@google/genai 1.35.0 → 1.37.0

package/dist/web/web.d.ts

@@ -417,11 +417,17 @@ export declare interface AudioChunk {
  */
 declare interface AudioContent {
     type: 'audio';
+    /**
+     * The audio content.
+     */
     data?: string;
     /**
      * The mime type of the audio.
      */
     mime_type?: AudioMimeType;
+    /**
+     * The URI of the audio.
+     */
     uri?: string;
 }
 
@@ -581,7 +587,7 @@ declare interface BaseCreateAgentInteractionParams {
     /**
      * Body param: The inputs for the interaction.
      */
-    input: string | Array<Content_2> | Array<Turn> | TextContent | ImageContent | AudioContent | DocumentContent | VideoContent | ThoughtContent | FunctionCallContent | FunctionResultContent | CodeExecutionCallContent | CodeExecutionResultContent | URLContextCallContent | URLContextResultContent | GoogleSearchCallContent | GoogleSearchResultContent | MCPServerToolCallContent | MCPServerToolResultContent | FileSearchResultContent;
+    input: string | Array<Content_2> | Array<Turn> | TextContent | ImageContent | AudioContent | DocumentContent | VideoContent | ThoughtContent | FunctionCallContent | FunctionResultContent | CodeExecutionCallContent | CodeExecutionResultContent | URLContextCallContent | URLContextResultContent | GoogleSearchCallContent | GoogleSearchResultContent | MCPServerToolCallContent | MCPServerToolResultContent | FileSearchCallContent | FileSearchResultContent;
     /**
      * Path param: Which version of the API to use.
      */
@@ -633,7 +639,7 @@ declare interface BaseCreateModelInteractionParams {
     /**
      * Body param: The inputs for the interaction.
      */
-    input: string | Array<Content_2> | Array<Turn> | TextContent | ImageContent | AudioContent | DocumentContent | VideoContent | ThoughtContent | FunctionCallContent | FunctionResultContent | CodeExecutionCallContent | CodeExecutionResultContent | URLContextCallContent | URLContextResultContent | GoogleSearchCallContent | GoogleSearchResultContent | MCPServerToolCallContent | MCPServerToolResultContent | FileSearchResultContent;
+    input: string | Array<Content_2> | Array<Turn> | TextContent | ImageContent | AudioContent | DocumentContent | VideoContent | ThoughtContent | FunctionCallContent | FunctionResultContent | CodeExecutionCallContent | CodeExecutionResultContent | URLContextCallContent | URLContextResultContent | GoogleSearchCallContent | GoogleSearchResultContent | MCPServerToolCallContent | MCPServerToolResultContent | FileSearchCallContent | FileSearchResultContent;
     /**
      * Body param: The name of the `Model` used for generating the interaction.
      */
@@ -1701,6 +1707,12 @@ export declare class ComputeTokensResponse {
     tokensInfo?: TokensInfo[];
 }
 
+/** Local tokenizer compute tokens result. */
+export declare interface ComputeTokensResult {
+    /** Lists of tokens info from the input. */
+    tokensInfo?: TokensInfo[];
+}
+
 declare type _ConditionalNodeReadableStream<R = any> = typeof globalThis extends {
     ReadableStream: any;
 } ? never : _NodeReadableStream<R>;
@@ -1720,10 +1732,10 @@ export declare interface Content {
 /**
  * The content of the response.
  */
-declare type Content_2 = TextContent | ImageContent | AudioContent | DocumentContent | VideoContent | ThoughtContent | FunctionCallContent | FunctionResultContent | CodeExecutionCallContent | CodeExecutionResultContent | URLContextCallContent | URLContextResultContent | GoogleSearchCallContent | GoogleSearchResultContent | MCPServerToolCallContent | MCPServerToolResultContent | FileSearchResultContent;
+declare type Content_2 = TextContent | ImageContent | AudioContent | DocumentContent | VideoContent | ThoughtContent | FunctionCallContent | FunctionResultContent | CodeExecutionCallContent | CodeExecutionResultContent | URLContextCallContent | URLContextResultContent | GoogleSearchCallContent | GoogleSearchResultContent | MCPServerToolCallContent | MCPServerToolResultContent | FileSearchCallContent | FileSearchResultContent;
 
 declare interface ContentDelta {
-    delta?: ContentDelta.TextDelta | ContentDelta.ImageDelta | ContentDelta.AudioDelta | ContentDelta.DocumentDelta | ContentDelta.VideoDelta | ContentDelta.ThoughtSummaryDelta | ContentDelta.ThoughtSignatureDelta | ContentDelta.FunctionCallDelta | ContentDelta.FunctionResultDelta | ContentDelta.CodeExecutionCallDelta | ContentDelta.CodeExecutionResultDelta | ContentDelta.URLContextCallDelta | ContentDelta.URLContextResultDelta | ContentDelta.GoogleSearchCallDelta | ContentDelta.GoogleSearchResultDelta | ContentDelta.MCPServerToolCallDelta | ContentDelta.MCPServerToolResultDelta | ContentDelta.FileSearchResultDelta;
+    delta?: ContentDelta.TextDelta | ContentDelta.ImageDelta | ContentDelta.AudioDelta | ContentDelta.DocumentDelta | ContentDelta.VideoDelta | ContentDelta.ThoughtSummaryDelta | ContentDelta.ThoughtSignatureDelta | ContentDelta.FunctionCallDelta | ContentDelta.FunctionResultDelta | ContentDelta.CodeExecutionCallDelta | ContentDelta.CodeExecutionResultDelta | ContentDelta.URLContextCallDelta | ContentDelta.URLContextResultDelta | ContentDelta.GoogleSearchCallDelta | ContentDelta.GoogleSearchResultDelta | ContentDelta.MCPServerToolCallDelta | ContentDelta.MCPServerToolResultDelta | ContentDelta.FileSearchCallDelta | ContentDelta.FileSearchResultDelta;
     /**
      * The event_id token to be used to resume the interaction stream, from this event.
      */
@@ -1921,6 +1933,13 @@ declare namespace ContentDelta {
             items?: Array<string | InteractionsAPI.ImageContent | unknown>;
         }
     }
+    interface FileSearchCallDelta {
+        type: 'file_search_call';
+        /**
+         * A unique ID for this specific tool call.
+         */
+        id?: string;
+    }
     interface FileSearchResultDelta {
         type: 'file_search_result';
         result?: Array<FileSearchResultDelta.Result>;
@@ -2103,6 +2122,12 @@ export declare class CountTokensResponse {
     cachedContentTokenCount?: number;
 }
 
+/** Local tokenizer count tokens result. */
+export declare interface CountTokensResult {
+    /** The total number of tokens. */
+    totalTokens?: number;
+}
+
 declare interface CreateAgentInteractionParamsNonStreaming extends BaseCreateAgentInteractionParams {
     /**
      * Body param: Input only. Whether the interaction will be streamed.
@@ -2754,11 +2779,17 @@ export { Document_2 as Document }
  */
 declare interface DocumentContent {
     type: 'document';
+    /**
+     * The document content.
+     */
     data?: string;
     /**
      * The mime type of the document.
      */
     mime_type?: DocumentMimeType;
+    /**
+     * The URI of the document.
+     */
     uri?: string;
 }
 
@@ -3399,6 +3430,17 @@ export declare interface FileSearch {
     metadataFilter?: string;
 }
 
+/**
+ * File Search content.
+ */
+declare interface FileSearchCallContent {
+    type: 'file_search_call';
+    /**
+     * A unique ID for this specific tool call.
+     */
+    id?: string;
+}
+
 /**
  * File Search result content.
  */
@@ -4560,6 +4602,10 @@ export declare interface GenerationConfig {
  * Configuration parameters for model interactions.
  */
 declare interface GenerationConfig_2 {
+    /**
+     * Configuration for image interaction.
+     */
+    image_config?: ImageConfig_2;
     /**
      * The maximum number of tokens to include in the response.
      */
@@ -5436,11 +5482,22 @@ export declare interface ImageConfig {
     outputCompressionQuality?: number;
 }
 
+/**
+ * The configuration for image interaction.
+ */
+declare interface ImageConfig_2 {
+    aspect_ratio?: '1:1' | '2:3' | '3:2' | '3:4' | '4:3' | '4:5' | '5:4' | '9:16' | '16:9' | '21:9';
+    image_size?: '1K' | '2K' | '4K';
+}
+
 /**
  * An image content block.
  */
 declare interface ImageContent {
     type: 'image';
+    /**
+     * The image content.
+     */
     data?: string;
     /**
      * The mime type of the image.
@@ -5450,6 +5507,9 @@ declare interface ImageContent {
      * The resolution of the media.
      */
     resolution?: 'low' | 'medium' | 'high' | 'ultra_high';
+    /**
+     * The URI of the image.
+     */
     uri?: string;
 }
 
@@ -5705,7 +5765,7 @@ export declare class Interactions extends BaseInteractions {
 }
 
 export declare namespace Interactions {
-    export { type AllowedTools as AllowedTools, type Annotation as Annotation, type AudioContent as AudioContent, type AudioMimeType as AudioMimeType, type CodeExecutionCallArguments as CodeExecutionCallArguments, type CodeExecutionCallContent as CodeExecutionCallContent, type CodeExecutionResultContent as CodeExecutionResultContent, type Content_2 as Content, type ContentDelta as ContentDelta, type ContentStart as ContentStart, type ContentStop as ContentStop, type DeepResearchAgentConfig as DeepResearchAgentConfig, type DocumentContent as DocumentContent, type DocumentMimeType as DocumentMimeType, type DynamicAgentConfig as DynamicAgentConfig, type ErrorEvent_2 as ErrorEvent, type FileSearchResultContent as FileSearchResultContent, type Function_2 as Function, type FunctionCallContent as FunctionCallContent, type FunctionResultContent as FunctionResultContent, type GenerationConfig_2 as GenerationConfig, type GoogleSearchCallArguments as GoogleSearchCallArguments, type GoogleSearchCallContent as GoogleSearchCallContent, type GoogleSearchResult as GoogleSearchResult, type GoogleSearchResultContent as GoogleSearchResultContent, type ImageContent as ImageContent, type ImageMimeType as ImageMimeType, type Interaction as Interaction, type InteractionEvent as InteractionEvent, type InteractionSSEEvent as InteractionSSEEvent, type InteractionStatusUpdate as InteractionStatusUpdate, type MCPServerToolCallContent as MCPServerToolCallContent, type MCPServerToolResultContent as MCPServerToolResultContent, type Model_2 as Model, type SpeechConfig_2 as SpeechConfig, type TextContent as TextContent, type ThinkingLevel_2 as ThinkingLevel, type ThoughtContent as ThoughtContent, type Tool_2 as Tool, type ToolChoice as ToolChoice, type ToolChoiceConfig as ToolChoiceConfig, type ToolChoiceType as ToolChoiceType, type Turn as Turn, type URLContextCallArguments as URLContextCallArguments, type URLContextCallContent as URLContextCallContent, type URLContextResult as URLContextResult, type URLContextResultContent as URLContextResultContent, type Usage as Usage, type VideoContent as VideoContent, type VideoMimeType as VideoMimeType, type InteractionDeleteResponse as InteractionDeleteResponse, type InteractionCreateParams as InteractionCreateParams, type CreateModelInteractionParamsNonStreaming as CreateModelInteractionParamsNonStreaming, type CreateModelInteractionParamsStreaming as CreateModelInteractionParamsStreaming, type CreateAgentInteractionParamsNonStreaming as CreateAgentInteractionParamsNonStreaming, type CreateAgentInteractionParamsStreaming as CreateAgentInteractionParamsStreaming, type InteractionDeleteParams as InteractionDeleteParams, type InteractionCancelParams as InteractionCancelParams, type InteractionGetParams as InteractionGetParams, type InteractionGetParamsNonStreaming as InteractionGetParamsNonStreaming, type InteractionGetParamsStreaming as InteractionGetParamsStreaming, };
+    export { type AllowedTools as AllowedTools, type Annotation as Annotation, type AudioContent as AudioContent, type AudioMimeType as AudioMimeType, type CodeExecutionCallArguments as CodeExecutionCallArguments, type CodeExecutionCallContent as CodeExecutionCallContent, type CodeExecutionResultContent as CodeExecutionResultContent, type Content_2 as Content, type ContentDelta as ContentDelta, type ContentStart as ContentStart, type ContentStop as ContentStop, type DeepResearchAgentConfig as DeepResearchAgentConfig, type DocumentContent as DocumentContent, type DocumentMimeType as DocumentMimeType, type DynamicAgentConfig as DynamicAgentConfig, type ErrorEvent_2 as ErrorEvent, type FileSearchCallContent as FileSearchCallContent, type FileSearchResultContent as FileSearchResultContent, type Function_2 as Function, type FunctionCallContent as FunctionCallContent, type FunctionResultContent as FunctionResultContent, type GenerationConfig_2 as GenerationConfig, type GoogleSearchCallArguments as GoogleSearchCallArguments, type GoogleSearchCallContent as GoogleSearchCallContent, type GoogleSearchResult as GoogleSearchResult, type GoogleSearchResultContent as GoogleSearchResultContent, type ImageConfig_2 as ImageConfig, type ImageContent as ImageContent, type ImageMimeType as ImageMimeType, type Interaction as Interaction, type InteractionEvent as InteractionEvent, type InteractionSSEEvent as InteractionSSEEvent, type InteractionStatusUpdate as InteractionStatusUpdate, type MCPServerToolCallContent as MCPServerToolCallContent, type MCPServerToolResultContent as MCPServerToolResultContent, type Model_2 as Model, type SpeechConfig_2 as SpeechConfig, type TextContent as TextContent, type ThinkingLevel_2 as ThinkingLevel, type ThoughtContent as ThoughtContent, type Tool_2 as Tool, type ToolChoice as ToolChoice, type ToolChoiceConfig as ToolChoiceConfig, type ToolChoiceType as ToolChoiceType, type Turn as Turn, type URLContextCallArguments as URLContextCallArguments, type URLContextCallContent as URLContextCallContent, type URLContextResult as URLContextResult, type URLContextResultContent as URLContextResultContent, type Usage as Usage, type VideoContent as VideoContent, type VideoMimeType as VideoMimeType, type InteractionDeleteResponse as InteractionDeleteResponse, type InteractionCreateParams as InteractionCreateParams, type CreateModelInteractionParamsNonStreaming as CreateModelInteractionParamsNonStreaming, type CreateModelInteractionParamsStreaming as CreateModelInteractionParamsStreaming, type CreateAgentInteractionParamsNonStreaming as CreateAgentInteractionParamsNonStreaming, type CreateAgentInteractionParamsStreaming as CreateAgentInteractionParamsStreaming, type InteractionDeleteParams as InteractionDeleteParams, type InteractionCancelParams as InteractionCancelParams, type InteractionGetParams as InteractionGetParams, type InteractionGetParamsNonStreaming as InteractionGetParamsNonStreaming, type InteractionGetParamsStreaming as InteractionGetParamsStreaming, };
 }
 
 declare namespace InteractionsAPI {
@@ -5728,6 +5788,7 @@ declare namespace InteractionsAPI {
         DocumentMimeType,
         DynamicAgentConfig,
         ErrorEvent_2 as ErrorEvent,
+        FileSearchCallContent,
         FileSearchResultContent,
         Function_2 as Function,
         FunctionCallContent,
@@ -5737,6 +5798,7 @@ declare namespace InteractionsAPI {
         GoogleSearchCallContent,
         GoogleSearchResult,
         GoogleSearchResultContent,
+        ImageConfig_2 as ImageConfig,
         ImageContent,
         ImageMimeType,
         Interaction,
@@ -6810,8 +6872,10 @@ export declare class LiveServerMessage {
     goAway?: LiveServerGoAway;
     /** Update of the session resumption state. */
     sessionResumptionUpdate?: LiveServerSessionResumptionUpdate;
-    /** Voice activity detection signal. */
+    /** Voice activity detection signal. Allowlisted only. */
     voiceActivityDetectionSignal?: VoiceActivityDetectionSignal;
+    /** Voice activity signal. */
+    voiceActivity?: VoiceActivity;
     /**
      * Returns the concatenation of all text parts from the server content if present.
      *
@@ -9828,6 +9892,7 @@ declare namespace types {
         TurnCompleteReason,
         MediaModality,
         VadSignalType,
+        VoiceActivityType,
         StartSensitivity,
         EndSensitivity,
         ActivityHandling,
@@ -10150,6 +10215,7 @@ declare namespace types {
         LiveServerGoAway,
         LiveServerSessionResumptionUpdate,
         VoiceActivityDetectionSignal,
+        VoiceActivity,
         LiveServerMessage,
         OperationFromAPIResponseParameters,
         GenerationConfigThinkingConfig,
@@ -10197,6 +10263,8 @@ declare namespace types {
         CreateAuthTokenConfig,
         CreateAuthTokenParameters,
         OperationGetParameters,
+        CountTokensResult,
+        ComputeTokensResult,
         CreateTuningJobParameters,
         UploadToFileSearchStoreResponse,
         UploadToFileSearchStoreOperation,
@@ -10837,6 +10905,9 @@ export declare enum VideoCompressionQuality {
  */
 declare interface VideoContent {
     type: 'video';
+    /**
+     * The video content.
+     */
     data?: string;
     /**
      * The mime type of the video.
@@ -10846,6 +10917,9 @@ declare interface VideoContent {
      * The resolution of the media.
      */
     resolution?: 'low' | 'medium' | 'high' | 'ultra_high';
+    /**
+     * The URI of the video.
+     */
     uri?: string;
 }
 
@@ -10925,11 +10999,33 @@ export declare interface VideoMetadata {
  */
 declare type VideoMimeType = 'video/mp4' | 'video/mpeg' | 'video/mov' | 'video/avi' | 'video/x-flv' | 'video/mpg' | 'video/webm' | 'video/wmv' | 'video/3gpp' | (string & {});
 
+/** Voice activity signal. */
+export declare interface VoiceActivity {
+    /** The type of the voice activity signal. */
+    voiceActivityType?: VoiceActivityType;
+}
+
 export declare interface VoiceActivityDetectionSignal {
     /** The type of the VAD signal. */
     vadSignalType?: VadSignalType;
 }
 
+/** The type of the voice activity signal. */
+export declare enum VoiceActivityType {
+    /**
+     * The default is VOICE_ACTIVITY_TYPE_UNSPECIFIED.
+     */
+    TYPE_UNSPECIFIED = "TYPE_UNSPECIFIED",
+    /**
+     * Start of sentence signal.
+     */
+    ACTIVITY_START = "ACTIVITY_START",
+    /**
+     * End of sentence signal.
+     */
+    ACTIVITY_END = "ACTIVITY_END"
+}
+
 export declare interface VoiceConfig {
     /** If true, the model will use a replicated voice for the response. */
     replicatedVoiceConfig?: ReplicatedVoiceConfig;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@google/genai",
-  "version": "1.35.0",
+  "version": "1.37.0",
   "description": "",
   "type": "module",
   "main": "dist/node/index.mjs",
@@ -34,6 +34,24 @@
       "types": "./dist/node/node.d.ts",
       "import": "./dist/node/index.mjs",
       "default": "./dist/node/index.mjs"
+    },
+    "./tokenizer": {
+      "node": {
+        "types": "./dist/tokenizer/node.d.ts",
+        "import": "./dist/tokenizer/node.mjs",
+        "require": "./dist/tokenizer/node.cjs",
+        "default": "./dist/tokenizer/node.mjs"
+      },
+      "types": "./dist/tokenizer/node.d.ts",
+      "import": "./dist/tokenizer/node.mjs",
+      "require": "./dist/tokenizer/node.cjs",
+      "default": "./dist/tokenizer/node.mjs"
+    },
+    "./tokenizer/node": {
+      "types": "./dist/tokenizer/node.d.ts",
+      "import": "./dist/tokenizer/node.mjs",
+      "require": "./dist/tokenizer/node.cjs",
+      "default": "./dist/tokenizer/node.mjs"
     }
   },
   "scripts": {
@@ -43,10 +61,12 @@
     "api-extractor:dev:main": "api-extractor run --local --verbose",
     "api-extractor:dev:node": "api-extractor run -c api-extractor.node.json --local --verbose",
     "api-extractor:dev:web": "api-extractor run -c api-extractor.web.json --local --verbose",
+    "api-extractor:dev:tokenizer-node": "api-extractor run -c api-extractor.tokenizer-node.json --local --verbose",
     "api-extractor:prod:main": "api-extractor run --verbose",
     "api-extractor:prod:node": "api-extractor run -c api-extractor.node.json --verbose",
     "api-extractor:prod:web": "api-extractor run -c api-extractor.web.json --verbose",
-    "unit-test": "tsc && jasmine dist/test/unit/**/*_test.js dist/test/unit/*_test.js",
+    "api-extractor:prod:tokenizer-node": "api-extractor run -c api-extractor.tokenizer-node.json --verbose",
+    "unit-test": "tsc && cp src/cross/sentencepiece/sentencepiece_model.pb.js dist/src/cross/sentencepiece/ && jasmine dist/test/unit/**/*_test.js dist/test/unit/**/**/*_test.js dist/test/unit/*_test.js",
     "system-test": "tsc && jasmine dist/test/system/**/*_test.js",
     "test-server-tests": "tsc && GOOGLE_CLOUD_PROJECT=googcloudproj GOOGLE_CLOUD_LOCATION=googcloudloc jasmine dist/test/system/node/*_test.js -- --test-server",
     "test-server-tests:record": "tsc && jasmine --fail-fast dist/test/system/node/*_test.js -- --test-server --record",
@@ -56,7 +76,8 @@
     "format": "prettier '**/*.ts' '**/*.mjs' '**/*.json' --write",
     "lint": "eslint '**/*.ts'",
     "lint-fix": "eslint --fix '**/*.ts'",
-    "coverage-report": "./test/generate_report.sh"
+    "coverage-report": "./test/generate_report.sh",
+    "generate-proto": "pbjs -t static-module -w es6 -o src/cross/sentencepiece/sentencepiece_model.pb.js src/cross/sentencepiece/sentencepiece_model.proto && pbts -o src/cross/sentencepiece/sentencepiece_model.pb.d.ts src/cross/sentencepiece/sentencepiece_model.pb.js && sed -i.bak 's/import \\* as \\$protobuf from \"protobufjs\\/minimal\"/import \\$protobuf from \"protobufjs\\/minimal.js\"/' src/cross/sentencepiece/sentencepiece_model.pb.js && rm src/cross/sentencepiece/sentencepiece_model.pb.js.bak"
   },
   "engines": {
     "node": ">=20.0.0"
@@ -76,13 +97,17 @@
     "dist/web/index.mjs",
     "dist/web/index.mjs.map",
     "dist/web/web.d.ts",
+    "dist/tokenizer/node.mjs",
+    "dist/tokenizer/node.cjs",
+    "dist/tokenizer/node.mjs.map",
+    "dist/tokenizer/node.d.ts",
     "node/package.json",
     "web/package.json"
   ],
   "devDependencies": {
     "@eslint/js": "9.20.0",
     "@microsoft/api-extractor": "^7.52.9",
-    "@modelcontextprotocol/sdk": "^1.24.0",
+    "@modelcontextprotocol/sdk": "^1.25.2",
     "@rollup/plugin-json": "^6.1.0",
     "@types/jasmine": "^5.1.2",
     "@types/node": "^20.9.0",
@@ -101,6 +126,7 @@
     "patch-package": "^8.0.1",
     "prettier": "3.3.3",
     "prettier-plugin-organize-imports": "^4.1.0",
+    "protobufjs-cli": "^1.1.3",
     "rollup-plugin-typescript2": "^0.36.0",
     "test-server-sdk": "^0.2.9",
     "ts-node": "^10.9.2",
@@ -116,10 +142,11 @@
   },
   "dependencies": {
     "google-auth-library": "^10.3.0",
+    "protobufjs": "^7.5.4",
     "ws": "^8.18.0"
   },
   "peerDependencies": {
-    "@modelcontextprotocol/sdk": "^1.24.0"
+    "@modelcontextprotocol/sdk": "^1.25.2"
   },
   "peerDependenciesMeta": {
     "@modelcontextprotocol/sdk": {