tento-event-classifier

tento-event-classifier classifies calendar event titles into coarse activity categories using an ONNX model and an optional Python training subprocess.

Public surface

• EventCategory is the stable category enum.
• Model::new(path) loads an ONNX model through ort.
• EventClassifier::classify(input) maps event text to categories.
• EventClassifier::train(training_data, out_model) delegates training to a Trainer when one was configured.
• EventClassifierConfig declares the root, virtualenv, and model path environment variables.

Boundaries

Inference is Rust-owned and should not depend on Python at runtime. Training is optional and isolated behind trainer::Trainer, which launches the Python RPC process.

Callers should pass plain text inputs and consume EventCategory. They should not couple to ONNX tensor names, Python module paths, or the training data file format except where they are explicitly training a model.