[BUG] AgentResult.metrics.tool_metrics skips previous tool inputs for repeatedly invoked tools

[RingoIngo2]

Checks

• I have updated to the lastest minor and patch version of Strands
• I have checked the documentation and this is not expected behavior
• I have searched ./issues and there are no duplicates of my issue

Strands Version

Commit 6a1ccea

Python Version

3.13.1

Operating System

Amazon linux 2

Installation Method

git clone

Steps to Reproduce

Run this script

from strands import Agent, tool
from strands.models import BedrockModel
import random

@tool
def weather_forecast(city: str, days: int = 3) -> str:
    """Get weather forecast for a city.

    Args:
        city: The name of the city
        days: Number of days for the forecast
    """
    weather = random.choice(["sunny", "rainy", "cloudy"])
    return f"Weather forecast for {city} for the next {days} days is {weather}."


def create_bedrock_model():
    bedrock_model = BedrockModel(
        boto_session=boto3.Session(region_name="us-west-2"),
        model_id="us.anthropic.claude-3-7-sonnet-20250219-v1:0",
        cache_prompt="default",
        cache_tools="default",
        temperature=0.0,
    )
    return bedrock_model


def convert_tool_metrics_to_dict(tool_metrics) -> dict:
    from dataclasses import asdict

    output = {}
    for key, value in tool_metrics.items():
        output[key] = asdict(value)
    return output


agent = Agent(
    model=create_bedrock_model(),
    system_prompt="You are a helpful AI assistant",
    tools=[weather_forecast],
)

response = agent(
    "Whats the weather in Berlin tomorrow? Use a tool. Afterwards check the weather in Paris for tomorrow. Then check the weather in Rome."
)
print(convert_tool_metrics_to_dict(response.metrics.tool_metrics))

Check the printed tool metrics. They look like

{
    'weather_forecast': {
        'tool': {
            'toolUseId': 'tooluse_nQBPOmkJTsOY3hXIbjTr1A',
            'name': 'weather_forecast',
            'input': {
                'city': 'Rome'
            }
        },
        'call_count': 3,
        'success_count': 3,
        'error_count': 0,
        'total_time': 0.0002276897430419922
    }
}

Expected Behavior

The tool metrics should either not mention the tool inputs ('city': 'Rome') or they should list all inputs from the three tool invocations, i.e., [('city': 'Berlin'), ('city': 'Paris'), ('city': 'Rome')].

Actual Behavior

The tool metrics show three invocations of the tool. However, only one input (the latest one) is included. Here, this is 'city': 'Rome'. However, the tool was also invoked for two other cities.

{
    'weather_forecast': {
        'tool': {
            'toolUseId': 'tooluse_nQBPOmkJTsOY3hXIbjTr1A',
            'name': 'weather_forecast',
            'input': {
                'city': 'Rome'
            }
        },
        'call_count': 3,
        'success_count': 3,
        'error_count': 0,
        'total_time': 0.0002276897430419922
    }
}

Additional Context

No response

Possible Solution

No response

Related Issues

No response