@elevenlabs/elevenlabs-js 2.39.0 → 2.41.0

package/BaseClient.js

@@ -41,8 +41,8 @@ function normalizeClientOptions(options) {
     const headers = (0, headers_1.mergeHeaders)({
         "X-Fern-Language": "JavaScript",
         "X-Fern-SDK-Name": "@elevenlabs/elevenlabs-js",
-        "X-Fern-SDK-Version": "v2.39.0",
-        "User-Agent": "@elevenlabs/elevenlabs-js/v2.39.0",
+        "X-Fern-SDK-Version": "v2.41.0",
+        "User-Agent": "@elevenlabs/elevenlabs-js/v2.41.0",
         "X-Fern-Runtime": core.RUNTIME.type,
         "X-Fern-Runtime-Version": core.RUNTIME.version,
         "xi-api-key": options === null || options === void 0 ? void 0 : options.apiKey,

package/Client.d.ts

@@ -2,6 +2,7 @@ import { AudioIsolationClient } from "./api/resources/audioIsolation/client/Clie
 import { AudioNativeClient } from "./api/resources/audioNative/client/Client";
 import { ConversationalAiClient } from "./api/resources/conversationalAi/client/Client";
 import { DubbingClient } from "./api/resources/dubbing/client/Client";
+import { EnvironmentVariablesClient } from "./api/resources/environmentVariables/client/Client";
 import { ForcedAlignmentClient } from "./api/resources/forcedAlignment/client/Client";
 import { HistoryClient } from "./api/resources/history/client/Client";
 import { ModelsClient } from "./api/resources/models/client/Client";
@@ -43,6 +44,7 @@ export declare class ElevenLabsClient {
     protected _user: UserClient | undefined;
     protected _voices: VoicesClient | undefined;
     protected _studio: StudioClient | undefined;
+    protected _music: MusicClient | undefined;
     protected _dubbing: DubbingClient | undefined;
     protected _models: ModelsClient | undefined;
     protected _audioNative: AudioNativeClient | undefined;
@@ -53,7 +55,7 @@ export declare class ElevenLabsClient {
     protected _speechToText: SpeechToTextClient | undefined;
     protected _forcedAlignment: ForcedAlignmentClient | undefined;
     protected _conversationalAi: ConversationalAiClient | undefined;
-    protected _music: MusicClient | undefined;
+    protected _environmentVariables: EnvironmentVariablesClient | undefined;
     protected _tokens: TokensClient | undefined;
     protected _workspace: WorkspaceClient | undefined;
     constructor(options?: ElevenLabsClient.Options);
@@ -68,6 +70,7 @@ export declare class ElevenLabsClient {
     get user(): UserClient;
     get voices(): VoicesClient;
     get studio(): StudioClient;
+    get music(): MusicClient;
     get dubbing(): DubbingClient;
     get models(): ModelsClient;
     get audioNative(): AudioNativeClient;
@@ -78,7 +81,7 @@ export declare class ElevenLabsClient {
     get speechToText(): SpeechToTextClient;
     get forcedAlignment(): ForcedAlignmentClient;
     get conversationalAi(): ConversationalAiClient;
-    get music(): MusicClient;
+    get environmentVariables(): EnvironmentVariablesClient;
     get tokens(): TokensClient;
     get workspace(): WorkspaceClient;
     /**

package/Client.js

@@ -48,26 +48,27 @@ const Client_1 = require("./api/resources/audioIsolation/client/Client");
 const Client_2 = require("./api/resources/audioNative/client/Client");
 const Client_3 = require("./api/resources/conversationalAi/client/Client");
 const Client_4 = require("./api/resources/dubbing/client/Client");
-const Client_5 = require("./api/resources/forcedAlignment/client/Client");
-const Client_6 = require("./api/resources/history/client/Client");
-const Client_7 = require("./api/resources/models/client/Client");
-const Client_8 = require("./api/resources/music/client/Client");
-const Client_9 = require("./api/resources/pronunciationDictionaries/client/Client");
-const Client_10 = require("./api/resources/samples/client/Client");
-const Client_11 = require("./api/resources/serviceAccounts/client/Client");
-const Client_12 = require("./api/resources/speechToSpeech/client/Client");
-const Client_13 = require("./api/resources/speechToText/client/Client");
-const Client_14 = require("./api/resources/studio/client/Client");
-const Client_15 = require("./api/resources/textToDialogue/client/Client");
-const Client_16 = require("./api/resources/textToSoundEffects/client/Client");
-const Client_17 = require("./api/resources/textToSpeech/client/Client");
-const Client_18 = require("./api/resources/textToVoice/client/Client");
-const Client_19 = require("./api/resources/tokens/client/Client");
-const Client_20 = require("./api/resources/usage/client/Client");
-const Client_21 = require("./api/resources/user/client/Client");
-const Client_22 = require("./api/resources/voices/client/Client");
-const Client_23 = require("./api/resources/webhooks/client/Client");
-const Client_24 = require("./api/resources/workspace/client/Client");
+const Client_5 = require("./api/resources/environmentVariables/client/Client");
+const Client_6 = require("./api/resources/forcedAlignment/client/Client");
+const Client_7 = require("./api/resources/history/client/Client");
+const Client_8 = require("./api/resources/models/client/Client");
+const Client_9 = require("./api/resources/music/client/Client");
+const Client_10 = require("./api/resources/pronunciationDictionaries/client/Client");
+const Client_11 = require("./api/resources/samples/client/Client");
+const Client_12 = require("./api/resources/serviceAccounts/client/Client");
+const Client_13 = require("./api/resources/speechToSpeech/client/Client");
+const Client_14 = require("./api/resources/speechToText/client/Client");
+const Client_15 = require("./api/resources/studio/client/Client");
+const Client_16 = require("./api/resources/textToDialogue/client/Client");
+const Client_17 = require("./api/resources/textToSoundEffects/client/Client");
+const Client_18 = require("./api/resources/textToSpeech/client/Client");
+const Client_19 = require("./api/resources/textToVoice/client/Client");
+const Client_20 = require("./api/resources/tokens/client/Client");
+const Client_21 = require("./api/resources/usage/client/Client");
+const Client_22 = require("./api/resources/user/client/Client");
+const Client_23 = require("./api/resources/voices/client/Client");
+const Client_24 = require("./api/resources/webhooks/client/Client");
+const Client_25 = require("./api/resources/workspace/client/Client");
 const BaseClient_1 = require("./BaseClient");
 const core = __importStar(require("./core"));
 const headers_1 = require("./core/headers");
@@ -80,11 +81,11 @@ class ElevenLabsClient {
     }
     get history() {
         var _a;
-        return ((_a = this._history) !== null && _a !== void 0 ? _a : (this._history = new Client_6.HistoryClient(this._options)));
+        return ((_a = this._history) !== null && _a !== void 0 ? _a : (this._history = new Client_7.HistoryClient(this._options)));
     }
     get textToSoundEffects() {
         var _a;
-        return ((_a = this._textToSoundEffects) !== null && _a !== void 0 ? _a : (this._textToSoundEffects = new Client_16.TextToSoundEffectsClient(this._options)));
+        return ((_a = this._textToSoundEffects) !== null && _a !== void 0 ? _a : (this._textToSoundEffects = new Client_17.TextToSoundEffectsClient(this._options)));
     }
     get audioIsolation() {
         var _a;
@@ -92,35 +93,39 @@ class ElevenLabsClient {
     }
     get samples() {
         var _a;
-        return ((_a = this._samples) !== null && _a !== void 0 ? _a : (this._samples = new Client_10.SamplesClient(this._options)));
+        return ((_a = this._samples) !== null && _a !== void 0 ? _a : (this._samples = new Client_11.SamplesClient(this._options)));
     }
     get textToSpeech() {
         var _a;
-        return ((_a = this._textToSpeech) !== null && _a !== void 0 ? _a : (this._textToSpeech = new Client_17.TextToSpeechClient(this._options)));
+        return ((_a = this._textToSpeech) !== null && _a !== void 0 ? _a : (this._textToSpeech = new Client_18.TextToSpeechClient(this._options)));
     }
     get textToDialogue() {
         var _a;
-        return ((_a = this._textToDialogue) !== null && _a !== void 0 ? _a : (this._textToDialogue = new Client_15.TextToDialogueClient(this._options)));
+        return ((_a = this._textToDialogue) !== null && _a !== void 0 ? _a : (this._textToDialogue = new Client_16.TextToDialogueClient(this._options)));
     }
     get speechToSpeech() {
         var _a;
-        return ((_a = this._speechToSpeech) !== null && _a !== void 0 ? _a : (this._speechToSpeech = new Client_12.SpeechToSpeechClient(this._options)));
+        return ((_a = this._speechToSpeech) !== null && _a !== void 0 ? _a : (this._speechToSpeech = new Client_13.SpeechToSpeechClient(this._options)));
     }
     get textToVoice() {
         var _a;
-        return ((_a = this._textToVoice) !== null && _a !== void 0 ? _a : (this._textToVoice = new Client_18.TextToVoiceClient(this._options)));
+        return ((_a = this._textToVoice) !== null && _a !== void 0 ? _a : (this._textToVoice = new Client_19.TextToVoiceClient(this._options)));
     }
     get user() {
         var _a;
-        return ((_a = this._user) !== null && _a !== void 0 ? _a : (this._user = new Client_21.UserClient(this._options)));
+        return ((_a = this._user) !== null && _a !== void 0 ? _a : (this._user = new Client_22.UserClient(this._options)));
     }
     get voices() {
         var _a;
-        return ((_a = this._voices) !== null && _a !== void 0 ? _a : (this._voices = new Client_22.VoicesClient(this._options)));
+        return ((_a = this._voices) !== null && _a !== void 0 ? _a : (this._voices = new Client_23.VoicesClient(this._options)));
     }
     get studio() {
         var _a;
-        return ((_a = this._studio) !== null && _a !== void 0 ? _a : (this._studio = new Client_14.StudioClient(this._options)));
+        return ((_a = this._studio) !== null && _a !== void 0 ? _a : (this._studio = new Client_15.StudioClient(this._options)));
+    }
+    get music() {
+        var _a;
+        return ((_a = this._music) !== null && _a !== void 0 ? _a : (this._music = new Client_9.MusicClient(this._options)));
     }
     get dubbing() {
         var _a;
@@ -128,7 +133,7 @@ class ElevenLabsClient {
     }
     get models() {
         var _a;
-        return ((_a = this._models) !== null && _a !== void 0 ? _a : (this._models = new Client_7.ModelsClient(this._options)));
+        return ((_a = this._models) !== null && _a !== void 0 ? _a : (this._models = new Client_8.ModelsClient(this._options)));
     }
     get audioNative() {
         var _a;
@@ -136,43 +141,43 @@ class ElevenLabsClient {
     }
     get usage() {
         var _a;
-        return ((_a = this._usage) !== null && _a !== void 0 ? _a : (this._usage = new Client_20.UsageClient(this._options)));
+        return ((_a = this._usage) !== null && _a !== void 0 ? _a : (this._usage = new Client_21.UsageClient(this._options)));
     }
     get pronunciationDictionaries() {
         var _a;
-        return ((_a = this._pronunciationDictionaries) !== null && _a !== void 0 ? _a : (this._pronunciationDictionaries = new Client_9.PronunciationDictionariesClient(this._options)));
+        return ((_a = this._pronunciationDictionaries) !== null && _a !== void 0 ? _a : (this._pronunciationDictionaries = new Client_10.PronunciationDictionariesClient(this._options)));
     }
     get serviceAccounts() {
         var _a;
-        return ((_a = this._serviceAccounts) !== null && _a !== void 0 ? _a : (this._serviceAccounts = new Client_11.ServiceAccountsClient(this._options)));
+        return ((_a = this._serviceAccounts) !== null && _a !== void 0 ? _a : (this._serviceAccounts = new Client_12.ServiceAccountsClient(this._options)));
     }
     get webhooks() {
         var _a;
-        return ((_a = this._webhooks) !== null && _a !== void 0 ? _a : (this._webhooks = new Client_23.WebhooksClient(this._options)));
+        return ((_a = this._webhooks) !== null && _a !== void 0 ? _a : (this._webhooks = new Client_24.WebhooksClient(this._options)));
     }
     get speechToText() {
         var _a;
-        return ((_a = this._speechToText) !== null && _a !== void 0 ? _a : (this._speechToText = new Client_13.SpeechToTextClient(this._options)));
+        return ((_a = this._speechToText) !== null && _a !== void 0 ? _a : (this._speechToText = new Client_14.SpeechToTextClient(this._options)));
     }
     get forcedAlignment() {
         var _a;
-        return ((_a = this._forcedAlignment) !== null && _a !== void 0 ? _a : (this._forcedAlignment = new Client_5.ForcedAlignmentClient(this._options)));
+        return ((_a = this._forcedAlignment) !== null && _a !== void 0 ? _a : (this._forcedAlignment = new Client_6.ForcedAlignmentClient(this._options)));
     }
     get conversationalAi() {
         var _a;
         return ((_a = this._conversationalAi) !== null && _a !== void 0 ? _a : (this._conversationalAi = new Client_3.ConversationalAiClient(this._options)));
     }
-    get music() {
+    get environmentVariables() {
         var _a;
-        return ((_a = this._music) !== null && _a !== void 0 ? _a : (this._music = new Client_8.MusicClient(this._options)));
+        return ((_a = this._environmentVariables) !== null && _a !== void 0 ? _a : (this._environmentVariables = new Client_5.EnvironmentVariablesClient(this._options)));
     }
     get tokens() {
         var _a;
-        return ((_a = this._tokens) !== null && _a !== void 0 ? _a : (this._tokens = new Client_19.TokensClient(this._options)));
+        return ((_a = this._tokens) !== null && _a !== void 0 ? _a : (this._tokens = new Client_20.TokensClient(this._options)));
     }
     get workspace() {
         var _a;
-        return ((_a = this._workspace) !== null && _a !== void 0 ? _a : (this._workspace = new Client_24.WorkspaceClient(this._options)));
+        return ((_a = this._workspace) !== null && _a !== void 0 ? _a : (this._workspace = new Client_25.WorkspaceClient(this._options)));
     }
     /**
      * @param {string} agent_id

package/api/errors/BadRequestError.d.ts

@@ -1,6 +1,5 @@
 import type * as core from "../../core";
 import * as errors from "../../errors/index";
-import type * as ElevenLabs from "../index";
 export declare class BadRequestError extends errors.ElevenLabsError {
-    constructor(body: ElevenLabs.BadRequestErrorBody, rawResponse?: core.RawResponse);
+    constructor(body?: unknown, rawResponse?: core.RawResponse);
 }

package/api/errors/UnprocessableEntityError.d.ts

@@ -1,6 +1,5 @@
 import type * as core from "../../core";
 import * as errors from "../../errors/index";
-import type * as ElevenLabs from "../index";
 export declare class UnprocessableEntityError extends errors.ElevenLabsError {
-    constructor(body: ElevenLabs.HttpValidationError, rawResponse?: core.RawResponse);
+    constructor(body?: unknown, rawResponse?: core.RawResponse);
 }

package/api/resources/audioIsolation/client/Client.js

@@ -99,12 +99,7 @@ class AudioIsolationClient {
             if (_response.error.reason === "status-code") {
                 switch (_response.error.statusCode) {
                     case 422:
-                        throw new ElevenLabs.UnprocessableEntityError(serializers.HttpValidationError.parseOrThrow(_response.error.body, {
-                            unrecognizedObjectKeys: "passthrough",
-                            allowUnrecognizedUnionMembers: true,
-                            allowUnrecognizedEnumValues: true,
-                            breadcrumbsPrefix: ["response"],
-                        }), _response.rawResponse);
+                        throw new ElevenLabs.UnprocessableEntityError(_response.error.body, _response.rawResponse);
                     default:
                         throw new errors.ElevenLabsError({
                             statusCode: _response.error.statusCode,
@@ -156,12 +151,7 @@ class AudioIsolationClient {
             if (_response.error.reason === "status-code") {
                 switch (_response.error.statusCode) {
                     case 422:
-                        throw new ElevenLabs.UnprocessableEntityError(serializers.HttpValidationError.parseOrThrow(_response.error.body, {
-                            unrecognizedObjectKeys: "passthrough",
-                            allowUnrecognizedUnionMembers: true,
-                            allowUnrecognizedEnumValues: true,
-                            breadcrumbsPrefix: ["response"],
-                        }), _response.rawResponse);
+                        throw new ElevenLabs.UnprocessableEntityError(_response.error.body, _response.rawResponse);
                     default:
                         throw new errors.ElevenLabsError({
                             statusCode: _response.error.statusCode,

package/api/resources/audioNative/client/Client.js

@@ -151,12 +151,7 @@ class AudioNativeClient {
             if (_response.error.reason === "status-code") {
                 switch (_response.error.statusCode) {
                     case 422:
-                        throw new ElevenLabs.UnprocessableEntityError(serializers.HttpValidationError.parseOrThrow(_response.error.body, {
-                            unrecognizedObjectKeys: "passthrough",
-                            allowUnrecognizedUnionMembers: true,
-                            allowUnrecognizedEnumValues: true,
-                            breadcrumbsPrefix: ["response"],
-                        }), _response.rawResponse);
+                        throw new ElevenLabs.UnprocessableEntityError(_response.error.body, _response.rawResponse);
                     default:
                         throw new errors.ElevenLabsError({
                             statusCode: _response.error.statusCode,
@@ -211,12 +206,7 @@ class AudioNativeClient {
             if (_response.error.reason === "status-code") {
                 switch (_response.error.statusCode) {
                     case 422:
-                        throw new ElevenLabs.UnprocessableEntityError(serializers.HttpValidationError.parseOrThrow(_response.error.body, {
-                            unrecognizedObjectKeys: "passthrough",
-                            allowUnrecognizedUnionMembers: true,
-                            allowUnrecognizedEnumValues: true,
-                            breadcrumbsPrefix: ["response"],
-                        }), _response.rawResponse);
+                        throw new ElevenLabs.UnprocessableEntityError(_response.error.body, _response.rawResponse);
                     default:
                         throw new errors.ElevenLabsError({
                             statusCode: _response.error.statusCode,
@@ -287,12 +277,7 @@ class AudioNativeClient {
             if (_response.error.reason === "status-code") {
                 switch (_response.error.statusCode) {
                     case 422:
-                        throw new ElevenLabs.UnprocessableEntityError(serializers.HttpValidationError.parseOrThrow(_response.error.body, {
-                            unrecognizedObjectKeys: "passthrough",
-                            allowUnrecognizedUnionMembers: true,
-                            allowUnrecognizedEnumValues: true,
-                            breadcrumbsPrefix: ["response"],
-                        }), _response.rawResponse);
+                        throw new ElevenLabs.UnprocessableEntityError(_response.error.body, _response.rawResponse);
                     default:
                         throw new errors.ElevenLabsError({
                             statusCode: _response.error.statusCode,
@@ -354,12 +339,7 @@ class AudioNativeClient {
             if (_response.error.reason === "status-code") {
                 switch (_response.error.statusCode) {
                     case 422:
-                        throw new ElevenLabs.UnprocessableEntityError(serializers.HttpValidationError.parseOrThrow(_response.error.body, {
-                            unrecognizedObjectKeys: "passthrough",
-                            allowUnrecognizedUnionMembers: true,
-                            allowUnrecognizedEnumValues: true,
-                            breadcrumbsPrefix: ["response"],
-                        }), _response.rawResponse);
+                        throw new ElevenLabs.UnprocessableEntityError(_response.error.body, _response.rawResponse);
                     default:
                         throw new errors.ElevenLabsError({
                             statusCode: _response.error.statusCode,

package/api/resources/conversationalAi/client/Client.js

@@ -215,12 +215,7 @@ class ConversationalAiClient {
             if (_response.error.reason === "status-code") {
                 switch (_response.error.statusCode) {
                     case 422:
-                        throw new ElevenLabs.UnprocessableEntityError(serializers.HttpValidationError.parseOrThrow(_response.error.body, {
-                            unrecognizedObjectKeys: "passthrough",
-                            allowUnrecognizedUnionMembers: true,
-                            allowUnrecognizedEnumValues: true,
-                            breadcrumbsPrefix: ["response"],
-                        }), _response.rawResponse);
+                        throw new ElevenLabs.UnprocessableEntityError(_response.error.body, _response.rawResponse);
                     default:
                         throw new errors.ElevenLabsError({
                             statusCode: _response.error.statusCode,
@@ -274,12 +269,7 @@ class ConversationalAiClient {
             if (_response.error.reason === "status-code") {
                 switch (_response.error.statusCode) {
                     case 422:
-                        throw new ElevenLabs.UnprocessableEntityError(serializers.HttpValidationError.parseOrThrow(_response.error.body, {
-                            unrecognizedObjectKeys: "passthrough",
-                            allowUnrecognizedUnionMembers: true,
-                            allowUnrecognizedEnumValues: true,
-                            breadcrumbsPrefix: ["response"],
-                        }), _response.rawResponse);
+                        throw new ElevenLabs.UnprocessableEntityError(_response.error.body, _response.rawResponse);
                     default:
                         throw new errors.ElevenLabsError({
                             statusCode: _response.error.statusCode,
@@ -334,12 +324,7 @@ class ConversationalAiClient {
             if (_response.error.reason === "status-code") {
                 switch (_response.error.statusCode) {
                     case 422:
-                        throw new ElevenLabs.UnprocessableEntityError(serializers.HttpValidationError.parseOrThrow(_response.error.body, {
-                            unrecognizedObjectKeys: "passthrough",
-                            allowUnrecognizedUnionMembers: true,
-                            allowUnrecognizedEnumValues: true,
-                            breadcrumbsPrefix: ["response"],
-                        }), _response.rawResponse);
+                        throw new ElevenLabs.UnprocessableEntityError(_response.error.body, _response.rawResponse);
                     default:
                         throw new errors.ElevenLabsError({
                             statusCode: _response.error.statusCode,
@@ -395,12 +380,7 @@ class ConversationalAiClient {
             if (_response.error.reason === "status-code") {
                 switch (_response.error.statusCode) {
                     case 422:
-                        throw new ElevenLabs.UnprocessableEntityError(serializers.HttpValidationError.parseOrThrow(_response.error.body, {
-                            unrecognizedObjectKeys: "passthrough",
-                            allowUnrecognizedUnionMembers: true,
-                            allowUnrecognizedEnumValues: true,
-                            breadcrumbsPrefix: ["response"],
-                        }), _response.rawResponse);
+                        throw new ElevenLabs.UnprocessableEntityError(_response.error.body, _response.rawResponse);
                     default:
                         throw new errors.ElevenLabsError({
                             statusCode: _response.error.statusCode,

package/api/resources/conversationalAi/conversation/Conversation.d.ts

@@ -66,6 +66,18 @@ export declare class Conversation extends EventEmitter {
      * This sends a ping to the orchestrator to reset the timeout timer.
      */
     registerUserActivity(): void;
+    /**
+     * Send a multimodal message combining text and a file reference to the agent.
+     *
+     * At least one of `text` or `fileId` must be provided.
+     *
+     * @param options.text Optional text message to include
+     * @param options.fileId Optional ElevenLabs file ID to include
+     */
+    sendMultimodalMessage(options: {
+        text?: string;
+        fileId?: string;
+    }): void;
     /**
      * Send a contextual update to the conversation.
      *

package/api/resources/conversationalAi/conversation/Conversation.js

@@ -120,6 +120,24 @@ class Conversation extends events_1.EventEmitter {
         };
         this.ws.send(JSON.stringify(event));
     }
+    /**
+     * Send a multimodal message combining text and a file reference to the agent.
+     *
+     * At least one of `text` or `fileId` must be provided.
+     *
+     * @param options.text Optional text message to include
+     * @param options.fileId Optional ElevenLabs file ID to include
+     */
+    sendMultimodalMessage(options) {
+        if (!this.ws || this.ws.readyState !== ws_1.default.OPEN) {
+            throw new Error("Session not started or websocket not connected");
+        }
+        if (!options.text && !options.fileId) {
+            throw new Error("At least one of text or fileId must be provided");
+        }
+        const event = Object.assign(Object.assign({ type: events_2.ClientToOrchestratorEvent.MULTIMODAL_MESSAGE }, (options.text && { text: { type: events_2.ClientToOrchestratorEvent.USER_MESSAGE, text: options.text } })), (options.fileId && { file: { type: "file_input", file_id: options.fileId } }));
+        this.ws.send(JSON.stringify(event));
+    }
     /**
      * Send a contextual update to the conversation.
      *

package/api/resources/conversationalAi/conversation/__tests__/Conversation.test.js

@@ -172,6 +172,31 @@ describe("Conversation", () => {
                 text: "User is looking at product page",
             }));
         });
+        it("should send multimodal message with text and file", () => {
+            conversation.sendMultimodalMessage({ text: "What is in this file?", fileId: "file_abc123" });
+            expect(mockWebSocket.send).toHaveBeenCalledWith(JSON.stringify({
+                type: "multimodal_message",
+                text: { type: "user_message", text: "What is in this file?" },
+                file: { type: "file_input", file_id: "file_abc123" },
+            }));
+        });
+        it("should send multimodal message with text only", () => {
+            conversation.sendMultimodalMessage({ text: "Hello" });
+            expect(mockWebSocket.send).toHaveBeenCalledWith(JSON.stringify({
+                type: "multimodal_message",
+                text: { type: "user_message", text: "Hello" },
+            }));
+        });
+        it("should send multimodal message with file only", () => {
+            conversation.sendMultimodalMessage({ fileId: "file_abc123" });
+            expect(mockWebSocket.send).toHaveBeenCalledWith(JSON.stringify({
+                type: "multimodal_message",
+                file: { type: "file_input", file_id: "file_abc123" },
+            }));
+        });
+        it("should throw error when sending multimodal message without text or fileId", () => {
+            expect(() => conversation.sendMultimodalMessage({})).toThrow("At least one of text or fileId must be provided");
+        });
         it("should throw error when sending message without active session", () => {
             conversation.endSession();
             expect(() => conversation.sendUserMessage("test")).toThrow("Session not started or websocket not connected");

package/api/resources/conversationalAi/conversation/events.d.ts

@@ -11,7 +11,9 @@ export declare enum ClientToOrchestratorEvent {
     CONTEXTUAL_UPDATE = "contextual_update",
     /** User text message. */
     USER_MESSAGE = "user_message",
-    USER_ACTIVITY = "user_activity"
+    USER_ACTIVITY = "user_activity",
+    /** Multimodal message combining text and a file reference. */
+    MULTIMODAL_MESSAGE = "multimodal_message"
 }
 /**
  * Base interface for all client-to-orchestrator events.
@@ -72,6 +74,21 @@ export interface FeedbackEvent extends BaseClientToOrchestratorEvent {
     feedback: string;
     rating?: number;
 }
+/**
+ * File component of a multimodal message.
+ */
+export interface MultimodalMessageFile {
+    type: "file_input";
+    file_id: string;
+}
+/**
+ * Event for sending multimodal messages combining text and a file reference.
+ */
+export interface MultimodalMessageClientToOrchestratorEvent extends BaseClientToOrchestratorEvent {
+    type: ClientToOrchestratorEvent.MULTIMODAL_MESSAGE;
+    text?: UserMessageClientToOrchestratorEvent;
+    file?: MultimodalMessageFile;
+}
 /**
  * Event for sending user audio chunks.
  */
@@ -81,4 +98,4 @@ export interface UserAudioChunkEvent {
 /**
  * Union type for all client-to-orchestrator events.
  */
-export type ClientToOrchestratorEventUnion = UserMessageClientToOrchestratorEvent | UserActivityClientToOrchestratorEvent | ContextualUpdateClientToOrchestratorEvent | ConversationInitiationClientDataEvent | ClientToolResultEvent | PongEvent | FeedbackEvent;
+export type ClientToOrchestratorEventUnion = UserMessageClientToOrchestratorEvent | UserActivityClientToOrchestratorEvent | ContextualUpdateClientToOrchestratorEvent | ConversationInitiationClientDataEvent | ClientToolResultEvent | PongEvent | FeedbackEvent | MultimodalMessageClientToOrchestratorEvent;

package/api/resources/conversationalAi/conversation/events.js

@@ -16,4 +16,6 @@ var ClientToOrchestratorEvent;
     /** User text message. */
     ClientToOrchestratorEvent["USER_MESSAGE"] = "user_message";
     ClientToOrchestratorEvent["USER_ACTIVITY"] = "user_activity";
+    /** Multimodal message combining text and a file reference. */
+    ClientToOrchestratorEvent["MULTIMODAL_MESSAGE"] = "multimodal_message";
 })(ClientToOrchestratorEvent || (exports.ClientToOrchestratorEvent = ClientToOrchestratorEvent = {}));

package/api/resources/conversationalAi/resources/agents/client/Client.d.ts

@@ -111,6 +111,7 @@ export declare class AgentsClient {
      *         search: "search",
      *         archived: true,
      *         showOnlyOwnedAgents: true,
+     *         createdByUserId: "created_by_user_id",
      *         sortDirection: "asc",
      *         sortBy: "name",
      *         cursor: "cursor"