PrefectHQ/fastmcp · error · ValueError

Unsupported content type for OpenAI: {type(item).__name__}

Error message

Unsupported content type for OpenAI: {type(item).__name__}

What it means

While converting MCP sampling messages to OpenAI format, tool-result messages may only contain text (and resource content that resolves to text); any other content type inside a tool result raises ValueError naming the Python class. This reflects OpenAI's requirement that tool messages carry plain text content.

Source

Thrown at fastmcp_slim/fastmcp/client/sampling/handlers/openai.py:249

                    elif isinstance(item, ToolResultContent):
                        # Collect tool results (added after assistant message)
                        content_text = ""
                        if item.content:
                            result_texts = [
                                sub_item.text
                                for sub_item in item.content
                                if isinstance(sub_item, TextContent)
                            ]
                            content_text = "\n".join(result_texts)
                        tool_messages.append(
                            ChatCompletionToolMessageParam(
                                role="tool",
                                tool_call_id=item.tool_use_id,
                                content=content_text,
                            )
                        )
                    else:
                        raise ValueError(
                            f"Unsupported content type for OpenAI: {type(item).__name__}"
                        )

                # Add assistant message with tool calls if present
                # OpenAI requires: assistant (with tool_calls) -> tool messages
                if tool_calls or content_parts:
                    if tool_calls:
                        has_multimodal = len(content_parts) > len(text_parts)
                        if has_multimodal:
                            raise ValueError(
                                "ImageContent/AudioContent is only supported "
                                "in user messages for OpenAI"
                            )
                        text_str = "\n".join(text_parts) or None
                        openai_messages.append(
                            ChatCompletionAssistantMessageParam(
                                role="assistant",
                                content=text_str,

View on GitHub (pinned to 1f02114297)

Solutions

1. Convert non-text tool results to text (e.g. describe or caption the image, transcribe audio) before including them.
2. Move images/audio into a subsequent user message instead of a tool message.
3. Omit unsupported content items from the tool result.
4. Use a sampling handler whose provider supports multimodal tool results.

Example fix

// before
SamplingMessage(role='tool', tool_use_id=id, content=[ImageContent(...)])
// after
SamplingMessage(role='tool', tool_use_id=id, content=[TextContent(type='text', text='Tool returned an image (see next message)')])

for m in messages:
    if m.role == 'tool':
        for c in m.content:
            if getattr(c, 'type', None) != 'text':
                raise ValueError(f"Tool result contains non-text content: {type(c).__name__}")

def is_text_only_tool_result(m) -> bool:
    return m.role != 'tool' or all(getattr(c, 'type', None) == 'text' for c in m.content)

try:
    result = await client.sample(...)
except ValueError as e:
    if 'Unsupported content type for OpenAI' in str(e):
        messages = stringify_tool_results(messages)
        result = await client.sample(messages=messages, ...)
    else:
        raise

When it happens

Trigger: A sampling message with role='tool' whose content list contains ImageContent, AudioContent, EmbeddedResource with non-text data, etc., passed through _convert_to_openai_messages during Client.sample().

Common situations: Tool results that return screenshots or audio clips fed back as tool messages; embedding binary resources in tool output; MCP servers returning rich tool results that OpenAI's tool-message schema can't represent.

Related errors

• ImageContent/AudioContent is only supported in user messages
• Unsupported content type: {type(content)}
• Invalid JSON in tool arguments for '{func.name}': {func.argu
• No content in response from completion (finish_reason={finis
• Unsupported image MIME type for OpenAI: {content.mime_type!r

AI-assisted analysis of PrefectHQ/fastmcp@1f02114297 (2026-08-29). Data as JSON: /api/errors/e59f642d3603eac0. Report an issue: GitHub.