Can't update AzureAISearchTool on Thread Level

[justuswolff-audit]

Filed as an issue at the azure-sdk-for-python repo.

Describe the bug
I set an AI Search on agent level. However, in my specific thread I want to update the filter which causes thre run to fail:

{'code': 'tool_user_error', 'message': 'Error: search_too_many_indexes; Too many indexes configured for Azure AI Search tool. Only one index can be attached to either the agent, thread, or run.'}

To Reproduce

agent_toolset = ToolSet()
agent_toolset.add(
    AzureAISearchTool(
        index_name="index_name",
        index_connection_id=get_env("SEARCH_ID"),
    )
)
agent = agents_client.create_agent(toolset=agent_toolset, model="gpt-4.1")


# create a thread toolset with ai search
thread_toolset = ToolSet()
thread_toolset.add(
    AzureAISearchTool(
        index_name="index_name",
        index_connection_id=get_env("SEARCH_ID"),
        filter="SessionId eq '11111'",  # example filter
    )
)
# set thread to use the thread toolset
thread = agents_client.threads.create(tool_resources=thread_toolset.resources)

# first run: search for some file
message = agents_client.messages.create(
    thread_id=thread.id,
    role=MessageRole.USER,
    content="Please use the ai search tool to search for 'hello world'",
)
run = agents_client.runs.create_and_process(thread_id=thread.id, agent_id=agent.id, toolset=thread_toolset)

# send a follow-up message to the thread
message = agents_client.messages.create(thread_id=thread.id, role=MessageRole.USER, content="What happened?")
run = agents_client.runs.create_and_process(thread_id=thread.id, agent_id=agent.id, toolset=thread_toolset)

Expected behavior
Overwrite the agent tools by the thread level ones on thread level.

From the tool overview

Screenshots

No tool call visible (which make sense since it failed)

[manojmaximum]

This is true, I am even receiving this error, when I tried to override the agent's tool with thread tool
Error: search_too_many_indexes; Too many indexes configured for Azure AI Search tool. Only one index can be attached to either the agent, thread, or run. and incomplete details reason: None

[manojmaximum]

@justuswolff-audit Can you please confirm whether run-level configurations override agent-level configurations when using the Azure AI Foundry SDK directly?

I'm encountering an issue with the Semantic Kernel SDK (version 1.35.3), which relies on azure-ai-agents (version 1.2.0b3) under the hood. Specifically, when creating an agent with a connected Azure AI Search tool, the filter passed via the tools parameter in agent.invoke (run-level configuration) is ignored. Additionally, if the agent is created without a connected Azure AI Search tool, it fails to recognize the tools provided at the run level.

For reference, see this related GitHub thread: [Relevant thread](Azure/azure-sdk-for-python#41912).