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
test(mcp): read the README as UTF-8 so the suite runs on the Windows lane
thread3d4cc765a645ca
diff --git a/tests/test_mcp.py b/tests/test_mcp.pyindex af5a4dd6..a70d93bd 100644--- a/tests/test_mcp.py+++ b/tests/test_mcp.py@@ -2180,7 +2180,11 @@ def test_batch_item_shape_as_documented():ok("schema/%s_item_shape" % tool, want in (by_name[tool].description or ""),"documents %r" % want)- readme = (Path(__file__).resolve().parents[1] / "README.md").read_text()+ # `encoding="utf-8"` explicitly: the README is UTF-8 and contains non-ASCII (the language+ # examples), while `Path.read_text()` defaults to the locale encoding -- cp1252 on the Windows+ # runner, where the decode raised `UnicodeDecodeError` and this suite failed. It went unseen+ # because the Windows lane ran this file without the `mcp` extra, so it skipped at the top.+ readme = (Path(__file__).resolve().parents[1] / "README.md").read_text(encoding="utf-8")marker = "one tool call takes an array of `{"ok("docs/readme_names_the_batch_shape", marker in readme, "sentence not found")if marker in readme: