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.
Python calibrationclassificationdecision-modelhuggingfacejevmodernbertmultilingualnlppythonpytorchroutingtyped-decisionszero-shot
import unittest
from laya.agent import Agent
from laya.common import render_options
class CriteriaTests(unittest.TestCase):
def test_noul_criteria_boolean_keys(self):
qdef = {
"type": "noul",
"instructions": "Is this a refund?",
"criteria": {True: "User requests money back", False: "User does not ask for money back"}
}
internal = Agent._to_internal(qdef)
self.assertIn("true", internal["crit"])
self.assertIn("false", internal["crit"])
opts = render_options(internal)
self.assertEqual(len(opts), 2)
self.assertTrue(opts[0].startswith("false: User does not ask for money back"))
self.assertTrue(opts[1].startswith("true: User requests money back"))
def test_choice_criteria_list_expansion(self):
qdef = {
"type": "choice",
"instructions": "Select urgency",
"criteria": ["low", "medium", "high"]
}
internal = Agent._to_internal(qdef)
self.assertEqual(list(internal["crit"].keys()), ["low", "medium", "high"])
if __name__ == "__main__":
unittest.main()