azd ai agent invoke : web_search tool throwing an error

[banrahan]

Service traces for request: https://dataexplorer.azure.com/dashboards/0b69c942-b3cd-4303-a087-96902864f967?p-_startTime=2days&p-_endTime=now&p-p_apim_request_id=v-cee89311-a325-47bf-b336-008837622037&p-p_take=v-1000#47505622-afa0-403f-a39e-d5c67eade5ea

Steps to repro

Make sure you

• azd ai agent init https://github.com/microsoft-foundry/foundry-samples/blob/main/samples/python/hosted-agents/agent-framework/agent-with-foundry-tools/agent.yaml
• Update the version of the azure-ai-agentserver-agentframework to 1.0.0.b15
• azd env set enableHostedAgentVNext=true
• Set an environment variable AZURE_OPENAI_CHAT_DEPLOYMENT_NAME, it wasn't picked up from the YAML
• use the below yaml and main.py

here is the error message

azd ai agent invoke "search the web for weather in roanoke, va"
Agent:        af-agent-with-foundry-tools (remote)
Message:      "search the web for weather in roanoke, va"
Session:      7aabdff5-5b92-4f91-9c4c-b00010c4b897
Conversation: conv_a3b64d25f700f00200hMjSNHCcLlScCP0YPoHlB1SLWj2UfT55

Trace ID: cee89311-a325-47bf-b336-008837622037

ERROR: agent error (server_error): <class 'agent_framework.azure._chat_client.AzureOpenAIChatClient'> service failed to complete the prompt: Error code: 400 - {'error': {'message': "Invalid 'messages[6].tool_calls[0].id': string too long. Expected a string with maximum length 40, but got a string with length 55 instead.", 'type': 'invalid_request_error', 'param': 'messages[6].tool_calls[0].id', 'code': 'string_above_max_length'}}

here is the main.py

import os
from dotenv import load_dotenv
from agent_framework.azure import AzureOpenAIChatClient

from azure.ai.agentserver.agentframework import from_agent_framework, FoundryToolsChatMiddleware
from azure.identity import DefaultAzureCredential, get_bearer_token_provider

# Load environment variables from .env file for local development
# load_dotenv()

# Create a token provider that refreshes tokens automatically for long-running servers
# This avoids 401 errors when the initial token expires (typically after 1 hour)
_credential = DefaultAzureCredential()
_token_provider = get_bearer_token_provider(_credential, "https://cognitiveservices.azure.com/.default")


def main():
    required_env_vars = [
        "AZURE_OPENAI_ENDPOINT",
        "AZURE_OPENAI_CHAT_DEPLOYMENT_NAME",
        "AZURE_AI_PROJECT_ENDPOINT",
    ]
    for env_var in required_env_vars:
        assert env_var in os.environ and os.environ[env_var], (
            f"{env_var} environment variable must be set."
        )

    tools=[{"type": "web_search"}]
    if project_tool_connection_id := os.environ.get("AZURE_AI_PROJECT_TOOL_CONNECTION_ID"):
        tools.append({"type": "mcp", "project_connection_id": project_tool_connection_id})

    # Use token provider for automatic token refresh in long-running servers
    chat_client = AzureOpenAIChatClient(ad_token_provider=_token_provider,
                                        middleware=FoundryToolsChatMiddleware(tools))
    agent = chat_client.create_agent(
        name="FoundryToolAgent",
        instructions="You are a helpful assistant with access to various tools."
    )

    from_agent_framework(agent).run()

if __name__ == "__main__":
    main()

Here is the agent.yaml

kind: hosted
name: af-agent-with-foundry-tools
description: |
    An AI agent that uses Azure OpenAI with a Hosted Model Context Protocol (MCP) server. The agent answers questions by searching Microsoft Learn documentation using MCP tools.
metadata:
    authors:
        - Microsoft
    tags:
        - Azure AI AgentServer
        - Microsoft Agent Framework
        - Model Context Protocol
        - MCP
protocols:
    - protocol: responses
      version: ""
environment_variables:
    - name: AZURE_OPENAI_ENDPOINT
      value: ${AZURE_OPENAI_ENDPOINT}
    - name: AZURE_OPENAI_CHAT_DEPLOYMENT_NAME
      value: gpt-5
    - name: AZURE_AI_PROJECT_ENDPOINT
      value: ${AZURE_AI_PROJECT_ENDPOINT}

[github-actions]

Thanks for the feedback! We are routing this to the appropriate team for follow-up. cc @JC-386 @lusu-msft.