regression in 2.55.0: pydantic-ai integration fails at startup (AttributeError: ToolManager has no attribute '_call_tool')

[jychp]

How do you use Sentry?

Sentry Saas (sentry.io)

Version

2.55.0

Steps to Reproduce

1. Install sentry-sdk==2.55.0 and pydantic-ai-slim==1.69.0
2. Initialize Sentry with PydanticAIIntegration
3. Start the application

Expected Result

Sentry initializes successfully with no errors.

Actual Result

File "/app/.venv/lib/python3.13/site-packages/sentry_sdk/integrations/pydantic_ai/patches/tools.py", line 41, in _patch_tool_execution                                                                                                                                                                    
original_call_tool = ToolManager._call_tool                                                                                                                                                                                                                                                               
AttributeError: type object 'ToolManager' has no attribute '_call_tool'

This is a regression introduced in 2.55.0. The same setup works fine with 2.54.0:

sentry-sdk	pydantic-ai-slim	Result
2.54.0	1.69.0	✅ Works
2.55.0	1.69.0	❌ Broken

[linear-code]

PY-2145

[jychp]

ref: #5561

[sentrivana]

Hey @jychp. I can't reproduce this with sentry-sdk==2.55.0 and pydantic-ai-slim==1.69.0. The patching guard is still there. Line 41 (as reported in your traceback) in the file is an empty line on 2.55.0 rather than an actual statement. Is it possible you're not running 2.55.0 after all? If you are, can you provide an example to helps us reproduce?

[sentrivana]

For the record, I tested it with this venv:

$ pip freeze
annotated-types==0.7.0
anyio==4.12.1
certifi==2026.1.4
genai-prices==0.0.54
griffelib==2.0.0
h11==0.16.0
httpcore==1.0.9
httpx==0.28.1
idna==3.11
importlib_metadata==8.7.1
logfire-api==4.25.0
opentelemetry-api==1.39.1
pydantic==2.12.5
pydantic-ai-slim==1.69.0
pydantic-graph==1.69.0
pydantic_core==2.41.5
sentry-sdk==2.55.0
typing-inspection==0.4.2
typing_extensions==4.15.0
urllib3==2.6.3
zipp==3.23.0

And this example:

import random
import sentry_sdk
from sentry_sdk.integrations.pydantic_ai import PydanticAIIntegration

from pydantic_ai import Agent
from pydantic_ai.models.test import TestModel

sentry_sdk.init(
    debug=True,
    traces_sample_rate=1,
    integrations=[PydanticAIIntegration()]
)

agent = Agent(
    TestModel(),
    instructions="You are a helpful assistant. Use the dice tool when asked to roll dice.",
)


@agent.tool_plain
def roll_dice(num_dice: int = 1, sides: int = 6) -> list[int]:
    """Roll one or more dice and return the results.

    Args:
        num_dice: Number of dice to roll (default 1).
        sides: Number of sides per die (default 6).
    """
    return [random.randint(1, sides) for _ in range(num_dice)]


if __name__ == "__main__":
    with sentry_sdk.start_transaction(name="test pydantic"):
        result = agent.run_sync("Roll 3 dice for me")
        print(result.output)