Failed to run deepseek-ocr model by api(/v1/chat/completion)

Nobody has claimed this yet.

LM Studio 0.3.31 build7
MLX version info:
- mlx-engine==c7cedfa
- mlx==0.29.3
- mlx-lm==0.28.4
- mlx-vlm==0.3.5

Run deepseek-ocr model by api(/v1/chat/completion), request with only one image and one text message
- in stream mode, error: TypeError: object of type 'NoneType' has no len()
- in none stream, error: Channel Error
```
2025-11-21 15:25:03 [INF0l Returning "data":["id":"deepseek-ocr","object":"model","owned by":"organization owner"object":"list"]
2025-11-21 15:26:41[INFO] [LM STUDI0 SERVER] Running chat completion on conversation with 2 messages.2025-11-21 15:26:41[LM STUDI0 SERVER]Streaming response...[INFO]
2025-11-21 15:26:41[ERROR] Error in iterating prediction stream: TypeError: object of type 'NoneType' has no len(). Error Data:n/a,Additional Data:n/a
2025-11-21 15:37:13[INFO] ILM STUDI0 SERVER] Running chat completion on conversation with 2 messages.
2025-11-21 15:37:13[ERROR]Error:Channel Error
2025-11-21 15:38:42[INF0] Server stopped.
```

Reproduce the deepseek-ocr request through the /v1/chat/completion entry point using one image and one text message, with streaming enabled and disabled. Trace the failure reported as a NoneType length error in stream mode and a Channel Error otherwise. Done means both request modes return a successful completion.

Written by the indexing model from the issue text.