fix: Image Pydantic schema to properly handle Union[str, Image] deserialization

[Sean-Kenneth-Doherty]

Why are these changes needed?

Fixes #7170 - UserMessage deserialization fails when content contains both string and Image data.

Problem

When UserMessage.content contains both string and Image in a list (e.g., [image, "describe this"]), JSON deserialization fails with:

Expected dict or Image instance, got <class 'str'>

Root Cause

The Image class's __get_pydantic_core_schema__ used core_schema.any_schema(), which accepts any input type. When Pydantic validates Union[str, Image], it would try the Image validator on strings before trying the str type, causing the validation to fail.

Solution

Changed the Image schema to use core_schema.union_schema with explicit types:

1. core_schema.is_instance_schema(cls) - for Image instances (pass through)
2. core_schema.dict_schema() with validator - for JSON dicts with 'data' key

This ensures the Image validator only processes dict/Image inputs, allowing strings to be handled by the str type in Union[str, Image].

Related issue number

Closes #7170

Checks

• I've included any doc changes needed for https://microsoft.github.io/autogen/. See https://github.com/microsoft/autogen/blob/main/CONTRIBUTING.md to build and test documentation locally.
• I've made sure all auto checks have passed.

Testing

Added comprehensive test file (test_image_mixed_content.py) with 5 test cases covering:

• String-only content ✅
• Image-only content ✅
• Mixed content (Image + string) - the exact bug scenario from when UserMessage have both string and Image data，JSON deserialization cause error #7170 ✅
• String before Image ✅
• Multiple strings and images interleaved ✅

All existing serialization tests continue to pass:

========================= 13 passed, 1 warning in 0.35s =========================

[Sean-Kenneth-Doherty]

@microsoft-github-policy-service agree