{ "system_prompt": "You produce a complete structured design description from an image. Return only valid JSON in the requested full schema.", "user_prompt": "Generate the complete JSON containing high_level_description, style_description, elements.background, and elements.layout. Every layout item must contain type, label, parent, bbox in xyxy_1000 coordinates, and desc. When exact visible text is available for an item of type \"text\", include it in the text field. Also include style_description.color_palette and preserve the full annotation schema: layout items include node_kind, color_palette, and role or role_family when applicable. The type value is the group_type or element_type of the node. Use null for unavailable style descriptions and [] for empty color palettes.", "enable_thinking": false, "max_model_len": 32768, "max_new_tokens": 20000, "max_pixels": 2359296, "gpu_memory_utilization": 0.82, "seed": 42, "temperature": 0.0, "top_p": 1.0, "top_k": 0, "repetition_penalty": 1.0, "frequency_penalty": 0.0, "dtype": "bfloat16", "tensor_parallel_size": 1, "max_num_seqs": 1, "gdn_prefill_backend": "triton", "compilation_config": { "mode": 0, "cudagraph_mode": 2, "cudagraph_capture_sizes": [ 128, 192, 256 ] }, "environment": { "VLLM_USE_DEEP_GEMM": "0", "VLLM_MOE_USE_DEEP_GEMM": "0", "VLLM_USE_FLASHINFER_SAMPLER": "0" }, "upstream_validated_runtime": { "python": "3.11", "vllm": "0.20.2" }, "output_bbox_format": "xyxy_1000", "output_nodes_path": "elements.layout" }