microsoft/autogen · error · ValueError

Image content cannot be converted to text

Error message

Image content cannot be converted to text

What it means

Teachability's text extraction cannot convert binary/image memory items to text. When a MemoryContent contains an Image object, the converter raises because advice extraction and retrieval operate on text only.

Source

Thrown at python/packages/autogen-ext/src/autogen_ext/experimental/task_centric_memory/utils/teachability.py:51

        return self._name

    def _extract_text(self, content_item: str | MemoryContent) -> str:
        """Extract searchable text from content."""
        if isinstance(content_item, str):
            return content_item

        content = content_item.content
        mime_type = content_item.mime_type

        if mime_type in [MemoryMimeType.TEXT, MemoryMimeType.MARKDOWN]:
            return str(content)
        elif mime_type == MemoryMimeType.JSON:
            if isinstance(content, dict):
                # Store original JSON string representation
                return str(content).lower()
            raise ValueError("JSON content must be a dict")
        elif isinstance(content, Image):
            raise ValueError("Image content cannot be converted to text")
        else:
            raise ValueError(f"Unsupported content type: {mime_type}")

    async def update_context(
        self,
        model_context: ChatCompletionContext,
    ) -> UpdateContextResult:
        """
        Extracts any advice from the last user turn to be stored in memory,
        and adds any relevant memories to the model context.
        """
        self._logger.enter_function()

        # Extract text from the user's last message
        messages = await model_context.get_messages()
        if not messages:
            self._logger.leave_function()
            return UpdateContextResult(memories=MemoryQueryResult(results=[]))

View on GitHub (pinned to 027ecf0a37)

Solutions

1. Filter out image contents before adding them to the teachability memory (check isinstance(content, Image)).
2. Store a textual description/caption of the image instead of the Image object.
3. Use a memory implementation that supports multimodal content for images.

Example fix

# before
await memory.add(MemoryContent(content=Image.from_uri(img_uri), mime_type=MemoryMimeType.IMAGE))
# after
await memory.add(MemoryContent(content=describe_image(img_uri), mime_type=MemoryMimeType.TEXT))

from autogen_core.models import Image
if isinstance(content, Image):
    raise SkipContent("image not supported by teachability")  # or convert to a caption
c = MemoryContent(content=content, mime_type=MemoryMimeType.TEXT)

from autogen_core.models import Image

def is_text_safe_memory_content(content) -> bool:
    return not isinstance(content, Image)

When it happens

Trigger: Passing MemoryContent(content=Image.from_uri(...)) (or any Image instance) into the teachability memory's update_context / query path, which routes through the text converter.

Common situations: Multimodal agents sharing a memory store with image artifacts; accidentally storing screenshot/image content in a text-only teachable memory.

Related errors

• Invalid aggregate message {reason}
• Unsupported content type {item.GetType()}
• Only TextContent and ImageContent are allowed in MultiModalM
• JSON content must be a dict
• Unsupported content type: {mime_type}

AI-assisted analysis of microsoft/autogen@027ecf0a37 (2026-08-15). Data as JSON: /api/errors/da79489ae50fd06a. Report an issue: GitHub.