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
docs(serve): _default_model_option docstring matches the 0.4.0 multilingual default
Bruce-Yiid6c100bae056
diff --git a/laya/serve.py b/laya/serve.pyindex fd3da64d..4ba72d15 100644--- a/laya/serve.py+++ b/laya/serve.py@@ -557,9 +557,9 @@ def _default_model_option(extra_models: Optional[Dict[str, Any]] = None) -> Dict"""Routing fallback from ``LAYA_DEFAULT_MODEL``, as a ``Router`` keyword; unset sends nothing.``Router.default`` answers the two states that carry no language evidence at all: no letters,- and Latin script too short to identify ("Quero cancelar", "Esqueci minha senha"). The README- tells a deployment whose traffic is mostly non-English to set ``Router(default="multilingual")``,- and this is the only way such a deployment can say so without writing its own server. Left out+ and Latin script too short to identify ("Quero cancelar", "Esqueci minha senha"). ``Router``+ defaults to ``multilingual`` since 0.4.0, and a deployment whose traffic is mostly English+ may set ``Router(default="english")`` instead, as the README describes. Left outof the constructor when unset, so the value cannot drift from ``Router``'s own default -- thesame reasoning as ``_resolve_max_loaded`` above.