NandhaKishorM/laya
Non-autoregressive System 1 decision engine. Typed choice, score and yes/no decisions over any text in a single forward pass, in 100+ languages, with a router that picks the right checkpoint per request.
Laya is a fast, multilingual decision engine that classifies text into structured answers (choice, score, yes/no) in a single forward pass — no text generation, so it's fast (33 ms per question) and doesn't hallucinate.
It works by taking a piece of text (email, ticket, review) and a set of typed questions, then directly outputs answers like "billing", "blocking", or a probability. A built-in router detects the language and picks the best of three pre-trained checkpoints (English, multilingual, or a fine-tuned domain model). It supports 100+ languages and long documents up to 8,192 tokens.
The models are trained with reinforcement learning against proper scoring rules (RLCD) and can be fine-tuned on your own domain data for higher accuracy. It ships as a Python package with a CLI, plus integrations for LangChain, LlamaIndex, CrewAI, and more.
Use cases: customer support triage, intent classification, content moderation, routing, and any task where you need fast, structured decisions from text.
Written about in
Related
- Dispatch
- Project
- Dispatch
- Project
- Project