Fix unbounded TypeAdapter cache causing memory leak in multi-threaded usage

src/openai/_models.py

@@ -3,6 +3,7 @@
 import os
 import inspect
 import weakref
+import threading
 from typing import (
     IO,
     TYPE_CHECKING,
@@ -54,7 +55,6 @@
     is_list,
     is_given,
     json_safe,
-    lru_cache,
     is_mapping,
     parse_date,
     coerce_boolean,
@@ -799,12 +799,21 @@ class GenericModel(BaseGenericModel, BaseModel):
 if not PYDANTIC_V1:
     from pydantic import TypeAdapter as _TypeAdapter
 
-    _CachedTypeAdapter = cast("TypeAdapter[object]", lru_cache(maxsize=None)(_TypeAdapter))
+    _type_adapter_cache: threading.local = threading.local()
+
+    def _get_cached_type_adapter(type_: type[_T]) -> _TypeAdapter[_T]:
+        cache: dict[type[Any], _TypeAdapter[Any]] = getattr(_type_adapter_cache, "adapters", None) or {}
+        _type_adapter_cache.adapters = cache
+        adapter = cache.get(type_)
+        if adapter is None:
+            adapter = _TypeAdapter(type_)
+            cache[type_] = adapter
+        return adapter
 
     if TYPE_CHECKING:
         from pydantic import TypeAdapter
     else:
-        TypeAdapter = _CachedTypeAdapter
+        TypeAdapter = _get_cached_type_adapter
 
     def _validate_non_model_type(*, type_: type[_T], value: object) -> _T:
         return TypeAdapter(type_).validate_python(value)

tests/test_models.py

@@ -961,3 +961,24 @@ def __getattr__(self, attr: str) -> Item: ...
     assert model.a.prop == 1
     assert isinstance(model.a, Item)
     assert model.other == "foo"
+
+
+@pytest.mark.skipif(PYDANTIC_V1, reason="TypeAdapter cache is only used in Pydantic v2")
+def test_type_adapter_cache_is_thread_local() -> None:
+    """Regression test for https://github.com/openai/openai-python/issues/2672
+
+    The TypeAdapter cache uses threading.local() to prevent memory leaks
+    in multi-threaded environments. Each thread maintains its own cache that
+    is cleaned up when the thread exits.
+    """
+    import threading
+
+    from openai._models import TypeAdapter, _type_adapter_cache
+
+    # Verify the cache is thread-local
+    assert isinstance(_type_adapter_cache, threading.local)
+
+    # Verify TypeAdapter returns a cached instance for the same type
+    adapter1 = TypeAdapter(int)
+    adapter2 = TypeAdapter(int)
+    assert adapter1 is adapter2, "TypeAdapter should return cached instances for the same type"