Merge pull request #103 from langchain-ai/mattf/remove-nvolveqa-40k

remove compatibility mode for nvolveqa_40k embedding model

libs/ai-endpoints/langchain_nvidia_ai_endpoints/_statics.py

@@ -494,8 +494,6 @@ def validate_client(self) -> "Model":
         endpoint="https://ai.api.nvidia.com/v1/retrieval/nvidia/embeddings",
         aliases=[
             "ai-embed-qa-4",
-            "playground_nvolveqa_40k",
-            "nvolveqa_40k",
         ],
     ),
     "nvidia/nv-embed-v1": Model(

libs/ai-endpoints/langchain_nvidia_ai_endpoints/embeddings.py

@@ -1,6 +1,5 @@
 """Embeddings Components Derived from NVEModel/Embeddings"""
 
-import warnings
 from typing import Any, List, Literal, Optional
 
 from langchain_core.embeddings import Embeddings
@@ -97,16 +96,6 @@ def __init__(self, **kwargs: Any):
         # same for base_url
         self.base_url = self._client.base_url
 
-        # todo: remove when nvolveqa_40k is removed from MODEL_TABLE
-        if "model" in kwargs and kwargs["model"] in [
-            "playground_nvolveqa_40k",
-            "nvolveqa_40k",
-        ]:
-            warnings.warn(
-                'Setting truncate="END" for nvolveqa_40k backward compatibility'
-            )
-            self.truncate = "END"
-
     @property
     def available_models(self) -> List[Model]:
         """

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

@@ -53,8 +53,6 @@ async def test_embed_documents_multiple_async(embedding_model: str, mode: dict)
 
 
 def test_embed_query_long_text(embedding_model: str, mode: dict) -> None:
-    if embedding_model in ["playground_nvolveqa_40k", "nvolveqa_40k"]:
-        pytest.skip("Skip test for nvolveqa-40k due to compat override of truncate")
     embedding = NVIDIAEmbeddings(model=embedding_model, **mode)
     text = "nvidia " * 10240
     with pytest.raises(Exception):
@@ -71,8 +69,6 @@ def test_embed_documents_batched_texts(embedding_model: str, mode: dict) -> None
 
 
 def test_embed_documents_mixed_long_texts(embedding_model: str, mode: dict) -> None:
-    if embedding_model in ["playground_nvolveqa_40k", "nvolveqa_40k"]:
-        pytest.skip("Skip test for nvolveqa-40k due to compat override of truncate")
     embedding = NVIDIAEmbeddings(model=embedding_model, **mode)
     count = _DEFAULT_BATCH_SIZE * 2 - 1
     texts = ["nvidia " * 32] * count
@@ -101,16 +97,5 @@ def test_embed_documents_truncate(
     assert len(output) == count
 
 
-@pytest.mark.parametrize("nvolveqa_40k", ["playground_nvolveqa_40k", "nvolveqa_40k"])
-def test_embed_nvolveqa_40k_compat(nvolveqa_40k: str, mode: dict) -> None:
-    if mode:
-        pytest.skip("Test only relevant for API Catalog")
-    with pytest.warns(UserWarning):
-        embedding = NVIDIAEmbeddings(model=nvolveqa_40k, truncate="NONE", **mode)
-    text = "nvidia " * 2048
-    output = embedding.embed_query(text)
-    assert len(output) > 3
-
-
 # todo: test max_length > max length accepted by the model
 # todo: test max_batch_size > max batch size accepted by the model