Merge pull request #321 from weaviate/chore/vector-syntax-amendments

Fix breaking changes from v3.7.0 release

Author: bevzzz

src/alias/journey.test.ts

@@ -21,10 +21,12 @@ requireAtLeast(1, 32, 0).describe('manages collection aliases', () => {
       .then((aliases) => {
         expect(aliases).not.toBeUndefined();
         expect(aliases).toHaveLength(2);
-        expect(aliases).toEqual<Alias[]>([
-          { collection: 'PaulHewson', alias: 'Bono' },
-          { collection: 'GeorgeBarnes', alias: 'MachineGunKelly' },
-        ]);
+        expect(aliases).toEqual(
+          expect.arrayContaining<Alias>([
+            { collection: 'PaulHewson', alias: 'Bono' },
+            { collection: 'GeorgeBarnes', alias: 'MachineGunKelly' },
+          ])
+        );
       });
   });
 

src/collections/aggregate/integration.test.ts

@@ -90,7 +90,6 @@ describe('Testing of the collection.aggregate methods', () => {
           // },
         ],
         vectorizers: weaviate.configure.vectors.text2VecContextionary({
-          vectorizeCollectionName: false,
           vectorIndexConfig: weaviate.configure.vectorIndex.hnsw({ maxConnections: 64 }),
         }),
       })

src/collections/config/types/generative.ts

@@ -88,6 +88,8 @@ export type GenerativePaLMConfig = GenerativeGoogleConfig;
 export type GenerativeGoogleConfig = {
   apiEndpoint?: string;
   maxOutputTokens?: number;
+  model?: string;
+  /** @deprecated Use `model` instead. */
   modelId?: string;
   projectId?: string;
   temperature?: number;

src/collections/config/types/vectorizer.ts

@@ -19,6 +19,7 @@ type Text2VecPalmVectorizer = 'text2vec-palm';
 
 export type Vectorizer =
   | 'img2vec-neural'
+  | 'multi2vec-nvidia'
   | 'multi2vec-clip'
   | 'multi2vec-cohere'
   | 'multi2vec-bind'
@@ -65,6 +66,32 @@ export type Multi2VecField = {
   weight?: number;
 };
 
+/** The configuration for multi-media vectorization using the NVIDIA module.
+ *
+ * See the [documentation](https://weaviate.io/developers/weaviate/model-providers/nvidia/embeddings-multimodal) for detailed usage.
+ */
+export type Multi2VecNvidiaConfig = {
+  /** The model to use. Defaults to `None`, which uses the server-defined default. */
+  model?: string;
+  /** The base URL where API requests should go. */
+  baseURL?: string;
+  /** Whether to apply truncation. */
+  truncation?: boolean;
+  /** Format in which the embeddings are encoded. Defaults to `None`, so the embeddings are represented as a list of floating-point numbers. */
+  output_encoding?: string;
+  /** The image fields used when vectorizing. */
+  imageFields?: string[];
+  /** The text fields used when vectorizing. */
+  textFields?: string[];
+  /** The weights of the fields used for vectorization. */
+  weights?: {
+    /** The weights of the image fields. */
+    imageFields?: number[];
+    /** The weights of the text fields. */
+    textFields?: number[];
+  };
+};
+
 /** The configuration for multi-media vectorization using the CLIP module.
  *
  * See the [documentation](https://weaviate.io/developers/weaviate/model-providers/transformers/embeddings-multimodal) for detailed usage.
@@ -175,6 +202,9 @@ export type Multi2VecGoogleConfig = {
   /** Length of a video interval in seconds. */
   videoIntervalSeconds?: number;
   /** The model ID in use. */
+  model?: string;
+  /** The model ID in use.
+   * @deprecated Use `model` instead.*/
   modelId?: string;
   /** The dimensionality of the vector once embedded. */
   dimensions?: number;
@@ -393,8 +423,6 @@ export type Text2MultiVecJinaAIConfig = {
   dimensions?: number;
   /** The model to use. */
   model?: string;
-  /** Whether to vectorize the collection name. */
-  vectorizeCollectionName?: boolean;
 };
 
 /** @deprecated Use `Text2VecJinaAIConfig` instead. */
@@ -476,6 +504,9 @@ export type Text2VecGoogleConfig = {
   /** The API endpoint to use without a leading scheme such as `http://`. */
   apiEndpoint?: string;
   /** The model ID to use. */
+  model?: string;
+  /** The model ID to use.
+   * @deprecated Use `model `instead.*/
   modelId?: string;
   /** The project ID to use. */
   projectId?: string;
@@ -565,6 +596,8 @@ export type VectorizerConfig =
 
 export type VectorizerConfigType<V> = V extends 'img2vec-neural'
   ? Img2VecNeuralConfig | undefined
+  : V extends 'multi2vec-nvidia'
+  ? Multi2VecNvidiaConfig | undefined
   : V extends 'multi2vec-clip'
   ? Multi2VecClipConfig | undefined
   : V extends 'multi2vec-cohere'

src/collections/configure/generative.ts

@@ -246,7 +246,14 @@ export default {
     console.warn('The `generative-palm` module is deprecated. Use `generative-google` instead.');
     return {
       name: 'generative-palm',
-      config,
+      // Do not populate config key if config === undefined.
+      config: config
+        ? {
+            ...config,
+            // Only create modelId key if either config.model or config.modelId are defined.
+            ...(config?.modelId || config?.model ? { modelId: config?.model ?? config?.model } : undefined),
+          }
+        : undefined,
     };
   },
   /**
@@ -262,7 +269,14 @@ export default {
   ): ModuleConfig<'generative-google', GenerativeGoogleConfig | undefined> => {
     return {
       name: 'generative-google',
-      config,
+      // Do not populate config key if config === undefined.
+      config: config
+        ? {
+            ...config,
+            // Only create modelId key if either config.model or config.modelId are defined.
+            ...(config?.modelId || config?.model ? { modelId: config?.model ?? config?.model } : undefined),
+          }
+        : undefined,
     };
   },
   /**

src/collections/configure/index.ts

@@ -15,7 +15,7 @@ import {
 import generative from './generative.js';
 import reranker from './reranker.js';
 import { configure as configureVectorIndex, reconfigure as reconfigureVectorIndex } from './vectorIndex.js';
-import { multiVectors, vectors } from './vectorizer.js';
+import { multiVectors, vectorizer, vectors } from './vectorizer.js';
 
 import { parseWithDefault } from './parsing.js';
 
@@ -61,9 +61,9 @@ const configure = {
   multiVectors,
   reranker,
   /**
-   * @deprecated Use `vectors` instead.
+   * @deprecated Use `configure.vectors` instead.
    */
-  vectorizer: vectors,
+  vectorizer: vectorizer,
   vectors,
   vectorIndex: configureVectorIndex,
   dataType,
@@ -311,5 +311,6 @@ export {
   tokenization,
   vectorDistances,
   configureVectorIndex as vectorIndex,
+  vectorizer,
   vectors,
 };

src/collections/configure/types/vectorizer.ts

@@ -112,6 +112,37 @@ export type ConfigureTextMultiVectorizerOptions<
 
 export type Img2VecNeuralConfigCreate = Img2VecNeuralConfig;
 
+// model: Optional[str] = None,
+// truncation: Optional[bool] = None,
+// output_encoding: Optional[str],
+// vectorize_collection_name: bool = True,
+// base_url: Optional[AnyHttpUrl] = None,
+// image_fields: Optional[Union[List[str], List[Multi2VecField]]] = None,
+// text_fields: Optional[Union[List[str], List[Multi2VecField]]] = None,
+
+// model: The model to use. Defaults to `None`, which uses the server-defined default.
+// output_encoding: Format in which the embeddings are encoded. Defaults to `None`, so the embeddings are represented as a list of floating-point numbers.
+// 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.
+// image_fields: The image fields to use in vectorization.
+// text_fields: The text fields to use in vectorization.
+
+/** The configuration for the `multi2vec-nvidia` vectorizer. */
+export type Multi2VecNvidiaConfigCreate = {
+  /** The model to use. Defaults to `None`, which uses the server-defined default. */
+  model?: string;
+  /** The base URL where API requests should go. */
+  baseURL?: string;
+  /** Whether to apply truncation. */
+  truncation?: boolean;
+  /** Format in which the embeddings are encoded. Defaults to `None`, so the embeddings are represented as a list of floating-point numbers. */
+  outputEncoding?: string;
+  /** The image fields to use in vectorization. Can be string of `Multi2VecField` type. If string, weight 0 will be assumed. */
+  imageFields?: string[] | Multi2VecField[];
+  /** The text fields to use in vectorization. Can be string of `Multi2VecField` type. If string, weight 0 will be assumed. */
+  textFields?: string[] | Multi2VecField[];
+};
+
 /** The configuration for the `multi2vec-clip` vectorizer. */
 export type Multi2VecClipConfigCreate = {
   /** The image fields to use in vectorization. Can be string of `Multi2VecField` type. If string, weight 0 will be assumed. */
@@ -261,6 +292,8 @@ export type Text2MultiVecJinaAIConfigCreate = Text2MultiVecJinaAIConfig;
 
 export type VectorizerConfigCreateType<V> = V extends 'img2vec-neural'
   ? Img2VecNeuralConfigCreate | undefined
+  : V extends 'multi2vec-nvidia'
+  ? Multi2VecNvidiaConfigCreate | undefined
   : V extends 'multi2vec-clip'
   ? Multi2VecClipConfigCreate | undefined
   : V extends 'multi2vec-cohere'