Trinity Agent Monitoring SDK

A comprehensive Python SDK for monitoring AI agents with advanced tracing, metrics collection, and observability capabilities powered by Langfuse.

Features

• Comprehensive Agent Monitoring: Track agent execution, tool usage, and performance metrics
• Rich Metrics Collection: Monitor tokens, costs, latency, and tool utilization
• Tool Usage Analytics: Detailed analysis of which tools your agents use and how often
• Easy Integration: Simple decorator-based monitoring with minimal code changes

Installation

pip install trinity-agent-monitoring

Quick Start

1. Initialize the Monitor

from trinity_agent_monitoring import create_monitor, set_monitor

# Create and set up the global monitor
monitor = create_monitor(
    public_key="your_trinity_public_key",
    secret_key="your_trinity_secret_key"
)

# Set as global monitor for easy access
set_monitor(monitor)

2. Monitor Your Agent

from trinity_agent_monitoring import get_monitor

# Get the global monitor
monitor = get_monitor()

# Use the decorator to monitor your agent
@monitor.monitor_agent(agent_name="my_ai_agent")
def my_agent_function(agent_executor,user_input):
    # Your agent logic here
    response = process_user_input(user_input)
    return response

# Call your monitored function
result = my_agent_function(agent_executor, "Hello, agent!")

3. Instance-based Monitoring

from trinity_agent_monitoring import TrinityAgentMonitor

# Create a dedicated monitor instance
monitor = TrinityAgentMonitor(
    public_key="your_public_key",
    secret_key="your_secret_key"
)

# Use with your agent
@monitor.monitor_agent("custom_agent")
def custom_agent_function():
    # Agent logic
    pass

Get Callback Handler for LangChain

# Get the callback handler for LangChain integration
callback_handler = monitor.get_callback_handler()

# Use with your LangChain agent
from langchain.agents import AgentExecutor
from langchain.llms import OpenAI

agent = AgentExecutor.from_agent_and_tools(
    agent=your_agent,
    tools=your_tools,
    callbacks=[callback_handler]
)

API Reference

TrinityAgentMonitor

Main monitoring class that provides comprehensive agent monitoring capabilities.

Utility Functions

• create_monitor(public_key, secret_key, host): Create a new monitor instance
• get_monitor(): Get the global monitor instance
• set_monitor(monitor): Set the global monitor instance

Metrics Collected

The SDK automatically collects the following metrics:

• Token Usage: Input, output, and total tokens
• Cost Analysis: Calculated costs for each operation
• Latency: Start time, end time, and duration
• Tool Usage: Which tools were used and how many times
• Input/Output: Raw input and output data for analysis
• Metadata: Custom metadata for additional context

Support

For support and questions, contact us at support@giggso.com