Crate weirwood 

weirwood — privacy-preserving XGBoost inference via Fully Homomorphic Encryption.

Load a trained XGBoost model and evaluate it either in plaintext (for testing) or encrypted under FHE so the server learns nothing about the input.

Quickstart — plaintext

use weirwood::{model::WeirwoodTree, eval::PlaintextEvaluator};

// `from_file` dispatches by extension: `.ubj` → Universal Binary JSON,
// anything else → JSON.
let model = WeirwoodTree::from_file("model.ubj")?;
let proba = PlaintextEvaluator.predict_proba(&model, &[1.0_f32, 0.5, 3.2, 0.1])?;

Quickstart — encrypted gRPC inference

With the transport Cargo feature enabled. The server holds the XGBoost model and reports its shape during session setup, so the client never has to load the model file:

use weirwood::transport::WeirwoodClient;

let mut client = WeirwoodClient::connect("http://127.0.0.1:9999").await?;
let proba = client.predict_proba(&[1.0_f32, 0.5, 3.2, 0.1]).await?;

Re-exports

pub use eval::fhe;

pub use error::Error;

Modules

error

eval

Evaluators for running inference over a loaded WeirwoodTree.

model

XGBoost model loading and internal representation.