Define HybridParams (#8935)

Co-authored-by: Erik Eldridge <[email protected]>

Author: gsiddh

common/api-review/vertexai.api.md

@@ -344,7 +344,7 @@ export class GenerativeModel extends VertexAIModel {
 }
 
 // @public
-export function getGenerativeModel(vertexAI: VertexAI, modelParams: ModelParams, requestOptions?: RequestOptions): GenerativeModel;
+export function getGenerativeModel(vertexAI: VertexAI, onCloudOrHybridParams: ModelParams | HybridParams, requestOptions?: RequestOptions): GenerativeModel;
 
 // @beta
 export function getImagenModel(vertexAI: VertexAI, modelParams: ImagenModelParams, requestOptions?: RequestOptions): ImagenModel;
@@ -416,6 +416,18 @@ export enum HarmSeverity {
     HARM_SEVERITY_NEGLIGIBLE = "HARM_SEVERITY_NEGLIGIBLE"
 }
 
+// @public
+export interface HybridParams {
+    // (undocumented)
+    mode?: InferenceMode;
+    // (undocumented)
+    onCloudParams?: ModelParams;
+    // Warning: (ae-forgotten-export) The symbol "LanguageModelCreateOptions" needs to be exported by the entry point index.d.ts
+    //
+    // (undocumented)
+    onDeviceParams?: LanguageModelCreateOptions;
+}
+
 // @beta
 export enum ImagenAspectRatio {
     LANDSCAPE_16x9 = "16:9",
@@ -500,6 +512,16 @@ export interface ImagenSafetySettings {
     safetyFilterLevel?: ImagenSafetyFilterLevel;
 }
 
+// @public
+export enum InferenceMode {
+    // (undocumented)
+    ONLY_ON_CLOUD = "ONLY_ON_CLOUD",
+    // (undocumented)
+    ONLY_ON_DEVICE = "ONLY_ON_DEVICE",
+    // (undocumented)
+    PREFER_ON_DEVICE = "PREFER_ON_DEVICE"
+}
+
 // @public
 export interface InlineDataPart {
     // (undocumented)

docs-devsite/_toc.yaml

@@ -536,6 +536,8 @@ toc:
     path: /docs/reference/js/vertexai.groundingattribution.md
   - title: GroundingMetadata
     path: /docs/reference/js/vertexai.groundingmetadata.md
+  - title: HybridParams
+    path: /docs/reference/js/vertexai.hybridparams.md
   - title: ImagenGCSImage
     path: /docs/reference/js/vertexai.imagengcsimage.md
   - title: ImagenGenerationConfig

docs-devsite/vertexai.hybridparams.md

@@ -0,0 +1,51 @@
+Project: /docs/reference/js/_project.yaml
+Book: /docs/reference/_book.yaml
+page_type: reference
+
+{% comment %}
+DO NOT EDIT THIS FILE!
+This is generated by the JS SDK team, and any local changes will be
+overwritten. Changes should be made in the source code at
+https://github.com/firebase/firebase-js-sdk
+{% endcomment %}
+
+# HybridParams interface
+Configures on-device and on-cloud inference.
+
+<b>Signature:</b>
+
+```typescript
+export interface HybridParams 
+```
+
+## Properties
+
+|  Property | Type | Description |
+|  --- | --- | --- |
+|  [mode](./vertexai.hybridparams.md#hybridparamsmode) | [InferenceMode](./vertexai.md#inferencemode) |  |
+|  [onCloudParams](./vertexai.hybridparams.md#hybridparamsoncloudparams) | [ModelParams](./vertexai.modelparams.md#modelparams_interface) |  |
+|  [onDeviceParams](./vertexai.hybridparams.md#hybridparamsondeviceparams) | LanguageModelCreateOptions |  |
+
+## HybridParams.mode
+
+<b>Signature:</b>
+
+```typescript
+mode?: InferenceMode;
+```
+
+## HybridParams.onCloudParams
+
+<b>Signature:</b>
+
+```typescript
+onCloudParams?: ModelParams;
+```
+
+## HybridParams.onDeviceParams
+
+<b>Signature:</b>
+
+```typescript
+onDeviceParams?: LanguageModelCreateOptions;
+```

docs-devsite/vertexai.md

@@ -19,7 +19,7 @@ The Vertex AI in Firebase Web SDK.
 |  <b>function(app, ...)</b> |
 |  [getVertexAI(app, options)](./vertexai.md#getvertexai_04094cf) | Returns a [VertexAI](./vertexai.vertexai.md#vertexai_interface) instance for the given app. |
 |  <b>function(vertexAI, ...)</b> |
-|  [getGenerativeModel(vertexAI, modelParams, requestOptions)](./vertexai.md#getgenerativemodel_e3037c9) | Returns a [GenerativeModel](./vertexai.generativemodel.md#generativemodel_class) class with methods for inference and other functionality. |
+|  [getGenerativeModel(vertexAI, onCloudOrHybridParams, requestOptions)](./vertexai.md#getgenerativemodel_202434f) | Returns a [GenerativeModel](./vertexai.generativemodel.md#generativemodel_class) class with methods for inference and other functionality. |
 |  [getImagenModel(vertexAI, modelParams, requestOptions)](./vertexai.md#getimagenmodel_812c375) | <b><i>(Public Preview)</i></b> Returns an [ImagenModel](./vertexai.imagenmodel.md#imagenmodel_class) class with methods for using Imagen.<!-- -->Only Imagen 3 models (named <code>imagen-3.0-*</code>) are supported. |
 
 ## Classes
@@ -55,6 +55,7 @@ The Vertex AI in Firebase Web SDK.
 |  [ImagenAspectRatio](./vertexai.md#imagenaspectratio) | <b><i>(Public Preview)</i></b> Aspect ratios for Imagen images.<!-- -->To specify an aspect ratio for generated images, set the <code>aspectRatio</code> property in your [ImagenGenerationConfig](./vertexai.imagengenerationconfig.md#imagengenerationconfig_interface)<!-- -->.<!-- -->See the the [documentation](http://firebase.google.com/docs/vertex-ai/generate-images) for more details and examples of the supported aspect ratios. |
 |  [ImagenPersonFilterLevel](./vertexai.md#imagenpersonfilterlevel) | <b><i>(Public Preview)</i></b> A filter level controlling whether generation of images containing people or faces is allowed.<!-- -->See the <a href="http://firebase.google.com/docs/vertex-ai/generate-images">personGeneration</a> documentation for more details. |
 |  [ImagenSafetyFilterLevel](./vertexai.md#imagensafetyfilterlevel) | <b><i>(Public Preview)</i></b> A filter level controlling how aggressively to filter sensitive content.<!-- -->Text prompts provided as inputs and images (generated or uploaded) through Imagen on Vertex AI are assessed against a list of safety filters, which include 'harmful categories' (for example, <code>violence</code>, <code>sexual</code>, <code>derogatory</code>, and <code>toxic</code>). This filter level controls how aggressively to filter out potentially harmful content from responses. See the [documentation](http://firebase.google.com/docs/vertex-ai/generate-images) and the [Responsible AI and usage guidelines](https://cloud.google.com/vertex-ai/generative-ai/docs/image/responsible-ai-imagen#safety-filters) for more details. |
+|  [InferenceMode](./vertexai.md#inferencemode) | Determines whether inference happens on-device or on-cloud. |
 |  [Modality](./vertexai.md#modality) | Content part modality. |
 |  [SchemaType](./vertexai.md#schematype) | Contains the list of OpenAPI data types as defined by the [OpenAPI specification](https://swagger.io/docs/specification/data-models/data-types/) |
 |  [VertexAIErrorCode](./vertexai.md#vertexaierrorcode) | Standardized error codes that [VertexAIError](./vertexai.vertexaierror.md#vertexaierror_class) can have. |
@@ -91,6 +92,7 @@ The Vertex AI in Firebase Web SDK.
 |  [GenerativeContentBlob](./vertexai.generativecontentblob.md#generativecontentblob_interface) | Interface for sending an image. |
 |  [GroundingAttribution](./vertexai.groundingattribution.md#groundingattribution_interface) |  |
 |  [GroundingMetadata](./vertexai.groundingmetadata.md#groundingmetadata_interface) | Metadata returned to client when grounding is enabled. |
+|  [HybridParams](./vertexai.hybridparams.md#hybridparams_interface) | Configures on-device and on-cloud inference. |
 |  [ImagenGCSImage](./vertexai.imagengcsimage.md#imagengcsimage_interface) | An image generated by Imagen, stored in a Cloud Storage for Firebase bucket.<!-- -->This feature is not available yet. |
 |  [ImagenGenerationConfig](./vertexai.imagengenerationconfig.md#imagengenerationconfig_interface) | <b><i>(Public Preview)</i></b> Configuration options for generating images with Imagen.<!-- -->See the [documentation](http://firebase.google.com/docs/vertex-ai/generate-images-imagen) for more details. |
 |  [ImagenGenerationResponse](./vertexai.imagengenerationresponse.md#imagengenerationresponse_interface) | <b><i>(Public Preview)</i></b> The response from a request to generate images with Imagen. |
@@ -99,10 +101,10 @@ The Vertex AI in Firebase Web SDK.
 |  [ImagenSafetySettings](./vertexai.imagensafetysettings.md#imagensafetysettings_interface) | <b><i>(Public Preview)</i></b> Settings for controlling the aggressiveness of filtering out sensitive content.<!-- -->See the [documentation](http://firebase.google.com/docs/vertex-ai/generate-images) for more details. |
 |  [InlineDataPart](./vertexai.inlinedatapart.md#inlinedatapart_interface) | Content part interface if the part represents an image. |
 |  [ModalityTokenCount](./vertexai.modalitytokencount.md#modalitytokencount_interface) | Represents token counting info for a single modality. |
-|  [ModelParams](./vertexai.modelparams.md#modelparams_interface) | Params passed to [getGenerativeModel()](./vertexai.md#getgenerativemodel_e3037c9)<!-- -->. |
+|  [ModelParams](./vertexai.modelparams.md#modelparams_interface) | Params passed to [getGenerativeModel()](./vertexai.md#getgenerativemodel_202434f)<!-- -->. |
 |  [ObjectSchemaInterface](./vertexai.objectschemainterface.md#objectschemainterface_interface) | Interface for [ObjectSchema](./vertexai.objectschema.md#objectschema_class) class. |
 |  [PromptFeedback](./vertexai.promptfeedback.md#promptfeedback_interface) | If the prompt was blocked, this will be populated with <code>blockReason</code> and the relevant <code>safetyRatings</code>. |
-|  [RequestOptions](./vertexai.requestoptions.md#requestoptions_interface) | Params passed to [getGenerativeModel()](./vertexai.md#getgenerativemodel_e3037c9)<!-- -->. |
+|  [RequestOptions](./vertexai.requestoptions.md#requestoptions_interface) | Params passed to [getGenerativeModel()](./vertexai.md#getgenerativemodel_202434f)<!-- -->. |
 |  [RetrievedContextAttribution](./vertexai.retrievedcontextattribution.md#retrievedcontextattribution_interface) |  |
 |  [SafetyRating](./vertexai.safetyrating.md#safetyrating_interface) | A safety rating associated with a [GenerateContentCandidate](./vertexai.generatecontentcandidate.md#generatecontentcandidate_interface) |
 |  [SafetySetting](./vertexai.safetysetting.md#safetysetting_interface) | Safety setting that can be sent as part of request parameters. |
@@ -160,22 +162,22 @@ export declare function getVertexAI(app?: FirebaseApp, options?: VertexAIOptions
 
 ## function(vertexAI, ...)
 
-### getGenerativeModel(vertexAI, modelParams, requestOptions) {:#getgenerativemodel_e3037c9}
+### getGenerativeModel(vertexAI, onCloudOrHybridParams, requestOptions) {:#getgenerativemodel_202434f}
 
 Returns a [GenerativeModel](./vertexai.generativemodel.md#generativemodel_class) class with methods for inference and other functionality.
 
 <b>Signature:</b>
 
 ```typescript
-export declare function getGenerativeModel(vertexAI: VertexAI, modelParams: ModelParams, requestOptions?: RequestOptions): GenerativeModel;
+export declare function getGenerativeModel(vertexAI: VertexAI, onCloudOrHybridParams: ModelParams | HybridParams, requestOptions?: RequestOptions): GenerativeModel;
 ```
 
 #### Parameters
 
 |  Parameter | Type | Description |
 |  --- | --- | --- |
 |  vertexAI | [VertexAI](./vertexai.vertexai.md#vertexai_interface) |  |
-|  modelParams | [ModelParams](./vertexai.modelparams.md#modelparams_interface) |  |
+|  onCloudOrHybridParams | [ModelParams](./vertexai.modelparams.md#modelparams_interface) \| [HybridParams](./vertexai.hybridparams.md#hybridparams_interface) |  |
 |  requestOptions | [RequestOptions](./vertexai.requestoptions.md#requestoptions_interface) |  |
 
 <b>Returns:</b>
@@ -489,6 +491,24 @@ export declare enum ImagenSafetyFilterLevel
 |  BLOCK\_NONE | <code>&quot;block_none&quot;</code> | <b><i>(Public Preview)</i></b> The least aggressive filtering level; blocks very few sensitive prompts and responses.<!-- -->Access to this feature is restricted and may require your case to be reviewed and approved by Cloud support. |
 |  BLOCK\_ONLY\_HIGH | <code>&quot;block_only_high&quot;</code> | <b><i>(Public Preview)</i></b> Blocks few sensitive prompts and responses. |
 
+## InferenceMode
+
+Determines whether inference happens on-device or on-cloud.
+
+<b>Signature:</b>
+
+```typescript
+export declare enum InferenceMode 
+```
+
+## Enumeration Members
+
+|  Member | Value | Description |
+|  --- | --- | --- |
+|  ONLY\_ON\_CLOUD | <code>&quot;ONLY_ON_CLOUD&quot;</code> |  |
+|  ONLY\_ON\_DEVICE | <code>&quot;ONLY_ON_DEVICE&quot;</code> |  |
+|  PREFER\_ON\_DEVICE | <code>&quot;PREFER_ON_DEVICE&quot;</code> |  |
+
 ## Modality
 
 Content part modality.

docs-devsite/vertexai.modelparams.md

@@ -10,7 +10,7 @@ https://github.com/firebase/firebase-js-sdk
 {% endcomment %}
 
 # ModelParams interface
-Params passed to [getGenerativeModel()](./vertexai.md#getgenerativemodel_e3037c9)<!-- -->.
+Params passed to [getGenerativeModel()](./vertexai.md#getgenerativemodel_202434f)<!-- -->.
 
 <b>Signature:</b>
 

docs-devsite/vertexai.requestoptions.md

@@ -10,7 +10,7 @@ https://github.com/firebase/firebase-js-sdk
 {% endcomment %}
 
 # RequestOptions interface
-Params passed to [getGenerativeModel()](./vertexai.md#getgenerativemodel_e3037c9)<!-- -->.
+Params passed to [getGenerativeModel()](./vertexai.md#getgenerativemodel_202434f)<!-- -->.
 
 <b>Signature:</b>
 

packages/vertexai/src/api.ts

@@ -23,6 +23,7 @@ import { VertexAIService } from './service';
 import { VertexAI, VertexAIOptions } from './public-types';
 import {
   ImagenModelParams,
+  HybridParams,
   ModelParams,
   RequestOptions,
   VertexAIErrorCode
@@ -70,16 +71,27 @@ export function getVertexAI(
  */
 export function getGenerativeModel(
   vertexAI: VertexAI,
-  modelParams: ModelParams,
+  onCloudOrHybridParams: ModelParams | HybridParams,
   requestOptions?: RequestOptions
 ): GenerativeModel {
-  if (!modelParams.model) {
+  // Disambiguates onCloudOrHybridParams input.
+  const hybridParams = onCloudOrHybridParams as HybridParams;
+  let onCloudParams: ModelParams;
+  if (hybridParams.mode) {
+    onCloudParams = hybridParams.onCloudParams || {
+      model: 'gemini-2.0-flash-lite'
+    };
+  } else {
+    onCloudParams = onCloudOrHybridParams as ModelParams;
+  }
+
+  if (!onCloudParams.model) {
     throw new VertexAIError(
       VertexAIErrorCode.NO_MODEL,
       `Must provide a model name. Example: getGenerativeModel({ model: 'my-model-name' })`
     );
   }
-  return new GenerativeModel(vertexAI, modelParams, requestOptions);
+  return new GenerativeModel(vertexAI, onCloudParams, requestOptions);
 }
 
 /**

packages/vertexai/src/types/enums.ts

@@ -240,3 +240,13 @@ export enum Modality {
    */
   DOCUMENT = 'DOCUMENT'
 }
+
+/**
+ * Determines whether inference happens on-device or on-cloud.
+ * @public
+ */
+export enum InferenceMode {
+  PREFER_ON_DEVICE = 'PREFER_ON_DEVICE',
+  ONLY_ON_DEVICE = 'ONLY_ON_DEVICE',
+  ONLY_ON_CLOUD = 'ONLY_ON_CLOUD'
+}

packages/vertexai/src/types/language-model.ts

@@ -38,12 +38,13 @@ enum Availability {
   'downloading',
   'available'
 }
-interface LanguageModelCreateCoreOptions {
+export interface LanguageModelCreateCoreOptions {
   topK?: number;
   temperature?: number;
   expectedInputs?: LanguageModelExpectedInput[];
 }
-interface LanguageModelCreateOptions extends LanguageModelCreateCoreOptions {
+export interface LanguageModelCreateOptions
+  extends LanguageModelCreateCoreOptions {
   signal?: AbortSignal;
   systemPrompt?: string;
   initialPrompts?: LanguageModelInitialPrompts;

packages/vertexai/src/types/requests.ts

@@ -17,11 +17,13 @@
 
 import { TypedSchema } from '../requests/schema-builder';
 import { Content, Part } from './content';
+import { LanguageModelCreateOptions } from './language-model';
 import {
   FunctionCallingMode,
   HarmBlockMethod,
   HarmBlockThreshold,
-  HarmCategory
+  HarmCategory,
+  InferenceMode
 } from './enums';
 import { ObjectSchemaInterface, SchemaRequest } from './schema';
 
@@ -213,3 +215,13 @@ export interface FunctionCallingConfig {
   mode?: FunctionCallingMode;
   allowedFunctionNames?: string[];
 }
+
+/**
+ * Configures on-device and on-cloud inference.
+ * @public
+ */
+export interface HybridParams {
+  mode?: InferenceMode;
+  onDeviceParams?: LanguageModelCreateOptions;
+  onCloudParams?: ModelParams;
+}