microsoft/semantic-kernel · error · ServiceInvalidExecutionSettingsError

Auto-invocation of tool calls may only be used with a OpenAI

Error message

Auto-invocation of tool calls may only be used with a OpenAIChatPromptExecutions.number_of_responses of 1.

What it means

Raised in _verify_function_choice_settings when auto-invocation of tool calls is enabled (function_choice configuration is set) but number_of_responses is greater than 1. The OpenAI API cannot reliably correlate tool-call requests/responses across multiple parallel completions, so Semantic Kernel blocks this combination at the configuration validation stage.

Source

Thrown at python/semantic_kernel/connectors/ai/open_ai/services/open_ai_chat_completion_base.py:145

                        inner_content=chunk,
                        ai_model_id=settings.ai_model_id,
                        metadata=chunk_metadata,
                        function_invoke_attempt=function_invoke_attempt,
                    )
                    for i in range(settings.number_of_responses or 1)
                ]
            else:
                yield [
                    self._create_streaming_chat_message_content(chunk, choice, chunk_metadata, function_invoke_attempt)
                    for choice in chunk.choices
                ]

    @override
    def _verify_function_choice_settings(self, settings: "PromptExecutionSettings") -> None:
        if not isinstance(settings, OpenAIChatPromptExecutionSettings):
            raise ServiceInvalidExecutionSettingsError("The settings must be an OpenAIChatPromptExecutionSettings.")
        if settings.number_of_responses is not None and settings.number_of_responses > 1:
            raise ServiceInvalidExecutionSettingsError(
                "Auto-invocation of tool calls may only be used with a "
                "OpenAIChatPromptExecutions.number_of_responses of 1."
            )

    @override
    def _update_function_choice_settings_callback(
        self,
    ) -> Callable[["FunctionCallChoiceConfiguration", "PromptExecutionSettings", FunctionChoiceType], None]:
        return update_settings_from_function_call_configuration

    @override
    def _reset_function_choice_settings(self, settings: "PromptExecutionSettings") -> None:
        if hasattr(settings, "tool_choice"):
            settings.tool_choice = None
        if hasattr(settings, "tools"):
            settings.tools = None

    # endregion

View on GitHub (pinned to c028a0c7dc)

Solutions

Set settings.number_of_responses = 1 (or leave it unset, which defaults to 1) when using auto-function-invocation
If you need multiple responses, disable auto-invocation and handle tool calls manually in a loop
Audit your settings object right before the call: if function calling is configured, force number_of_responses to 1

Example fix

# before
settings = OpenAIChatPromptExecutionSettings(number_of_responses=3)
# auto-invocation configured elsewhere -> raises at verify time
# after
settings = OpenAIChatPromptExecutionSettings()  # number_of_responses defaults to 1
# or explicitly
settings.number_of_responses = 1

Defensive patterns

Strategy: validation

Validate before calling

from semantic_kernel.connectors.ai.open_ai import OpenAIChatPromptExecutionSettings

if isinstance(settings, OpenAIChatPromptExecutionSettings):
    if settings.number_of_responses is not None and settings.number_of_responses > 1:
        if hasattr(settings, 'function_choice_behavior') and settings.function_choice_behavior:
            settings.number_of_responses = 1  # reconcile conflict

Try / catch

from semantic_kernel.exceptions.service_exceptions import ServiceInvalidExecutionSettingsError

try:
    await kernel.invoke(...)
except ServiceInvalidExecutionSettingsError as e:
    if 'number_of_responses of 1' in str(e):
        settings.number_of_responses = 1
        await kernel.invoke(...)  # retry

Prevention

Treat number_of_responses > 1 and function-calling as mutually exclusive by design
Add a settings-validation helper that checks this invariant before the kernel call

When it happens

Trigger: Setting OpenAIChatPromptExecutionSettings.number_of_responses (OpenAI's n parameter) to a value > 1 while also configuring auto-function-invocation (e.g., FunctionChoiceBehavior.Auto with kernel/plugins registered).

Common situations: A developer sets n > 1 to get multiple candidate responses for selection, then later adds function calling to the same settings without resetting n; copying settings from a template that defaults number_of_responses to a higher value.

Related errors

AI-assisted analysis of microsoft/semantic-kernel@c028a0c7dc (2026-08-13). Data as JSON: /api/errors/97bb5e33b2465874. Report an issue: GitHub.