InferenceSettings

class rapidtools.gui.InferenceSettings(raster_path: Path, output_dir: Path, prompt: str, backend: Literal['gemini', 'claude', 'openai', 'muse_spark', 'qwen', 'gemma4', 'llama', 'muse_glimmer', 'qwen_vl', 'hf'] = 'gemini', api_key: str = '', model_id: str = '', max_workers: int = 5, batch_size: int = 4, temperature: float = 0.4, max_tokens: int = 2048, json_mode: bool = False, load_in_4bit: bool = True, overlay_asset_outline: bool = True, outline_shape: str = 'geometry', outline_buffer: float | str = 0, outline_width: int | float | str = 6, outline_color: str = 'red', image_prefix: str = 'aerial', only_analyze_inference_crops: bool = True, imagery: str = 'aerial', min_footprint_coverage: float | None = None, pad_edges: bool | None = None, street_search_radius_m: float = 50.0, street_max_images: int = 1, street_vertical_crop: tuple[float, float] = (0.0, 1.0), mapillary_token: str = '', mapillary_start_date: str = '', mapillary_end_date: str = '', mapillary_rapid_only: bool = True, mapillary_max_images: int = 4, object_image_size: str = '2048', object_crop_buffer: float | str = '25%')

Bases: object

User-configurable options for the VLM inference stage.

property backend_spec: dict[str, Any]

The MODEL_BACKENDS record describing the selected backend.