laya-mlx
Copyright 2026 laya-mlx contributors

This product includes software derived from Laya:
https://github.com/NandhaKishorM/laya
Copyright Convai Innovations and Laya contributors. Licensed under Apache-2.0.
Original upstream source revision: 573e5b62696ba441230cd6be71d593331b5d23af

The token sequence construction, question rendering, confidence calculation,
presets, email utilities and language router are adapted from Laya.
The neural network is reimplemented using Apple's MLX, following Laya's
DecisionModel and the ModernBERT architecture in Hugging Face Transformers.
Model weights are downloaded separately from Convai Innovations on Hugging Face;
they are not included in this repository.

Selected inference, input validation, routing, language and email fixes adapted
from upstream revision 4aa6761 (v0.3.23 source tree). This is a selective sync,
not an implementation of every upstream API. The neural architecture parity
reference remains 573e5b62696ba441230cd6be71d593331b5d23af.

Additional lifecycle/registry, language, email, tournament and confidence-gate
behavior adapted from upstream revision
1adc59f7e371deb601fcfa18a14e25db238addcc (v0.4.1). Parallel option layout and masks
are reimplemented in MLX and checked against that revision. Sequential model
parity remains checked against the original reference above.
