microsoft/semantic-kernel · error · ValueError

No FunctionResultContent found in the message items

Error message

No FunctionResultContent found in the message items

What it means

Raised in _format_tool_message when formatting a TOOL-role ChatMessageContent whose first item is not a FunctionResultContent. The Azure AI Inference connector expects each tool message to carry exactly one tool result (it warns if there are more/fewer items, and hard-fails if the first item is not the expected result type), because it must extract .result and .id to build the API's tool message payload.

Source

Thrown at python/semantic_kernel/connectors/ai/azure_ai_inference/services/utils.py:144

def _format_tool_message(message: ChatMessageContent) -> ToolMessage:
    """Format a tool message to the expected object for the client.

    Args:
        message: The tool message.

    Returns:
        The formatted tool message.
    """
    if len(message.items) != 1:
        logger.warning(
            "Unsupported number of items in Tool message while formatting chat history for Azure AI"
            f" Inference: {len(message.items)}"
        )

    if not isinstance(message.items[0], FunctionResultContent):
        raise ValueError("No FunctionResultContent found in the message items")

    # The API expects the result to be a string, so we need to convert it to a string
    return ToolMessage(
        content=str(message.items[0].result), tool_call_id=message.items[0].id if message.items[0].id else "None"
    )


MESSAGE_CONVERTERS: dict[AuthorRole, Callable[[ChatMessageContent], ChatRequestMessage]] = {
    AuthorRole.SYSTEM: _format_system_message,
    AuthorRole.USER: _format_user_message,
    AuthorRole.ASSISTANT: _format_assistant_message,
    AuthorRole.TOOL: _format_tool_message,
    AuthorRole.DEVELOPER: _format_developer_message,
}

View on GitHub (pinned to c028a0c7dc)

Solutions

1. Ensure each TOOL message contains exactly one FunctionResultContent with .result (and ideally .id matching the tool_call_id).
2. If you have multiple results, split into separate TOOL messages or ensure only FunctionResultContent items are present.
3. Use SK's function-invocation middleware to produce tool messages rather than constructing them manually.

Example fix

# before
history.add_message(ChatMessageContent(
    role=AuthorRole.TOOL,
    items=[TextContent(text="42")],  # wrong item type → error
))

# after
from semantic_kernel.contents import FunctionResultContent
history.add_message(ChatMessageContent(
    role=AuthorRole.TOOL,
    items=[FunctionResultContent(id=call_id, name="calc", result="42")],
))

from semantic_kernel.contents import FunctionResultContent

def tool_messages_have_result(history):
    for m in history:
        if m.role.name == "TOOL":
            assert m.items and isinstance(m.items[0], FunctionResultContent), \
                "Each TOOL message must start with a FunctionResultContent"

# call before sending history to the service

from semantic_kernel.contents import FunctionResultContent

def is_valid_tool_message(message) -> bool:
    return (
        message.role.name == "TOOL"
        and len(message.items) == 1
        and isinstance(message.items[0], FunctionResultContent)
    )

try:
    await service.get_chat_message_contents(history=history, settings=settings)
except ValueError as e:
    if "FunctionResultContent" in str(e):
        # rebuild the offending tool message with a FunctionResultContent item
        ...

When it happens

Trigger: A TOOL message whose items list is empty or whose first item is a different content type (e.g. a stray TextContent, FunctionCallContent, or ImageContent). Often a result of manually constructing a tool message without wrapping the result in FunctionResultContent.

Common situations: Building tool messages by hand instead of through the function-invocation pipeline; mixing content types into a tool message; bugs in middleware that replace items on tool messages.

Related errors

• Tool message found without a preceding message.
• Tool message found after a user or system message.
• Auto invocation of tool calls may only be used with a single
• Unsupported role in chat history: {curr_message.role}
• If response_format has type 'json_schema', 'json_schema' mus

AI-assisted analysis of microsoft/semantic-kernel@c028a0c7dc (2026-08-13). Data as JSON: /api/errors/aa23f0d4aef248f8. Report an issue: GitHub.