Merge pull request #10 from langchain-ai/mattf/test-chat-serialization

add a serialization test of ChatNVIDIA using dumps/loads

libs/ai-endpoints/tests/integration_tests/test_chat_models.py

@@ -2,6 +2,8 @@
 import warnings
 
 import pytest
+from langchain_core.load.dump import dumps
+from langchain_core.load.load import loads
 from langchain_core.messages import BaseMessage, HumanMessage, SystemMessage
 
 from langchain_nvidia_ai_endpoints.chat_models import ChatNVIDIA
@@ -273,3 +275,11 @@ def test_ai_endpoints_invoke_top_p_positive(chat_model: str, mode: dict) -> None
     result1 = llm1.invoke("What's in a top_p?")
     assert isinstance(result1.content, str)
     assert result0.content != result1.content
+
+
+def test_serialize_chatnvidia() -> None:
+    model = loads(
+        dumps(ChatNVIDIA()), valid_namespaces=["langchain_nvidia_ai_endpoints"]
+    )
+    result = model.invoke("What is there if there is nothing?")
+    assert isinstance(result.content, str)

libs/ai-endpoints/tests/unit_tests/test_serialization.py

@@ -0,0 +1,16 @@
+from langchain_core.load.dump import dumps
+from langchain_core.load.load import loads
+
+from langchain_nvidia_ai_endpoints import ChatNVIDIA
+
+
+def test_serialize_chatnvidia() -> None:
+    secret = "a-bogus-key"
+    x = ChatNVIDIA(nvidia_api_key=secret)
+    y = loads(
+        dumps(x),
+        secrets_map={"NVIDIA_API_KEY": secret},
+        valid_namespaces=["langchain_nvidia_ai_endpoints"],
+    )
+    assert x == y
+    assert isinstance(y, ChatNVIDIA)