Agent Metrics and Performance Monitoring

This example demonstrates how to collect and analyze agent metrics including message-level metrics, run metrics, and session metrics for performance monitoring.

Code

agent_metrics.py

from agno.agent import Agent
from agno.db.postgres import PostgresDb
from agno.models.openai import OpenAIChat
from agno.tools.duckduckgo import DuckDuckGoTools
from agno.utils import pprint

agent = Agent(
    model=OpenAIChat(id="gpt-5-mini"),
    tools=[DuckDuckGoTools(stock_price=True)],
    markdown=True,
    session_id="test-session-metrics",
    db=PostgresDb(db_url="postgresql+psycopg://ai:ai@localhost:5532/ai"),
)

# Get the run response directly from the non-streaming call
run_response = agent.run("What is the stock price of NVDA")
print("Tool execution completed successfully!")

# Print metrics per message
if run_response and run_response.messages:
    for message in run_response.messages:
        if message.role == "assistant":
            if message.content:
                print(
                    f"Message: {message.content[:100]}..."
                )  # Truncate for readability
            elif message.tool_calls:
                print(f"Tool calls: {len(message.tool_calls)} tool call(s)")
            print("---" * 5, "Message Metrics", "---" * 5)
            if message.metrics:
                pprint(message.metrics)
            else:
                print("No metrics available for this message")
            print("---" * 20)

# Print the run metrics
print("---" * 5, "Run Metrics", "---" * 5)
if run_response and run_response.metrics:
    pprint(run_response.metrics)
else:
    print("No run metrics available")

# Print the session metrics
print("---" * 5, "Session Metrics", "---" * 5)
try:
    session_metrics = agent.get_session_metrics()
    if session_metrics:
        pprint(session_metrics)
    else:
        print("No session metrics available")
except Exception as e:
    print(f"Error getting session metrics: {e}")

Usage

1. Create a virtual environment

   Open the Terminal and create a python virtual environment.

   Mac

   python3 -m venv .venv
   source .venv/bin/activate
   
2. Install libraries

   pip install -U agno openai ddgs psycopg
   
3. Setup PostgreSQL

   # Make sure PostgreSQL is running
   # Update connection string in the code as needed
   
4. Export your OpenAI API key

   Mac/Linux

     export OPENAI_API_KEY="your_openai_api_key_here"
   
5. Create a Python file

   Create a Python file and add the above code.

   touch agent_metrics.py
   
6. Run Agent

   Mac

   python agent_metrics.py
   
7. Find All Cookbooks

   Explore all the available cookbooks in the Agno repository. Click the link below to view the code on GitHub:

   Agno Cookbooks on GitHub