Crate cufflink_model 

cufflink-model

Run small trained text classifiers inside Cufflink services. Training happens offline (the Python toolkit in ml/); a service embeds the exported file and classifies with it, natively or in WASM, without calling out to anything.

use cufflink_model::Model;

static GENRES: &[u8] = include_bytes!("../models/genre.cflm");

let model = Model::from_bytes(GENRES)?;
let p = model.predict("The Dragon of Ember Hill");
if p.confidence >= model.threshold().unwrap_or(0.9) {
    // use p.label
}

A WASM guest is a fresh instance on every call, so Model::from_bytes reads the file in place: it checks the layout once and keeps borrowed slices, with no tables rebuilt. Every offset is bounds-checked at load, and nothing in this crate panics on a malformed file or unusual input.

The file format is described in docs/design-notes/models.md.

Modules

featurise

Turning text into the n-gram features a model scores.

Structs

Model

A loaded model, borrowing the file it was read from.

Prediction

A model’s answer for one text.

Enums

Kind

The kind of model a file holds.

ModelError

Why a model file couldn’t be read.

Constants

FORMAT_VERSION

The file layout this crate reads.

MAGIC

The first four bytes of every model file.