## What I'd like

Please add first-class support for **Ollama text-to-image generation** in `litellm.image_generation(...)`.

Today, calling `litellm.image_generation(model=\"ollama/...\")` appears to route to the Ollama provider but returns an empty payload (`ImageResponse.data=[]`) instead of generated image data.

## Why this matters

Ollama now exposes image-generation models (for example FLUX variants), and it would be very useful if LiteLLM could normalize this path the same way it does for other image providers.

Relevant Ollama URLs: - Model page: https://ollama.com/x/flux2-klein - API docs (`/api/generate`): https://docs.ollama.com/api/generate - Related Ollama issue discussing API text-to-image support: https://github.com/ollama/ollama/issues/16061

## Minimal repro

```python import litellm

response = litellm.image_generation( model="ollama/x/flux2-klein:4b-bf16", prompt="A red apple on a wooden table", n=1, size="1024x1024", api_base="http://localhost:11434", timeout=30, )

print(type(response).__name__) print(len(response.data)) print(response) ```

## Observed behavior

- No exception is raised. - `type(response)` is `ImageResponse`. - `respon…