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
fix(mcp): batch tools resolve model names through core's registry
Nandakishorabb979c5bd53
diff --git a/laya/mcp/tools.py b/laya/mcp/tools.pyindex d0d40e0b..1cb69356 100644--- a/laya/mcp/tools.py+++ b/laya/mcp/tools.py@@ -491,14 +491,13 @@ def _validate_batch_model(value: Any, where: str) -> str | None:separate from ``validate_model``, whose "auto" means "answer viarouter.predict" on the single-request path."""- if value is None or value == "auto":+ if value is None:return None- if value not in VALID_MODELS:- raise ToolError(- "invalid_model",- f"{where}['model'] must be one of {sorted(VALID_MODELS)}, got {value!r}",- )- return value+ try:+ name = validate_model(value)+ except ToolError as error:+ raise ToolError("invalid_model", "%s['model']: %s" % (where, error)) from None+ return None if name == AUTO else namedef _validate_batch_str(value: Any, where: str, allow_empty: bool = False) -> str: