feat: update default embedding model to nvidia/nv-embedqa-e5-v5 (#1015)

* add tests for preview embedding apis

* update default model to nvidia/nv-embedqa-e5-v5

* lint fixes w/ latest versions of black, isort, hatch

integrations/nvidia/src/haystack_integrations/components/embedders/nvidia/document_embedder.py

@@ -93,7 +93,7 @@ def __init__(
         self._initialized = False
 
         if is_hosted(api_url) and not self.model:  # manually set default model
-            self.model = "NV-Embed-QA"
+            self.model = "nvidia/nv-embedqa-e5-v5"
 
     def default_model(self):
         """Set default model in local NIM mode."""

integrations/nvidia/src/haystack_integrations/components/embedders/nvidia/text_embedder.py

@@ -77,7 +77,7 @@ def __init__(
         self._initialized = False
 
         if is_hosted(api_url) and not self.model:  # manually set default model
-            self.model = "NV-Embed-QA"
+            self.model = "nvidia/nv-embedqa-e5-v5"
 
     def default_model(self):
         """Set default model in local NIM mode."""

integrations/nvidia/tests/conftest.py

@@ -2,9 +2,10 @@
 
 import pytest
 from haystack.utils import Secret
-from haystack_integrations.utils.nvidia import Model, NimBackend
 from requests_mock import Mocker
 
+from haystack_integrations.utils.nvidia import Model, NimBackend
+
 
 class MockBackend(NimBackend):
     def __init__(self, model: str, api_key: Optional[Secret] = None, model_kwargs: Optional[Dict[str, Any]] = None):

integrations/nvidia/tests/test_base_url.py

@@ -1,4 +1,5 @@
 import pytest
+
 from haystack_integrations.components.embedders.nvidia import NvidiaDocumentEmbedder, NvidiaTextEmbedder
 from haystack_integrations.components.generators.nvidia import NvidiaGenerator
 

integrations/nvidia/tests/test_document_embedder.py

@@ -3,6 +3,7 @@
 import pytest
 from haystack import Document
 from haystack.utils import Secret
+
 from haystack_integrations.components.embedders.nvidia import EmbeddingTruncateMode, NvidiaDocumentEmbedder
 
 from . import MockBackend
@@ -14,7 +15,7 @@ def test_init_default(self, monkeypatch):
         embedder = NvidiaDocumentEmbedder()
 
         assert embedder.api_key == Secret.from_env_var("NVIDIA_API_KEY")
-        assert embedder.model == "NV-Embed-QA"
+        assert embedder.model == "nvidia/nv-embedqa-e5-v5"
         assert embedder.api_url == "https://ai.api.nvidia.com/v1/retrieval/nvidia"
         assert embedder.prefix == ""
         assert embedder.suffix == ""
@@ -372,15 +373,34 @@ def test_run_integration_with_nim_backend(self):
             assert isinstance(doc.embedding, list)
             assert isinstance(doc.embedding[0], float)
 
+    @pytest.mark.parametrize(
+        "model, api_url",
+        [
+            ("NV-Embed-QA", None),
+            ("snowflake/arctic-embed-l", "https://integrate.api.nvidia.com/v1"),
+            ("nvidia/nv-embed-v1", "https://integrate.api.nvidia.com/v1"),
+            ("nvidia/nv-embedqa-mistral-7b-v2", "https://integrate.api.nvidia.com/v1"),
+            ("nvidia/nv-embedqa-e5-v5", "https://integrate.api.nvidia.com/v1"),
+            ("baai/bge-m3", "https://integrate.api.nvidia.com/v1"),
+        ],
+        ids=[
+            "NV-Embed-QA",
+            "snowflake/arctic-embed-l",
+            "nvidia/nv-embed-v1",
+            "nvidia/nv-embedqa-mistral-7b-v2",
+            "nvidia/nv-embedqa-e5-v5",
+            "baai/bge-m3",
+        ],
+    )
     @pytest.mark.skipif(
         not os.environ.get("NVIDIA_API_KEY", None),
         reason="Export an env var called NVIDIA_API_KEY containing the NVIDIA API key to run this test.",
     )
     @pytest.mark.integration
-    def test_run_integration_with_api_catalog(self):
+    def test_run_integration_with_api_catalog(self, model, api_url):
         embedder = NvidiaDocumentEmbedder(
-            model="NV-Embed-QA",
-            api_url="https://ai.api.nvidia.com/v1/retrieval/nvidia",
+            model=model,
+            **({"api_url": api_url} if api_url else {}),
             api_key=Secret.from_env_var("NVIDIA_API_KEY"),
         )
         embedder.warm_up()

integrations/nvidia/tests/test_generator.py

@@ -5,9 +5,10 @@
 
 import pytest
 from haystack.utils import Secret
-from haystack_integrations.components.generators.nvidia import NvidiaGenerator
 from requests_mock import Mocker
 
+from haystack_integrations.components.generators.nvidia import NvidiaGenerator
+
 
 @pytest.fixture
 def mock_local_chat_completion(requests_mock: Mocker) -> None:

integrations/nvidia/tests/test_text_embedder.py

@@ -2,6 +2,7 @@
 
 import pytest
 from haystack.utils import Secret
+
 from haystack_integrations.components.embedders.nvidia import EmbeddingTruncateMode, NvidiaTextEmbedder
 
 from . import MockBackend
@@ -169,15 +170,34 @@ def test_run_integration_with_nim_backend(self):
         assert all(isinstance(x, float) for x in embedding)
         assert "usage" in meta
 
+    @pytest.mark.parametrize(
+        "model, api_url",
+        [
+            ("NV-Embed-QA", None),
+            ("snowflake/arctic-embed-l", "https://integrate.api.nvidia.com/v1"),
+            ("nvidia/nv-embed-v1", "https://integrate.api.nvidia.com/v1"),
+            ("nvidia/nv-embedqa-mistral-7b-v2", "https://integrate.api.nvidia.com/v1"),
+            ("nvidia/nv-embedqa-e5-v5", "https://integrate.api.nvidia.com/v1"),
+            ("baai/bge-m3", "https://integrate.api.nvidia.com/v1"),
+        ],
+        ids=[
+            "NV-Embed-QA",
+            "snowflake/arctic-embed-l",
+            "nvidia/nv-embed-v1",
+            "nvidia/nv-embedqa-mistral-7b-v2",
+            "nvidia/nv-embedqa-e5-v5",
+            "baai/bge-m3",
+        ],
+    )
     @pytest.mark.skipif(
         not os.environ.get("NVIDIA_API_KEY", None),
         reason="Export an env var called NVIDIA_API_KEY containing the NVIDIA API key to run this test.",
     )
     @pytest.mark.integration
-    def test_run_integration_with_api_catalog(self):
+    def test_run_integration_with_api_catalog(self, model, api_url):
         embedder = NvidiaTextEmbedder(
-            model="NV-Embed-QA",
-            api_url="https://ai.api.nvidia.com/v1/retrieval/nvidia",
+            model=model,
+            **({"api_url": api_url} if api_url else {}),
             api_key=Secret.from_env_var("NVIDIA_API_KEY"),
         )
         embedder.warm_up()