Unknown parameter in toolConfig.tools[0]: "systemTool", must be one of: toolSpec, cachePoint

[corona10]

I am not sure this is aioboto3 issue, but if I try to add systemTool to toolConfig,
Following error is emitted.

Unknown parameter in toolConfig.tools[0]: "systemTool", must be one of: toolSpec, cachePoint

[corona10]

Environment

Type "help", "copyright", "credits" or "license" for more information.
>>> import boto3
>>> boto3.__version__
'1.40.61'
>>> import aioboto3
>>> aioboto3.__version__
'15.5.0'

aioboto3 + boto3

import asyncio
import json

import aioboto3


async def main():
    # Initialize the aioboto3 session
    session = aioboto3.Session()

    # Define the system prompt
    system_prompt = """You are an expert data analyst who can use available tools to analyze datasets and generate insights.
  Use the appropriate tools to complete the requested analysis efficiently and provide comprehensive results."""

    # Define the user prompt
    user_prompt = """Analyze the sales performance for Q4 2024 and create a summary report with visualizations.

  The sales database contains tables: orders, customers, products, and sales_reps.
  Focus on revenue trends, top-performing products, and regional performance."""

    # Prepare the request
    request_body = {
        "system": [{"text": system_prompt}],
        "messages": [{"role": "user", "content": [{"text": user_prompt}]}],
        "toolConfig": {
            "tools": [
                {"systemTool": {"name": "nova_grounding"}},
                {"systemTool": {"name": "nova_code_interpreter"}},
            ]
        },
        "additionalModelRequestFields": {
            "reasoningConfig": {"type": "enabled", "maxReasoningEffort": "low"}
        },
        "inferenceConfig": {"temperature": 0.3, "topP": 0.9, "maxTokens": 10000},
    }

    # Make the API call
    async with session.client("bedrock-runtime", region_name="us-west-2") as bedrock_runtime:
        response = await bedrock_runtime.converse(
            modelId="us.amazon.nova-2-lite-v1:0", **request_body
        )

        # Print the response
        print(json.dumps(response, indent=2, default=str))


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

It only make trouble when aioboto3 is installed. With pure installation of boto3 without aioboto3, it does not make any trouble.

Only boto3 is installed.

import boto3
import json

# Initialize the Bedrock Runtime client
bedrock_runtime = boto3.client('bedrock-runtime', region_name='us-west-2')

# Define the system prompt
system_prompt = """You are an expert data analyst who can use available tools to analyze datasets and generate insights.
  Use the appropriate tools to complete the requested analysis efficiently and provide comprehensive results."""

# Define the user prompt
user_prompt = """Analyze the sales performance for Q4 2024 and create a summary report with visualizations.
  
  The sales database contains tables: orders, customers, products, and sales_reps.
  Focus on revenue trends, top-performing products, and regional performance."""

# Prepare the request
request_body = {
    "system": [
        {
            "text": system_prompt
        }
    ],
    "messages": [
        {
            "role": "user",
            "content": [
                {
                    "text": user_prompt
                }
            ]
        }
    ],
    "toolConfig": {
        "tools": [
            {
                "systemTool": {
                    "name": "nova_grounding"
                }
            },
            {
                "systemTool": {
                    "name": "nova_code_interpreter"
                }
            }
        ]
    },
    "additionalModelRequestFields": {
        "reasoningConfig": {
            "type": "enabled",
            "maxReasoningEffort": "low"
        }
    },
    "inferenceConfig": {
        "temperature": 0.3,
        "topP": 0.9,
        "maxTokens": 10000
    }
}

# Make the API call
response = bedrock_runtime.converse(
    modelId="us.amazon.nova-2-lite-v1:0",
    **request_body
)

# Print the response
print(json.dumps(response, indent=2, default=str))

[corona10]

Upgrading boto3 to 1.42.x by force resolve the issue :(

[terricain]

Its possible that functionality is only available in a newer boto3/botocore version. I plan to bump the version of aiobotocore we bundle which will in turn bump the constraint on boto3, that being said, if its very new, we might still need a newer version of aiobotocore to be released

[jakob-keller]

Its possible that functionality is only available in a newer boto3/botocore version. I plan to bump the version of aiobotocore we bundle which will in turn bump the constraint on boto3, that being said, if its very new, we might still need a newer version of aiobotocore to be released

Recent releases of aiobotocore should provide support for system tools in Bedrock Runtime.

[corona10]

@terricain Hi Terri, when do you plan to bump up the version? if just simple bumping up is needed, I think that I can send a patch for it :)