upd openai version (#277)

* upd openai version

* narrow type of underpopulated classes field on `SplitReadinessResult`

Author: voorhs

pyproject.toml

@@ -39,7 +39,7 @@ dependencies = [
     "pathlib (>=1.0.1,<2.0.0)",
     "pydantic (>=2.10.5,<3.0.0)",
     "faiss-cpu (>=1.9.0,<2.0.0)",
-    "openai (>=1.59.6,<2.0.0)",
+    "openai (>=2,<3)",
     "datasets (>=3.2.0,<5.0.0)",
     "xxhash (>=3.5.0,<4.0.0)",
     "python-dotenv (>=1.0.1,<2.0.0)",

src/autointent/context/data_handler/_stratification.py

@@ -55,7 +55,7 @@ class SplitReadinessResult:
     """
 
     ready: bool
-    underpopulated_classes: list[tuple[int | str | None, int]]
+    underpopulated_classes: list[tuple[int, int]]
     min_samples_per_class_required: int
     reason: str | None
 
@@ -378,9 +378,9 @@ def _get_adjusted_test_size(self, n: int, k: int) -> float:
 
 def _check_multiclass_counts(
     dataset: HFDataset, label_feature: str, min_samples_per_class: int
-) -> list[tuple[int | str | None, int]]:
+) -> list[tuple[int, int]]:
     """Return (label, count) for each class with fewer than min_samples_per_class samples."""
-    labels = dataset[label_feature]
+    labels: list[int] = dataset[label_feature]
     counts = Counter(labels)
     return [(label, count) for label, count in counts.items() if count < min_samples_per_class]