AWS Bedrock Claude Tool Call Error: The tool 'autogen-tools_get_weather' is not available.

[GxWwT]

import boto3
import asyncio
from botocore.config import Config

from autogen_core.models import ModelInfo, ModelFamily
from autogen_ext.models.semantic_kernel import SKChatCompletionAdapter

from semantic_kernel import Kernel
from semantic_kernel.connectors.ai.bedrock import BedrockChatCompletion, BedrockChatPromptExecutionSettings
from semantic_kernel.memory.null_memory import NullMemory

my_config = Config(
    region_name = 'us-east-1',
    signature_version = 'v4',
    retries = {
        'max_attempts': 10,
        'mode': 'standard'
    },
)

# Create the custom boto3 client
bedrock_runtime_client = boto3.client(service_name='bedrock-runtime', config=my_config)
bedrock_client = boto3.client("bedrock", config=my_config)

sk_client = BedrockChatCompletion(
    model_id='anthropic.claude-3-5-sonnet-20240620-v1:0',
    runtime_client=bedrock_runtime_client,
    client=bedrock_client,
)

# Configure execution settings
settings = BedrockChatPromptExecutionSettings(
    temperature=0.7,
    max_tokens=1000,
)

model_info = ModelInfo(vision=False, function_calling=True, json_output=True, family=ModelFamily.UNKNOWN)
model_client = SKChatCompletionAdapter(
    sk_client,
    kernel=Kernel(memory=NullMemory()),
    prompt_settings=settings,
    model_info=model_info,
)

from autogen_agentchat.agents import AssistantAgent
from autogen_agentchat.messages import TextMessage
from autogen_core import CancellationToken

async def get_weather(city: str) -> str:
    """Get the current weather for a given city"""
    return f"The weather in {city} is 73 degrees and Sunny."

async def main() -> None:
    weather_agent = AssistantAgent(
        name="assistant",
        model_client=model_client,
        tools=[get_weather],
        system_message="You are a helpful AI assistant that can provide weather information.",
        model_client_stream=True,
    )
    # print("Registered tools:", [tool.name for tool in weather_agent._tools])

    stream = weather_agent.on_messages_stream(
        [TextMessage(content="Weather in Shanghai", source="user")], CancellationToken()
    )
    async for response in stream:
        print(response)

asyncio.run(main())

Error

source='assistant' models_usage=None content="Certainly! I'd be happy to" type='ModelClientStreamingChunkEvent'
source='assistant' models_usage=None content=' provide you with the current weather' type='ModelClientStreamingChunkEvent'
source='assistant' models_usage=None content=' information for Shanghai. To get the most' type='ModelClientStreamingChunkEvent'
source='assistant' models_usage=None content=' up-to-date an' type='ModelClientStreamingChunkEvent'
source='assistant' models_usage=None content="d accurate weather data, I'll nee" type='ModelClientStreamingChunkEvent'
source='assistant' models_usage=None content='d to use the weather' type='ModelClientStreamingChunkEvent'
source='assistant' models_usage=None content=' tool. Let me fetch' type='ModelClientStreamingChunkEvent'
source='assistant' models_usage=None content=' that information for you right' type='ModelClientStreamingChunkEvent'
source='assistant' models_usage=None content=' away.' type='ModelClientStreamingChunkEvent'
source='assistant' models_usage=RequestUsage(prompt_tokens=0, completion_tokens=0) content=[FunctionCall(id='tooluse_VN13BhhhTLuwwFxmtQUdqA', arguments='{}', name='autogen-tools_get_weather')] type='ToolCallRequestEvent'
source='assistant' models_usage=None content=[FunctionExecutionResult(content="Error: The tool 'autogen-tools_get_weather' is not available.", call_id='tooluse_VN13BhhhTLuwwFxmtQUdqA')] type='ToolCallExecutionEvent'
Response(chat_message=ToolCallSummaryMessage(source='assistant', models_usage=None, content="Error: The tool 'autogen-tools_get_weather' is not available.", type='ToolCallSummaryMessage'), inner_messages=[ToolCallRequestEvent(source='assistant', models_usage=RequestUsage(prompt_tokens=0, completion_tokens=0), content=[FunctionCall(id='tooluse_VN13BhhhTLuwwFxmtQUdqA', arguments='{}', name='autogen-tools_get_weather')], type='ToolCallRequestEvent'), ToolCallExecutionEvent(source='assistant', models_usage=None, content=[FunctionExecutionResult(content="Error: The tool 'autogen-tools_get_weather' is not available.", call_id='tooluse_VN13BhhhTLuwwFxmtQUdqA')], type='ToolCallExecutionEvent')])

[ekzhu]

Created an issue #5439

[lspinheiro]

The issue is being caused by SK prepending the plugin name to the function name during the client invocation which causes a mismatch with the assistant agent registered tool name. This should fix the issue: #5444

[GxWwT]

Thanks

[GxWwT]

Keep Going:

I have removed """ plugin_name="autogen_tools" """ from """ kernel_function = KernelFunctionFromTool(tool, plugin_name="autogen_tools") """ in SKChatCompletionAdapter.
Also some other issue you committed.

However Still Error:

source='assistant' models_usage=RequestUsage(prompt_tokens=0, completion_tokens=0) content=[FunctionCall(id='tooluse_oh4KMIFxShy1WQZGPWErRg', arguments='{}', name='get-weather')] type='ToolCallRequestEvent'
source='assistant' models_usage=None content=[FunctionExecutionResult(content="Error: The tool 'get-weather' is not available.", call_id='tooluse_oh4KMIFxShy1WQZGPWErRg')] type='ToolCallExecutionEvent'
Response(chat_message=ToolCallSummaryMessage(source='assistant', models_usage=None, content="Error: The tool 'get-weather' is not available.", type='ToolCallSummaryMessage'), inner_messages=[ToolCallRequestEvent(source='assistant', models_usage=RequestUsage(prompt_tokens=0, completion_tokens=0), content=[FunctionCall(id='tooluse_oh4KMIFxShy1WQZGPWErRg', arguments='{}', name='get-weather')], type='ToolCallRequestEvent'), ToolCallExecutionEvent(source='assistant', models_usage=None, content=[FunctionExecutionResult(content="Error: The tool 'get-weather' is not available.", call_id='tooluse_oh4KMIFxShy1WQZGPWErRg')], type='ToolCallExecutionEvent')])

Another Problem:

I found arguments='{}'. For this case arguments should include required input parameter, e.g. city = Shanghai.

[GxWwT]

Give you an Example:

import boto3
from botocore.config import Config

my_config = Config(
    region_name = 'us-east-1',
    signature_version = 'v4',
    retries = {
        'max_attempts': 10,
        'mode': 'standard'
    },
)

client = boto3.client(service_name='bedrock-runtime', config=my_config)

messages = [
    {
        'role': 'user',
        'content': [
            {
                'text': "Weather in Shanghai?"
            }
        ]
    }
]

response = client.converse_stream(
    modelId='anthropic.claude-3-5-sonnet-20240620-v1:0',
    messages=messages,
    inferenceConfig={
        'maxTokens': 2000
    },
    toolConfig={
        'tools': [
            {
                'toolSpec': {
                    'name': 'get_weather',
                    'description': 'Get the current weather in a given location',
                    'inputSchema': {
                        'json': {
                            'type': 'object',
                            'properties': {
                                'location': {
                                    'type': 'string',
                                    'description': 'The city and state, e.g. San Francisco, CA',
                                }
                            },
                            'required': ['location']
                        }
                    }
                }
            }
        ]
    },
)
print(response)
stream = response.get("stream")
for chunk in stream:
    print(chunk)
    

## Response
{'ResponseMetadata': {'RequestId': 'b499a667-a88b-4197-a1bd-15431076edd3', 'HTTPStatusCode': 200, 'HTTPHeaders': {'date': 'Sat, 08 Feb 2025 03:18:46 GMT', 'content-type': 'application/vnd.amazon.eventstream', 'transfer-encoding': 'chunked', 'connection': 'keep-alive', 'x-amzn-requestid': 'b499a667-a88b-4197-a1bd-15431076edd3'}, 'RetryAttempts': 0}, 'stream': <botocore.eventstream.EventStream object at 0x000001D26DC98BF0>}
{'messageStart': {'role': 'assistant'}}
{'contentBlockDelta': {'delta': {'text': 'Certainly! I'}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': ' can'}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': ' help you get the'}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': ' current'}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': ' weather information'}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': ' for Shanghai. To'}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': ' '}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': 'do this'}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': ", I'll use"}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': ' the get'}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': '_weather function'}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': '.'}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': ' Let'}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': ' me'}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': ' fetch'}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': ' that'}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': ' information'}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': ' for'}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': ' you right'}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': ' away'}, 'contentBlockIndex': 0}}
{'contentBlockDelta': {'delta': {'text': '.'}, 'contentBlockIndex': 0}}
{'contentBlockStop': {'contentBlockIndex': 0}}
{'contentBlockStart': {'start': {'toolUse': {'toolUseId': 'tooluse_r5VChZCzSaiAkrCxPPkcWQ', 'name': 'get_weather'}}, 'contentBlockIndex': 1}}
{'contentBlockDelta': {'delta': {'toolUse': {'input': ''}}, 'contentBlockIndex': 1}}
{'contentBlockDelta': {'delta': {'toolUse': {'input': '{"locatio'}}, 'contentBlockIndex': 1}}
{'contentBlockDelta': {'delta': {'toolUse': {'input': 'n": '}}, 'contentBlockIndex': 1}}
{'contentBlockDelta': {'delta': {'toolUse': {'input': '"Shanghai, '}}, 'contentBlockIndex': 1}}
{'contentBlockDelta': {'delta': {'toolUse': {'input': 'China"}'}}, 'contentBlockIndex': 1}}
{'contentBlockStop': {'contentBlockIndex': 1}}
{'messageStop': {'stopReason': 'tool_use'}}
{'metadata': {'usage': {'inputTokens': 379, 'outputTokens': 78, 'totalTokens': 457}, 'metrics': {'latencyMs': 2453}}}

[lspinheiro]

Looking into it now. Apparently, we are not receiving the function arguments from the semantic kernel client. I still need to identify what is the cause. Will update once I have identified the problem.

[lspinheiro]

The streaming function in the adapter seems to be returning early and before the chunks with the arguments are included. It ended up being a valid but incomplete result. There should be a fix soon: #5449

Thanks for your patience in this and for helping us identify those issues.