mlflow/mlflow · error · MlflowException

Unknown content type: {content_type['type']}. Please make su

Error message

Unknown content type: {content_type['type']}. Please make sure the message is a valid Anthropic message object. If it is a valid type, contact to the MLflow maintainer via https://github.com/mlflow/mlflow/issues/new/choose for requesting support for a new message type.

What it means

Within a list content, each block's 'type' must be one MLflow knows: text, image, tool_use, tool_result, or thinking (Claude 3.7 extended thinking, mapped to text). An unrecognized block type cannot be converted to a ContentPart, so MLflow raises this error and asks you to file an issue if the type is genuinely valid Anthropic content.

Source

Thrown at mlflow/anthropic/chat.py:110

    content_type = content.get("type")
    if content_type == "text":
        return TextContentPart(text=content["text"], type="text")
    elif content_type == "image":
        source = content["source"]
        return ImageContentPart(
            image_url=ImageUrl(
                url=f"data:{source['media_type']};{source['type']},{source['data']}"
            ),
            type="image_url",
        )
    # Claude 3.7 added new "thinking" content block, which is essentially a text block as of now.
    # TODO: We should consider adding a new ContentPart type if more providers support this.
    # https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking
    elif content_type == "thinking":
        return TextContentPart(text=content["thinking"], type="text")
    else:
        raise MlflowException.invalid_parameter_value(
            f"Unknown content type: {content_type['type']}. Please make sure the message "
            "is a valid Anthropic message object. If it is a valid type, contact to the "
            "MLflow maintainer via https://github.com/mlflow/mlflow/issues/new/choose for "
            "requesting support for a new message type."
        )


def convert_tool_to_mlflow_chat_tool(tool: dict[str, Any]) -> ChatTool:
    """
    Convert Anthropic tool definition into MLflow's standard format (OpenAI compatible).

    Ref: https://docs.anthropic.com/en/docs/build-with-claude/tool-use

    Args:
        tool: A dictionary represents a single tool definition in the input request.

    Returns:
        ChatTool: MLflow's standard tool definition object.

View on GitHub (pinned to 6a27f2decc)

Solutions

Upgrade MLflow to the latest version, which may support the new block type
Strip or pre-convert unsupported blocks to {'type': 'text', 'text': ...} before logging
If the type is valid Anthropic content, file an issue at https://github.com/mlflow/mlflow/issues/new/choose
Verify each block dict has a correctly spelled 'type' key

Example fix

// before
blocks = [{"type": "document", "source": {...}}]
// after
blocks = [{"type": "text", "text": extract_text_from_document()}]  # or upgrade mlflow

Defensive patterns

Strategy: try-catch

Validate before calling

KNOWN = {"text", "image", "tool_use", "tool_result", "thinking"}
unsupported = [b["type"] for m in messages if isinstance(m.get("content"), list)
               for b in m["content"] if isinstance(b, dict) and b.get("type") not in KNOWN]
assert not unsupported, f"Unsupported block types: {unsupported}"

Type guard

def is_supported_block(b: dict) -> bool:
    return isinstance(b, dict) and b.get("type") in {"text", "image", "tool_use", "tool_result", "thinking"}

Try / catch

try:
    trace = model_to_chat(response)
except MlflowException as e:
    if "Unknown content type" in str(e):
        response.content = [b for b in response.content if getattr(b, "type", None) in {"text", "tool_use", "thinking"}]
        trace = model_to_chat(response)
    else:
        raise

Prevention

Keep MLflow updated when adopting new Anthropic features (documents, web search)
Filter or flatten unsupported blocks before logging traces
Check mlflow release notes for new Anthropic content-type support

When it happens

Trigger: A content block with 'type' like 'document', 'server_tool_use', 'web_search_tool_result', 'redacted_thinking', or any new Anthropic block type not yet handled by your MLflow version; passing through _parse_content with a dict lacking/misspelling 'type' (content_type becomes None).

Common situations: Using new Anthropic API features (documents/PDFs, web search, code execution blocks) with an older MLflow; a typo in the 'type' field; 'type' key missing entirely so None is reported.

Related errors

AI-assisted analysis of mlflow/mlflow@6a27f2decc (2026-08-29). Data as JSON: /api/errors/d36efb0dd55a759c. Report an issue: GitHub.