Merge pull request #2102 from weaviate/pr/2090

Add endpoint setting for text2vec-openai and text2vec-morph

Author: dirkkul

test/collection/test_config.py

@@ -319,6 +319,21 @@ def test_basic_config():
             }
         },
     ),
+    (
+        Configure.Vectorizer.text2vec_openai(
+            vectorize_collection_name=False,
+            model="ada",
+            endpoint="/api/v3/embeddings",
+        ),
+        {
+            "text2vec-openai": {
+                "vectorizeClassName": False,
+                "model": "ada",
+                "endpoint": "/api/v3/embeddings",
+                "isAzure": False,
+            }
+        },
+    ),
     (
         Configure.Vectorizer.text2vec_mistral(
             vectorize_collection_name=False,
@@ -1759,6 +1774,28 @@ def test_vector_config_flat_pq() -> None:
             }
         },
     ),
+    (
+        [
+            Configure.NamedVectors.text2vec_openai(
+                name="test",
+                source_properties=["prop"],
+                endpoint="/api/v3/embeddings",
+            )
+        ],
+        {
+            "test": {
+                "vectorizer": {
+                    "text2vec-openai": {
+                        "properties": ["prop"],
+                        "vectorizeClassName": True,
+                        "endpoint": "/api/v3/embeddings",
+                        "isAzure": False,
+                    }
+                },
+                "vectorIndexType": "hnsw",
+            }
+        },
+    ),
     (
         [Configure.NamedVectors.text2vec_mistral(name="test", source_properties=["prop"])],
         {
@@ -2361,6 +2398,28 @@ def test_config_with_named_vectors(
             }
         },
     ),
+    (
+        [
+            Configure.Vectors.text2vec_openai(
+                name="test",
+                source_properties=["prop"],
+                endpoint="/api/v3/embeddings",
+            )
+        ],
+        {
+            "test": {
+                "vectorizer": {
+                    "text2vec-openai": {
+                        "properties": ["prop"],
+                        "vectorizeClassName": True,
+                        "endpoint": "/api/v3/embeddings",
+                        "isAzure": False,
+                    }
+                },
+                "vectorIndexType": "hnsw",
+            }
+        },
+    ),
     (
         [Configure.Vectors.text2vec_mistral(name="test", source_properties=["prop"])],
         {
@@ -2408,6 +2467,27 @@ def test_config_with_named_vectors(
             }
         },
     ),
+    (
+        [
+            Configure.Vectors.text2vec_morph(
+                name="test",
+                source_properties=["prop"],
+                endpoint="/api/v3/embeddings",
+            )
+        ],
+        {
+            "test": {
+                "vectorizer": {
+                    "text2vec-morph": {
+                        "vectorizeClassName": True,
+                        "properties": ["prop"],
+                        "endpoint": "/api/v3/embeddings",
+                    }
+                },
+                "vectorIndexType": "hnsw",
+            }
+        },
+    ),
     (
         [
             Configure.Vectors.text2vec_google(

weaviate/collections/classes/config_named_vectors.py

@@ -401,6 +401,7 @@ def text2vec_openai(
         *,
         base_url: Optional[AnyHttpUrl] = None,
         dimensions: Optional[int] = None,
+        endpoint: Optional[str] = None,
         model: Optional[Union[OpenAIModel, str]] = None,
         model_version: Optional[str] = None,
         type_: Optional[OpenAIType] = None,
@@ -424,6 +425,7 @@ def text2vec_openai(
             vectorize_collection_name: Whether to vectorize the collection name. Defaults to `True`.
             base_url: The base URL to use where API requests should go. Defaults to `None`, which uses the server-defined default.
             dimensions: Number of dimensions. Applicable to v3 OpenAI models only. Defaults to `None`, which uses the server-defined default.
+            endpoint: The API path to append to `base_url`, e.g. `/api/v3/embeddings`. Defaults to `None`, which uses the server-defined default.
 
         Raises:
             pydantic.ValidationError: If `type_` is not a valid value from the `OpenAIType` type.
@@ -438,6 +440,7 @@ def text2vec_openai(
                 type_=type_,
                 vectorizeClassName=vectorize_collection_name,
                 dimensions=dimensions,
+                endpoint=endpoint,
             ),
             vector_index_config=vector_index_config,
         )

weaviate/collections/classes/config_vectorizers.py

@@ -310,6 +310,7 @@ class _Text2VecMorphConfig(_VectorizerConfigCreate):
     model: Optional[str]
     vectorizeClassName: bool
     baseURL: Optional[AnyHttpUrl]
+    endpoint: Optional[str]
 
     def _to_dict(self) -> Dict[str, Any]:
         ret_dict = super()._to_dict()
@@ -336,6 +337,7 @@ class _Text2VecOpenAIConfig(_VectorizerConfigCreate):
     )
     baseURL: Optional[AnyHttpUrl]
     dimensions: Optional[int]
+    endpoint: Optional[str]
     model: Optional[str]
     modelVersion: Optional[str]
     type_: Optional[OpenAIType]
@@ -1137,6 +1139,7 @@ def text2vec_openai(
         vectorize_collection_name: bool = True,
         base_url: Optional[AnyHttpUrl] = None,
         dimensions: Optional[int] = None,
+        endpoint: Optional[str] = None,
     ) -> _VectorizerConfigCreate:
         """Create a `_Text2VecOpenAIConfigCreate` object for use when vectorizing using the `text2vec-openai` model.
 
@@ -1150,6 +1153,7 @@ def text2vec_openai(
             vectorize_collection_name: Whether to vectorize the collection name. Defaults to `True`.
             base_url: The base URL to use where API requests should go. Defaults to `None`, which uses the server-defined default.
             dimensions: Number of dimensions. Applicable to v3 OpenAI models only. Defaults to `None`, which uses the server-defined default.
+            endpoint: The API path to append to `base_url`, e.g. `/api/v3/embeddings`. Defaults to `None`, which uses the server-defined default.
 
         Raises:
             pydantic.ValidationError: If `type_` is not a valid value from the `OpenAIType` type.
@@ -1161,6 +1165,7 @@ def text2vec_openai(
             type_=type_,
             vectorizeClassName=vectorize_collection_name,
             dimensions=dimensions,
+            endpoint=endpoint,
         )
 
     @staticmethod

weaviate/collections/classes/config_vectors.py

@@ -664,6 +664,7 @@ def text2vec_morph(
         quantizer: Optional[_QuantizerConfigCreate] = None,
         base_url: Optional[AnyHttpUrl] = None,
         model: Optional[str] = None,
+        endpoint: Optional[str] = None,
         source_properties: Optional[List[str]] = None,
         vector_index_config: Optional[_VectorIndexConfigCreate] = None,
         vectorize_collection_name: bool = True,
@@ -678,6 +679,7 @@ def text2vec_morph(
             quantizer: The quantizer to use for the vector index. If not provided, no quantization will be applied.
             base_url: The base URL to use where API requests should go. Defaults to `None`, which uses the server-defined default.
             model: The model to use. Defaults to `None`, which uses the server-defined default.
+            endpoint: The API path to append to `base_url`, e.g. `/api/v3/embeddings`. Defaults to `None`, which uses the server-defined default.
             source_properties: Which properties should be included when vectorizing. By default all text properties are included.
             vector_index_config: The configuration for Weaviate's vector index. Use `wvc.config.Configure.VectorIndex` to create a vector index configuration. None by default
             vectorize_collection_name: Whether to vectorize the collection name. Defaults to `True`.
@@ -688,6 +690,7 @@ def text2vec_morph(
             vectorizer=_Text2VecMorphConfig(
                 baseURL=base_url,
                 model=model,
+                endpoint=endpoint,
                 vectorizeClassName=vectorize_collection_name,
             ),
             vector_index_config=_IndexWrappers.single(vector_index_config, quantizer),
@@ -739,6 +742,7 @@ def text2vec_openai(
         quantizer: Optional[_QuantizerConfigCreate] = None,
         base_url: Optional[AnyHttpUrl] = None,
         dimensions: Optional[int] = None,
+        endpoint: Optional[str] = None,
         model: Optional[Union[OpenAIModel, str]] = None,
         model_version: Optional[str] = None,
         type_: Optional[OpenAIType] = None,
@@ -756,6 +760,7 @@ def text2vec_openai(
             quantizer: The quantizer to use for the vector index. If not provided, no quantization will be applied.
             base_url: The base URL to use where API requests should go. Defaults to `None`, which uses the server-defined default.
             dimensions: Number of dimensions. Applicable to v3 OpenAI models only. Defaults to `None`, which uses the server-defined default.
+            endpoint: The API path to append to `base_url`, e.g. `/api/v3/embeddings`. Defaults to `None`, which uses the server-defined default.
             model: The model to use. Defaults to `None`, which uses the server-defined default.
             model_version: The model version to use. Defaults to `None`, which uses the server-defined default.
             type_: The type of model to use. Defaults to `None`, which uses the server-defined default.
@@ -776,6 +781,7 @@ def text2vec_openai(
                 type_=type_,
                 vectorizeClassName=vectorize_collection_name,
                 dimensions=dimensions,
+                endpoint=endpoint,
             ),
             vector_index_config=_IndexWrappers.single(vector_index_config, quantizer),
         )