raullenchai

Native Rapid-MLX provider for DeepSeek Harness — teaches DSH what the local server already knows (memory-fitted context via max_model_len), plus model-management tools and a /rapid-mlx overview.