PydanticAI

Introduction

Logfire supports instrumenting PydanticAI.

import logfire
from pydantic_ai import Agent, RunContext

logfire.configure()
Agent.instrument_all()

Note

You will need to provide your LOGFIRE_TOKEN and LLM-appropriate API key environment variables.

With that you get:

• A span around all LLM calls your agent makes which records duration and captures any exceptions that might occur
• Human-readable display of the conversation with the agent
• Details of the request/response, including the number of tokens used

You can also instrument specific agents with this approach:

import logfire
from pydantic_ai import Agent, RunContext

logfire.configure()


roulette_agent = Agent(
    'openai:gpt-4o',
    deps_type=int,
    result_type=bool,
    system_prompt=(
        'Use the `roulette_wheel` function to see if the '
        'customer has won based on the number they provide.'
    ),
    instrument=True
)


@roulette_agent.tool
async def roulette_wheel(ctx: RunContext[int], square: int) -> str:
    """check if the square is a winner"""
    return 'winner' if square == ctx.deps else 'loser'


# Run the agent
success_number = 18
result = roulette_agent.run_sync('Put my money on square eighteen', deps=success_number)
print(result.data)
#> True

result = roulette_agent.run_sync('I bet five is the winner', deps=success_number)
print(result.data)
#> False

The above example displays like this in Logfire:

PydanticAI instrumented and displayed in Logfire

You can use PydanticAI with a large variety of LLMs, the example just happens to show gpt-4o.

For more information, see the PydanticAI docs on instrumenting with Logfire.