Support multimodal tasks in AGS

[victordibia]

AgentChat supports MultiModalMessages, AGS should.
This will enable tasks such as

• Analyze this image and write a report .. which is then addressed by a team

How

• Update UI to enable addition of images
• On submit, convert UserMessge(content=[str, Image])
• ...

[rajantripathi]

I'd like to help implement this feature. My use case is a team of agents analyzing medical images (pathology slides, radiology) with multilingual clinical queries.

What I can contribute:

1. Image Upload UI: Drag-and-drop for DICOM/PNG/JPEG with preview
2. Message Conversion: UserMessage(content=[str, Image]) with proper base64 encoding
3. Multilingual Support: Handle image queries in non-English languages (Chinese, Arabic, Uzbek medical terminology)

Architecture suggestion:

from autogen_agentchat.messages import MultiModalMessage, Image

async def handle_image_query(image_path: str, query: str, language: str = "en"):
    image = Image.from_file(image_path)
    message = MultiModalMessage(
        content=[query, image],
        language=language  # For prompt adaptation
    )
    return await agent_team.run(message)

Cross-reference: Similar multimodal message handling in RAGAS (vibrantlabsai/ragas#2651) where we added language adaptation for evaluation prompts.

Happy to contribute a PR for the image handling layer once the UI design is finalized.