agentscope-ai/agentscope · error · ValueError

Unsupported image source type: {type(source)}

Error message

Unsupported image source type: {type(source)}

What it means

The Moonshot formatter only accepts URLSource (remote URLs are converted to base64 data URLs by downloading) and Base64Source for images; any other source type (e.g. LocalFileSource, raw bytes) hits this ValueError in _moonshot_format_image_source.

Source

Thrown at src/agentscope/formatter/_moonshot_formatter.py:51

    """
    if isinstance(source, Base64Source):
        url = f"data:{source.media_type};base64,{source.data}"

    elif isinstance(source, URLSource):
        url_str = str(source.url)
        if url_str.startswith("file://"):
            local_path = url_str.removeprefix("file://")
            with open(local_path, "rb") as f:
                encoded = base64.b64encode(f.read()).decode("utf-8")
            url = f"data:{source.media_type};base64,{encoded}"
        else:
            response = requests.get(url_str, timeout=30)
            response.raise_for_status()
            encoded = base64.b64encode(response.content).decode("utf-8")
            url = f"data:{source.media_type};base64,{encoded}"

    else:
        raise ValueError(f"Unsupported image source type: {type(source)}")

    return {
        "type": "image_url",
        "image_url": {"url": url},
    }


class MoonshotChatFormatter(_OpenAIFormatterBase):
    """The Moonshot AI formatter for chatbot scenario.

    Moonshot's API is OpenAI-compatible, but thinking models (``kimi-k2.6``,
    ``kimi-k2-thinking``) return a ``reasoning_content`` field alongside
    ``content`` in assistant messages.  This formatter preserves that field
    when re-sending assistant messages back to the API so that the
    *Preserved Thinking* feature works correctly in multi-turn conversations.
    """

    input_types: list[str] = Field(

View on GitHub (pinned to e90f1c7592)

Solutions

1. Convert the local image to a Base64Source before formatting
2. Host the image and use URLSource with the public URL
3. Pre-check source types and provide a text fallback for unsupported blocks

Example fix

# before
ImageBlock(source=LocalFileSource(path="cat.jpg"), alt="cat")

# after
import base64
ImageBlock(source=Base64Source(data=base64.b64encode(open("cat.jpg","rb").read()).decode(), media_type="image/jpeg"), alt="cat")

from agentscope.message import URLSource, Base64Source
assert isinstance(block.source, (URLSource, Base64Source))

def is_moonshot_image_source(src) -> bool:
    return isinstance(src, (URLSource, Base64Source))

try:
    formatter.format(msgs)
except ValueError as e:
    if "Unsupported image source type" in str(e):
        block.source = local_path_to_base64_source(path)  # then retry
    else:
        raise

When it happens

Trigger: Passing an ImageBlock with a LocalFileSource or non-URL/base64 source to the Moonshot formatter; using a source class the Moonshot path doesn't isinstance-match.

Common situations: Local image paths fed into Moonshot pipelines; sharing message-building code across providers where some accept local files; stale source objects after library upgrades that renamed source classes.

Related errors

• Unsupported source type: {type(source)}
• Unsupported image source type: {type(source)}
• Unsupported audio source type: {type(source)}
• Unsupported source type: {type(source)}
• Unsupported audio source type: {type(source)}.

AI-assisted analysis of agentscope-ai/agentscope@e90f1c7592 (2026-08-28). Data as JSON: /api/errors/35e975a0cb854a78. Report an issue: GitHub.