cloud	`FUNCTION_NAME` or `K_SERVICE` variables	Processes Cloud Function request
Azure Functions	`AZURE_FUNCTIONS_*` variables	Handles Azure Function request

Logging Configuration

The SDK includes a central logging system that automatically configures based on the deployment environment:

# Logging is automatically configured based on environment
# No manual setup required in most cases

# Optional: Override logging mode via environment variable
# SIGNALWIRE_LOG_MODE=off      # Disable all logging
# SIGNALWIRE_LOG_MODE=stderr   # Log to stderr
# SIGNALWIRE_LOG_MODE=default  # Use default logging
# SIGNALWIRE_LOG_MODE=auto     # Auto-detect (default)

The logging system automatically:

CGI Mode: Sets logging to 'off' to avoid interfering with HTTP headers
Lambda Mode: Configures appropriate logging for serverless environment
Server Mode: Uses structured logging with timestamps and levels
Debug Mode: Enhanced logging when debug flags are set

Create an Agent​

Run Agent​

Deployment with run()​

Deployment Modes​

HTTP Server Mode​

CGI Mode​

AWS Lambda Mode​

Environment Detection​

Logging Configuration​