Pure-Rust LightGBM inference: parses the text model format directly — no FFI, zero deps. Optional ONNX Runtime and CatBoost backends behind the same small Model trait.
# Pin formatting so `cargo fmt` is reproducible for all contributors,# independent of any global ~/.rustfmt.toml.max_width=70array_width=70chain_width=70fn_call_width=70short_array_element_width_threshold=20