Cache calls to Open AI during tests (#91)

pyproject.toml

@@ -57,7 +57,7 @@ optional = true
 langchain-openai = ">=0.0.3,<0.1"
 
 [tool.ruff]
-select = [
+lint.select = [
   "E",  # pycodestyle
   "F",  # pyflakes
   "I",  # isort

tests/integration_tests/test_vectorstores.py

@@ -59,9 +59,28 @@ def weaviate_client(docker_ip, docker_services):
     client.close()
 
 
-@pytest.fixture
+@pytest.fixture(scope="session")
 def embedding_openai():
-    yield OpenAIEmbeddings()
+    class MemoizedOpenAIEmbeddings(OpenAIEmbeddings):
+        def __init__(self):
+            super().__init__()
+            object.__setattr__(self, "cache", {})
+
+        def embed_query(self, query):
+            if query not in self.cache:
+                # Call the base class method if result not cached
+                self.cache[query] = super().embed_query(query)
+            return self.cache[query]
+
+        def embed_documents(self, documents):
+            # Use tuple to allow documents list to be hashable
+            hashable_docs = tuple(documents)
+            if hashable_docs not in self.cache:
+                # Call the base class method if result not cached
+                self.cache[hashable_docs] = super().embed_documents(documents)
+            return self.cache[hashable_docs]
+
+    yield MemoizedOpenAIEmbeddings()
 
 
 @pytest.fixture
@@ -622,7 +641,7 @@ def test_embedding_property(weaviate_client, embedding_openai):
         embedding=embedding_openai,
     )
 
-    assert type(docsearch.embeddings) == OpenAIEmbeddings
+    assert type(docsearch.embeddings) == type(embedding_openai)
 
 
 def test_documents_with_many_properties(weaviate_client, embedding_openai):