Open-weight route
Laya
ConvAI Innovations · Self-hostable
Independent guide · verified Sep 20, 2026
Laya is an open-weight decision-model project from ConvAI Innovations. It is described as a non-autoregressive route for typed choices, scores and boolean decisions.
Independent guide · not affiliated with ConvAI Innovations.
Quick answer
Laya takes state plus typed questions and returns structured decision outputs rather than generated prose. Its project materials describe choice, score and noul primitives and an open/self-host route. Facts on this page are project-reported and should be re-verified before production use.
Who built it
Open-weight route
ConvAI Innovations · Self-hostable
Laya can be reviewed and self-hosted through the project’s published repositories and model cards. That changes who controls deployment, but it does not remove infrastructure, model-serving or evaluation work.
Open-source and self-hostable; compute and operational costs still apply.
What Laya does
Choose one option from a known list.
Rate something on an ordered scale.
Answer a yes/no proposition with a probability.
Laya Python Quickstart
This site does not run Laya. Verify the package, model card and hardware requirements with the official project before using it locally.
import laya
agent = laya.load("convaiinnovations/laya")
result = agent.predict(
{"message": "My payment failed twice"},
laya.triage_questions(),
)
print(result)Try the same task here
Laya is not connected to this site, so there is no fake Laya playground. You can run the same bounded classification task with the hosted Jev route.
No. JevClassify does not run Laya in its live classifier. This page is an independent, source-linked guide and sends users to the real Jev classifier for an interactive test.
The Laya project and Hugging Face model card describe the model as Apache 2.0. Verify the current repository and model-card terms before shipping.
No. Self-hosting still requires compute, storage, deployment, monitoring and maintenance.