Kg vectorize support (#38)

* embeddingApiKey support

* added new vectroize parameters support

* updated api report

* added a couple small tests

* added some vectorize tests

* updated env.example

* linting fixes

* majorly updated vectorize tests

* fixed couple breaking tests

* updated build report

.env.example

@@ -1,14 +1,18 @@
+################################################################################
+# READ THE DEVGUIDE.MD FILE FOR MORE INFORMATION ON HOW TO CONFIGURE THIS FILE #
+################################################################################
+
 # Astra API endpoint
 ASTRA_URI=https://<db_id>-<region>.apps.astra.datastax.com
 
 # Application token, used to authenticate with the Astra API
 APPLICATION_TOKEN=AstraCS:<rest_of_token>
 
 # Set this to some value to enable running tests that require a $vectorize enabled environment
-ASTRA_RUN_VECTORIZE_TESTS=
+ASTRA_RUN_VECTORIZE_TESTS=1
 
 # Set this to some value to enable running long-running tests
-ASTRA_RUN_LONG_TESTS=
+ASTRA_RUN_LONG_TESTS=1
 
 # Set this to some value to enable running admin tests
 ASTRA_RUN_ADMIN_TESTS=

.gitignore

@@ -136,3 +136,5 @@ build.zip
 
 temp
 tsdoc-metadata.json
+
+tests/vectorize_credentials.json

DEVGUIDE.md

@@ -72,6 +72,33 @@ If a new tag really, really, needs to be added, it can be done by adding a new e
 format, and updating the `assertTestsEnabled` function. However, this should be done sparingly, as it can make the
 test suite harder to manage.
 
+### Running vectorize tests
+To run vectorize tests, you need to have a vectorize-enabled kube running, with the correct tags enabled.
+You must create a file, `tests/vectorize_tests.json`, with the following format:
+
+```ts
+interface Config {
+  [providerName: string]: {
+    apiKey?: string,
+    providerKey?: string,
+    parameters?: {
+      [modelName: string]: Record<string, string>
+    },
+  }
+}
+```
+
+where:
+- `providerName` is the name of the provider (e.g. `nvidia`, `openai`, etc.) as found in `findEmbeddingProviders`
+- `apiKey` is the API key for the provider (which will be passed in through the header) 
+  - optional if no header auth test wanted
+- `providerKey` is the provider key for the provider (which will be passed in @ collection creation) 
+  - optional if no KMS auth test wanted
+- `parameters` is a mapping of model names to their corresponding parameters
+  - optional if not required. `azureOpenAI`, for example, will need this.
+
+This file is gitignored by default and will not be checked into VCS.
+
 ### Coverage testing
 
 To run coverage testing, run the following command:

etc/astra-db-ts.api.md

@@ -175,9 +175,10 @@ export type Caller = [name: string, version?: string];
 // @public
 export class Collection<Schema extends SomeDoc = SomeDoc> {
     // Warning: (ae-forgotten-export) The symbol "DataAPIHttpClient" needs to be exported by the entry point index.d.ts
+    // Warning: (ae-forgotten-export) The symbol "CollectionSpawnOptions" needs to be exported by the entry point index.d.ts
     //
     // @internal
-    constructor(db: Db, httpClient: DataAPIHttpClient, name: string, namespace: string | undefined);
+    constructor(db: Db, httpClient: DataAPIHttpClient, name: string, opts: CollectionSpawnOptions | undefined);
     bulkWrite(operations: AnyBulkWriteOperation<Schema>[], options?: BulkWriteOptions): Promise<BulkWriteResult<Schema>>;
     readonly collectionName: string;
     countDocuments(filter: Filter<Schema>, upperBound: number, options?: WithTimeout): Promise<number>;
@@ -287,7 +288,7 @@ export interface CostInfo {
 }
 
 // @public
-export interface CreateCollectionOptions<Schema extends SomeDoc> extends WithTimeout, CollectionOptions<Schema>, WithNamespace {
+export interface CreateCollectionOptions<Schema extends SomeDoc> extends WithTimeout, CollectionOptions<Schema>, CollectionSpawnOptions {
     checkExists?: boolean;
 }
 
@@ -482,7 +483,7 @@ export class Db {
     // @internal
     constructor(endpoint: string, options: InternalRootClientOpts);
     admin(options?: AdminSpawnOptions): AstraDbAdmin;
-    collection<Schema extends SomeDoc = SomeDoc>(name: string, options?: WithNamespace): Collection<Schema>;
+    collection<Schema extends SomeDoc = SomeDoc>(name: string, options?: CollectionSpawnOptions): Collection<Schema>;
     collections(options?: WithNamespace & WithTimeout): Promise<Collection[]>;
     command(command: Record<string, any>, options?: RunCommandOptions): Promise<RawDataAPIResponse>;
     createCollection<Schema extends SomeDoc = SomeDoc>(collectionName: string, options?: CreateCollectionOptions<Schema>): Promise<Collection<Schema>>;
@@ -962,8 +963,9 @@ export interface ReplaceOneOptions extends WithTimeout {
 export type ReplaceOneResult<Schema extends SomeDoc> = InternalUpdateResult<Schema, 0 | 1>;
 
 // @public
-export interface RunCommandOptions extends WithNamespace, WithTimeout {
+export interface RunCommandOptions extends WithTimeout {
     collection?: string;
+    namespace?: string | null;
 }
 
 // @public
@@ -1170,7 +1172,9 @@ export interface VectorDoc {
 
 // @alpha
 export interface VectorizeServiceOptions {
+    authentication?: Record<string, string | undefined>;
     modelName: string;
+    parameters?: Record<string, unknown>;
     provider: string;
 }
 

src/api/clients/data-api-http-client.ts

@@ -15,6 +15,7 @@
 
 import {
   DEFAULT_DATA_API_AUTH_HEADER,
+  DEFAULT_EMBEDDING_API_KEY_HEADER,
   DEFAULT_NAMESPACE,
   DEFAULT_TIMEOUT,
   hrTimeMs,
@@ -23,7 +24,7 @@ import {
   HttpMethods,
   RawDataAPIResponse,
 } from '@/src/api';
-import { DataAPIResponseError, DataAPITimeoutError, ObjectId, UUID, WithNamespace } from '@/src/data-api';
+import { DataAPIResponseError, DataAPITimeoutError, ObjectId, UUID } from '@/src/data-api';
 import { TimeoutManager, TimeoutOptions } from '@/src/api/timeout-managers';
 import { CommandFailedEvent, CommandStartedEvent, CommandSucceededEvent } from '@/src/data-api/events';
 import { CollectionNotFoundError, DataAPIHttpError, mkRespErrorFromResponse } from '@/src/data-api/errors';
@@ -32,16 +33,20 @@ import { CollectionNotFoundError, DataAPIHttpError, mkRespErrorFromResponse } fr
  * @internal
  */
 export interface DataAPIRequestInfo {
-  url: string;
-  collection?: string;
-  namespace?: string;
-  command: Record<string, any>;
-  timeoutManager: TimeoutManager;
+  url: string,
+  collection?: string,
+  namespace?: string | null,
+  command: Record<string, any>,
+  timeoutManager: TimeoutManager,
 }
 
 interface ExecuteCommandOptions {
-  collection?: string;
-  namespace?: string;
+  namespace?: string | null,
+  collection?: string,
+}
+
+interface DataAPIHttpClientOptions extends HTTPClientOptions {
+  namespace: string | undefined,
 }
 
 /**
@@ -50,16 +55,16 @@ interface ExecuteCommandOptions {
 export class DataAPIHttpClient extends HttpClient {
   public collection?: string;
   public namespace?: string;
-  readonly #props: HTTPClientOptions & WithNamespace;
+  readonly #props: DataAPIHttpClientOptions;
 
-  constructor(props: HTTPClientOptions & WithNamespace) {
-    super(props, mkAuthHeader);
+  constructor(props: DataAPIHttpClientOptions, embeddingApiKey?: string) {
+    super(props, mkHeaders(embeddingApiKey));
     this.namespace = props.namespace;
     this.#props = props;
   }
 
-  public withCollection(namespace: string, collection: string): DataAPIHttpClient {
-    const clone = new DataAPIHttpClient(this.#props);
+  public forCollection(namespace: string, collection: string, embeddingApiKey: string | undefined): DataAPIHttpClient {
+    const clone = new DataAPIHttpClient(this.#props, embeddingApiKey);
     clone.collection = collection;
     clone.namespace = namespace;
     return clone;
@@ -86,9 +91,12 @@ export class DataAPIHttpClient extends HttpClient {
 
     try {
       info.collection ||= this.collection;
-      info.namespace ||= this.namespace || DEFAULT_NAMESPACE;
 
-      const keyspacePath = `/${info.namespace}`;
+      if (info.namespace !== null) {
+        info.namespace ||= this.namespace || DEFAULT_NAMESPACE;
+      }
+
+      const keyspacePath = info.namespace ? `/${info.namespace}` : '';
       const collectionPath = info.collection ? `/${info.collection}` : '';
       info.url += keyspacePath + collectionPath;
 
@@ -117,7 +125,7 @@ export class DataAPIHttpClient extends HttpClient {
 
       if (data.errors && data?.errors?.length > 0 && data?.errors[0]?.errorCode === 'COLLECTION_NOT_EXIST') {
         const name = data?.errors[0]?.message.split(': ')[1];
-        throw new CollectionNotFoundError(info.namespace, name);
+        throw new CollectionNotFoundError(info.namespace!, name);
       }
 
       if (data?.errors && data?.errors.length > 0) {
@@ -193,6 +201,15 @@ export function reviver(_: string, value: any): any {
   return value;
 }
 
-function mkAuthHeader(token: string): Record<string, any> {
-  return { [DEFAULT_DATA_API_AUTH_HEADER]: token };
+function mkHeaders(embeddingApiKey: string | undefined) {
+  if (embeddingApiKey) {
+    return (token: string) => ({
+      [DEFAULT_EMBEDDING_API_KEY_HEADER]: embeddingApiKey,
+      [DEFAULT_DATA_API_AUTH_HEADER]: token,
+    });
+  } else {
+    return (token: string) => ({
+      [DEFAULT_DATA_API_AUTH_HEADER]: token,
+    });
+  }
 }

src/api/clients/devops-api-http-client.ts

@@ -55,7 +55,7 @@ interface DevopsAPIResponse {
  */
 export class DevOpsAPIHttpClient extends HttpClient {
   constructor(opts: HTTPClientOptions) {
-    super(opts, mkAuthHeader);
+    super(opts, mkHeaders);
   }
 
   public async request(req: DevOpsAPIRequestInfo, options: TimeoutOptions | undefined, started: number = 0): Promise<DevopsAPIResponse> {
@@ -187,6 +187,6 @@ export class DevOpsAPIHttpClient extends HttpClient {
   }
 }
 
-function mkAuthHeader(token: string) {
+function mkHeaders(token: string) {
   return { [DEFAULT_DEVOPS_API_AUTH_HEADER]: `Bearer ${token}` };
 }

src/api/clients/http-client.ts

@@ -16,7 +16,7 @@ import { CLIENT_USER_AGENT, RAGSTACK_REQUESTED_WITH } from '@/src/api/constants'
 import { Caller, DataAPIClientEvents } from '@/src/client';
 import TypedEmitter from 'typed-emitter';
 import { FetchCtx, ResponseInfo } from '@/src/api/fetch/types';
-import { AuthHeaderFactory, HTTPClientOptions, HTTPRequestInfo } from '@/src/api/clients/types';
+import { MkBaseHeaders, HTTPClientOptions, HTTPRequestInfo } from '@/src/api/clients/types';
 
 /**
  * @internal
@@ -29,7 +29,7 @@ export abstract class HttpClient {
   readonly #applicationToken: string;
   readonly baseHeaders: Record<string, any>;
 
-  protected constructor(options: HTTPClientOptions, mkAuthHeader: AuthHeaderFactory) {
+  protected constructor(options: HTTPClientOptions, mkAuthHeader: MkBaseHeaders) {
     this.#applicationToken = options.applicationToken;
     this.baseUrl = options.baseUrl;
     this.emitter = options.emitter;

src/api/clients/types.ts

@@ -18,7 +18,7 @@ export interface HTTPClientOptions {
 /**
  * @internal
  */
-export type AuthHeaderFactory = (token: string) => Record<string, any>;
+export type MkBaseHeaders = (token: string) => Record<string, any>;
 
 /**
  * @internal

src/api/constants.ts

@@ -53,6 +53,11 @@ export const DEFAULT_NAMESPACE = 'default_keyspace';
  */
 export const DEFAULT_TIMEOUT = 30000;
 
+/**
+ * @internal
+ */
+export const DEFAULT_EMBEDDING_API_KEY_HEADER = 'x-embedding-api-key';
+
 /**
  * @internal
  */

src/data-api/collection.ts

@@ -75,6 +75,7 @@ import { FindOneAndDeleteCommand } from '@/src/data-api/types/find/find-one-dele
 import { FindOneAndUpdateCommand } from '@/src/data-api/types/find/find-one-update';
 import { InsertManyCommand } from '@/src/data-api/types/insert/insert-many';
 import { Mutable } from '@/src/data-api/types/utils';
+import { CollectionSpawnOptions } from '@/src/data-api/types/collections/spawn-collection';
 
 /**
  * Represents the interface to a collection in the database.
@@ -118,19 +119,19 @@ export class Collection<Schema extends SomeDoc = SomeDoc> {
    *
    * @internal
    */
-  constructor(db: Db, httpClient: DataAPIHttpClient, name: string, namespace: string | undefined) {
+  constructor(db: Db, httpClient: DataAPIHttpClient, name: string, opts: CollectionSpawnOptions | undefined) {
     Object.defineProperty(this, 'collectionName', {
       value: name,
       writable: false,
     });
 
     Object.defineProperty(this, 'namespace', {
-      value: namespace ?? db.namespace,
+      value: opts?.namespace ?? db.namespace,
       writable: false,
     });
 
     Object.defineProperty(this, '_httpClient', {
-      value: httpClient.withCollection(this.namespace, this.collectionName),
+      value: httpClient.forCollection(this.namespace, this.collectionName, opts?.embeddingApiKey),
       enumerable: false,
     });
 

src/data-api/db.ts

@@ -30,6 +30,7 @@ import { extractDbIdFromUrl, validateOption } from '@/src/data-api/utils';
 import { CreateCollectionCommand } from '@/src/data-api/types/collections/create-collection';
 import { ListCollectionsCommand } from '@/src/data-api/types/collections/list-collection';
 import { InternalRootClientOpts } from '@/src/client/types';
+import { CollectionSpawnOptions } from '@/src/data-api/types/collections/spawn-collection';
 
 /**
  * Represents an interface to some Astra database instance. This is the entrypoint for database-level DML, such as
@@ -244,8 +245,8 @@ export class Db {
    * @see SomeDoc
    * @see VectorDoc
    */
-  public collection<Schema extends SomeDoc = SomeDoc>(name: string, options?: WithNamespace): Collection<Schema> {
-    return new Collection<Schema>(this, this._httpClient, name, options?.namespace);
+  public collection<Schema extends SomeDoc = SomeDoc>(name: string, options?: CollectionSpawnOptions): Collection<Schema> {
+    return new Collection<Schema>(this, this._httpClient, name, options);
   }
 
   /**

src/data-api/types/collections/collections-common.ts

@@ -62,6 +62,18 @@ export interface VectorizeServiceOptions {
    * @alpha
    */
   modelName: string,
+  /**
+   * NOTE: This feature is under current development.
+   *
+   * @alpha
+   */
+  authentication?: Record<string, string | undefined>,
+  /**
+   * NOTE: This feature is under current development.
+   *
+   * @alpha
+   */
+  parameters?: Record<string, unknown>,
 }
 
 /**

src/data-api/types/collections/command.ts

@@ -12,7 +12,6 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-import { WithNamespace } from '@/src/data-api';
 import { WithTimeout } from '@/src/common';
 
 /**
@@ -25,9 +24,13 @@ import { WithTimeout } from '@/src/common';
  *
  * @public
  */
-export interface RunCommandOptions extends WithNamespace, WithTimeout {
+export interface RunCommandOptions extends WithTimeout {
   /**
    * The collection to run the command on. If not provided, the command is run on the database.
    */
-  collection?: string
+  collection?: string,
+  /**
+   * The namespace (aka keyspace) to use for the db operation.
+   */
+  namespace?: string | null,
 }

src/data-api/types/collections/create-collection.ts

@@ -12,9 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-import { SomeDoc, WithNamespace } from '@/src/data-api';
+import { SomeDoc } from '@/src/data-api';
 import { CollectionOptions } from '@/src/data-api/types';
 import { WithTimeout } from '@/src/common/types';
+import { CollectionSpawnOptions } from '@/src/data-api/types/collections/spawn-collection';
 
 /** @internal */
 export interface CreateCollectionCommand {
@@ -38,7 +39,7 @@ export interface CreateCollectionCommand {
  *
  * @public
  */
-export interface CreateCollectionOptions<Schema extends SomeDoc> extends WithTimeout, CollectionOptions<Schema>, WithNamespace {
+export interface CreateCollectionOptions<Schema extends SomeDoc> extends WithTimeout, CollectionOptions<Schema>, CollectionSpawnOptions {
   /**
    * If `true` or unset, runs an additional existence check before creating the collection, failing if the collection
    * with the same name already exists, raising a {@link CollectionAlreadyExistsError}.