### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### The Feature

Support provider-specific uploaded image references (ms://file-id) for Moonshot/Kimi Vision when using OpenAI-compatible image inputs through LiteLLM Proxy

### Motivation, pitch

We are using DSPy -> LiteLLM Proxy -> Moonshot/Kimi Vision for a multimodal endpoint.

Kimi Vision does not support arbitrary remote HTTP image URLs in image_url.url. According to Moonshot/Kimi docs, images must be sent either as base64 data URLs or as uploaded file references like ms://<file-id>.

In our production path, DSPy formats dspy.Image(url=...) into OpenAI-compatible content:

{ "type": "image_url", "image_url": { "url": "https://picaso-public.oss-cn-beijing.aliyuncs.com/..." } }

LiteLLM Proxy routes this to a Moonshot model group:

model_group: aesthetic_plan_conclusion litellm_params: model: moonshot/...

The request fails with:

litellm.exceptions.BadRequestError: OpenAIException - litellm.BadRequestError: MoonshotException - Invalid request: unsupported image url: https://picaso-public.oss-cn-beijing.aliyuncs…