feat(llmobs): [MLOB-2989] add pydantic_ai instrumentation

[Kyle-Verhoog]

This PR implements APM tracing for the PydanticAI library. Currently, the following methods are traced:

• agent.Agent.iter: captures all ways of running the agent, including agent.run, agent.run_sync, agent.run_stream, and agent.iter
• tools.Tool.run: captures tool invocations

The only llmobs related tags set on these spans are pydantic_ai.request.model and pydantic_ai.request.provider (for Pydantic Agent spans only).

Below is an example script using PydanticAI for a simple math problem. Running this script with DD_SERVICE="nicole-test" DD_ENV="nicole-test" ddtrace-run python3 agent_iter_example.py produces the following APM trace with a span representing the agent invocation, a span representing the tool invocation, and two Open AI LLM calls that occurred during the execution.

import asyncio
from pydantic_ai import Agent

def calculate_square(x: int) -> int:
    return x * x

agent = Agent(  
    'openai:gpt-4o',
    system_prompt='Use one of the tools provided to calculate the mathematical operation.',  
    tools=[calculate_square],
)

async def main():
    async with agent.iter('What is the square of 5?') as response:
        async for _ in response:
            pass

if __name__ == '__main__':
    asyncio.run(main())

Checklist

• PR author has checked that all the criteria below are met
• The PR description includes an overview of the change
• The PR description articulates the motivation for the change
• The change includes tests OR the PR description describes a testing strategy
• The PR description notes risks associated with the change, if any
• Newly-added code is easy to change
• The change follows the library release note guidelines
• The change includes or references documentation updates if necessary
• Backport labels are set (if applicable)

Reviewer Checklist

• Reviewer has checked that all the criteria below are met
• Title is accurate
• All changes are related to the pull request's stated goal
• Avoids breaking API changes
• Testing strategy adequately addresses listed risks
• Newly-added code is easy to change
• Release note makes sense to a user of the library
• If necessary, author has acknowledged and discussed the performance implications of this PR as reported in the benchmarks PR comment
• Backport labels are set in a manner that is consistent with the release branch maintenance policy

[github-actions]

CODEOWNERS have been resolved as:

.riot/requirements/111353f.txt                                          @DataDog/apm-python
.riot/requirements/14191ba.txt                                          @DataDog/apm-python
.riot/requirements/16cf7dd.txt                                          @DataDog/apm-python
.riot/requirements/1721127.txt                                          @DataDog/apm-python
.riot/requirements/187928c.txt                                          @DataDog/apm-python
.riot/requirements/1a98a60.txt                                          @DataDog/apm-python
.riot/requirements/2c6ccbc.txt                                          @DataDog/apm-python
.riot/requirements/3459dc4.txt                                          @DataDog/apm-python
.riot/requirements/d7a340e.txt                                          @DataDog/apm-python
.riot/requirements/f346262.txt                                          @DataDog/apm-python
ddtrace/contrib/_pydantic_ai.py                                         @DataDog/ml-observability
ddtrace/contrib/internal/pydantic_ai/patch.py                           @DataDog/ml-observability
ddtrace/contrib/internal/pydantic_ai/utils.py                           @DataDog/ml-observability
ddtrace/llmobs/_integrations/pydantic_ai.py                             @DataDog/ml-observability
releasenotes/notes/pydantic-ai-apm-integration-279845b06ab611dc.yaml    @DataDog/apm-python
tests/contrib/pydantic_ai/__init__.py                                   @DataDog/ml-observability
tests/contrib/pydantic_ai/cassettes/agent_iter.yaml                     @DataDog/ml-observability
tests/contrib/pydantic_ai/cassettes/agent_run_stream.yaml               @DataDog/ml-observability
tests/contrib/pydantic_ai/cassettes/agent_with_tools.yaml               @DataDog/ml-observability
tests/contrib/pydantic_ai/conftest.py                                   @DataDog/ml-observability
tests/contrib/pydantic_ai/test_pydantic_ai.py                           @DataDog/ml-observability
tests/contrib/pydantic_ai/test_pydantic_ai_patch.py                     @DataDog/ml-observability
tests/snapshots/tests.contrib.pydantic_ai.test_pydantic_ai.test_agent_run.json  @DataDog/apm-python
tests/snapshots/tests.contrib.pydantic_ai.test_pydantic_ai.test_agent_run_error.json  @DataDog/apm-python
tests/snapshots/tests.contrib.pydantic_ai.test_pydantic_ai.test_agent_with_tools.json  @DataDog/apm-python
.github/CODEOWNERS                                                      @DataDog/python-guild @DataDog/apm-core-python
ddtrace/_monkey.py                                                      @DataDog/apm-core-python
ddtrace/contrib/integration_registry/registry.yaml                      @DataDog/apm-core-python @DataDog/apm-idm-python
ddtrace/llmobs/_llmobs.py                                               @DataDog/ml-observability
ddtrace/settings/_config.py                                             @DataDog/apm-core-python
riotfile.py                                                             @DataDog/apm-python
supported_versions_output.json                                          @DataDog/apm-core-python
supported_versions_table.csv                                            @DataDog/apm-core-python
tests/contrib/openai/test_openai_v1.py                                  @DataDog/ml-observability
tests/llmobs/suitespec.yml                                              @DataDog/ml-observability

[github-actions]

Bootstrap import analysis

Comparison of import times between this PR and base.

Summary

The average import time from this PR is: 274 ± 2 ms.

The average import time from base is: 276 ± 2 ms.

The import time difference between this PR and base is: -1.16 ± 0.09 ms.

Import time breakdown

The following import paths have shrunk:

ddtrace.auto 1.923 ms (0.70%)

ddtrace.bootstrap.sitecustomize 1.247 ms (0.45%)

ddtrace.bootstrap.preload 1.247 ms (0.45%)

ddtrace.internal.remoteconfig.client 0.642 ms (0.23%)

ddtrace 0.676 ms (0.25%)

ddtrace.internal._unpatched 0.031 ms (0.01%)

json 0.031 ms (0.01%)

json.decoder 0.031 ms (0.01%)

re 0.031 ms (0.01%)

enum 0.031 ms (0.01%)

types 0.031 ms (0.01%)

[pr-commenter]

Benchmarks

Benchmark execution time: 2025-06-30 16:18:14

Comparing candidate commit 5120620 in PR branch kylev/pydantic with baseline commit 4b87ee8 in branch main.

Found 0 performance improvements and 0 performance regressions! Performance is the same for 569 metrics, 3 unstable metrics.

[Kyle-Verhoog]

just one thought about the span type and kind but otherwise so happy with how clean this was! nice work 😄

[brettlangdon]

left a thought, otherwise the files owned by core lgtm