Skip to main content

Pydantic AI via OpenTelemetry

The same agent as the Pydantic AI Quickstart, traced through OpenTelemetry instead. Nothing from TraceAgent is imported: the agent only exports OpenTelemetry traces to the server. It uses a fake model, so it runs without an API key.

Run​

pippip install pydantic-ai-slim opentelemetry-sdk opentelemetry-exporter-otlp-proto-http
uvuv add pydantic-ai-slim opentelemetry-sdk opentelemetry-exporter-otlp-proto-http

Start the TraceAgent server at http://127.0.0.1:8000, save the script below as otel_quickstart.py and run:

python otel_quickstart.py

The Script​

from opentelemetry.exporter.otlp.proto.http.trace_exporter import OTLPSpanExporter
from opentelemetry.sdk.trace import TracerProvider
from opentelemetry.sdk.trace.export import BatchSpanProcessor
from pydantic_ai import Agent
from pydantic_ai.capabilities import Instrumentation
from pydantic_ai.messages import ModelResponse, TextPart, ToolCallPart
from pydantic_ai.models.function import FunctionModel
from pydantic_ai.models.instrumented import InstrumentationSettings


def fake_model(messages, info):
    """Stands in for an LLM: asks for the tool first, then answers."""
    if len(messages) == 1:
        return ModelResponse(parts=[ToolCallPart("weather", {"city": "Madrid"})])
    return ModelResponse(parts=[TextPart("It is sunny in Madrid, 23 degrees.")])


# 1. Send OpenTelemetry traces to the TraceAgent server. Nothing from TraceAgent is imported.
provider = TracerProvider()
provider.add_span_processor(BatchSpanProcessor(OTLPSpanExporter(endpoint="http://127.0.0.1:8000/v1/traces")))

# 2. Your usual agent, with Pydantic AI's own instrumentation.
agent = Agent(
    FunctionModel(fake_model),  # use a real model string, e.g. "openai:gpt-4o", in your own agent
    name="otel-quickstart",
    capabilities=[Instrumentation(settings=InstrumentationSettings(tracer_provider=provider))],
)


@agent.tool_plain
def weather(city: str) -> str:
    """Return the weather of a city."""
    return f"Sunny, 23 degrees in {city}."


# 3. Use it as always.
result = agent.run_sync("What is the weather in Madrid?")
print(result.output)

# 4. A short script must flush before exiting, or the traces are lost.
provider.shutdown()

Output:

It is sunny in Madrid, 23 degrees.

What It Does​

  1. Configures an OpenTelemetry exporter that points at the TraceAgent server
  2. Builds a Pydantic AI agent with its own instrumentation
  3. The fake model asks for the weather tool, then answers
  4. Flushes the exporter so no span is lost

The server creates the run by itself, named otel-quickstart, with the same events as the hooks version.

Learn More​

💬 Comments