@azure/ai-voicelive 1.0.0-alpha.20251117.2

package/dist/esm/protocol/messageParser.d.ts

@@ -0,0 +1,42 @@
+import type { ServerEventUnion, ClientEventUnion } from "../models/index.js";
+/**
+ * Parsed message containing event data and metadata
+ */
+export interface ParsedMessage {
+    /** Type of event: client or server */
+    type: "client" | "server";
+    /** Parsed event data */
+    event: ClientEventUnion | ServerEventUnion;
+    /** Original raw data for debugging */
+    raw: string | ArrayBuffer;
+}
+/**
+ * Parses and serializes Voice Live protocol messages
+ */
+export declare class VoiceLiveMessageParser {
+    /**
+     * Parses incoming WebSocket message data into typed events
+     */
+    parseIncomingMessage(data: string | ArrayBuffer): ParsedMessage | null;
+    /**
+     * Serializes outgoing client events for WebSocket transmission
+     */
+    serializeOutgoingMessage(event: ClientEventUnion): string;
+    /**
+     * Checks if the parsed data represents a server event
+     */
+    private _isServerEvent;
+    /**
+     * Checks if the parsed data represents a client event
+     */
+    private _isClientEvent;
+    /**
+     * Validates if the type string represents a valid server event type
+     */
+    private _isValidServerEventType;
+    /**
+     * Validates if the type string represents a valid client event type
+     */
+    private _isValidClientEventType;
+}
+//# sourceMappingURL=messageParser.d.ts.map

package/dist/esm/protocol/messageParser.js

@@ -0,0 +1,150 @@
+// Copyright (c) Microsoft Corporation.
+// Licensed under the MIT License.
+import { clientEventUnionSerializer, serverEventUnionDeserializer } from "../models/models.js";
+/**
+ * Parses and serializes Voice Live protocol messages
+ */
+export class VoiceLiveMessageParser {
+    /**
+     * Parses incoming WebSocket message data into typed events
+     */
+    parseIncomingMessage(data) {
+        try {
+            let messageText;
+            if (data instanceof ArrayBuffer) {
+                messageText = new TextDecoder().decode(data);
+            }
+            else {
+                messageText = data;
+            }
+            const parsedData = JSON.parse(messageText);
+            // Validate and type the message based on the 'type' field
+            if (this._isServerEvent(parsedData)) {
+                return {
+                    type: "server",
+                    event: serverEventUnionDeserializer(parsedData),
+                    raw: data,
+                };
+            }
+            // Handle unexpected client events (shouldn't normally receive these from server)
+            if (this._isClientEvent(parsedData)) {
+                return {
+                    type: "client",
+                    event: parsedData,
+                    raw: data,
+                };
+            }
+            return null; // Unknown message format
+        }
+        catch (error) {
+            // Invalid JSON or parsing error
+            return null;
+        }
+    }
+    /**
+     * Serializes outgoing client events for WebSocket transmission
+     */
+    serializeOutgoingMessage(event) {
+        return JSON.stringify(clientEventUnionSerializer(event));
+    }
+    /**
+     * Checks if the parsed data represents a server event
+     */
+    _isServerEvent(data) {
+        // Check if data matches ServerEventUnion structure
+        return data && typeof data.type === "string" && this._isValidServerEventType(data.type);
+    }
+    /**
+     * Checks if the parsed data represents a client event
+     */
+    _isClientEvent(data) {
+        // Check if data matches ClientEventUnion structure
+        return data && typeof data.type === "string" && this._isValidClientEventType(data.type);
+    }
+    /**
+     * Validates if the type string represents a valid server event type
+     */
+    _isValidServerEventType(type) {
+        // Based on the comprehensive analysis in EXISTING_TYPES_ANALYSIS.md
+        const validServerTypes = [
+            // Error handling
+            "error",
+            // Session management
+            "session.created",
+            "session.updated",
+            "session.avatar.connecting",
+            // Audio buffer events
+            "input_audio_buffer.committed",
+            "input_audio_buffer.cleared",
+            "input_audio_buffer.speech_started",
+            "input_audio_buffer.speech_stopped",
+            // Conversation item events
+            "conversation.item.created",
+            "conversation.item.truncated",
+            "conversation.item.deleted",
+            "conversation.item.retrieved",
+            "conversation.item.input_audio_transcription.completed",
+            "conversation.item.input_audio_transcription.failed",
+            "conversation.item.input_audio_transcription.delta",
+            // Response lifecycle events
+            "response.created",
+            "response.done",
+            "response.output_item.added",
+            "response.output_item.done",
+            // Content streaming events
+            "response.content_part.added",
+            "response.content_part.done",
+            "response.text.delta",
+            "response.text.done",
+            // Audio streaming events
+            "response.audio_transcript.delta",
+            "response.audio_transcript.done",
+            "response.audio.delta",
+            "response.audio.done",
+            // Animation events
+            "response.animation.blendshape.delta",
+            "response.animation.blendshape.done",
+            "response.animation.viseme.delta",
+            "response.animation.viseme.done",
+            // Timestamp events
+            "response.audio.timestamp.delta",
+            "response.audio.timestamp.done",
+            // Function call events
+            "response.function_call_arguments.delta",
+            "response.function_call_arguments.done",
+        ];
+        return validServerTypes.includes(type);
+    }
+    /**
+     * Validates if the type string represents a valid client event type
+     */
+    _isValidClientEventType(type) {
+        // Based on the comprehensive analysis in EXISTING_TYPES_ANALYSIS.md
+        const validClientTypes = [
+            // Session management
+            "session.update",
+            "session.avatar.connect",
+            // Turn-based audio
+            "input_audio_turn.start",
+            "input_audio_turn.append",
+            "input_audio_turn.end",
+            "input_audio_turn.cancel",
+            // Buffer-based audio
+            "input_audio_buffer.append",
+            "input_audio_buffer.commit",
+            "input_audio_buffer.clear",
+            // Audio control
+            "input_audio.clear",
+            // Conversation management
+            "conversation.item.create",
+            "conversation.item.truncate",
+            "conversation.item.delete",
+            "conversation.item.retrieve",
+            // Response management
+            "response.create",
+            "response.cancel",
+        ];
+        return validClientTypes.includes(type);
+    }
+}
+//# sourceMappingURL=messageParser.js.map

package/dist/esm/protocol/messageParser.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"messageParser.js","sourceRoot":"","sources":["../../../src/protocol/messageParser.ts"],"names":[],"mappings":"AAAA,uCAAuC;AACvC,kCAAkC;AAGlC,OAAO,EAAE,0BAA0B,EAAE,4BAA4B,EAAE,MAAM,qBAAqB,CAAC;AAc/F;;GAEG;AACH,MAAM,OAAO,sBAAsB;IACjC;;OAEG;IACH,oBAAoB,CAAC,IAA0B;QAC7C,IAAI,CAAC;YACH,IAAI,WAAmB,CAAC;YAExB,IAAI,IAAI,YAAY,WAAW,EAAE,CAAC;gBAChC,WAAW,GAAG,IAAI,WAAW,EAAE,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC;YAC/C,CAAC;iBAAM,CAAC;gBACN,WAAW,GAAG,IAAI,CAAC;YACrB,CAAC;YAED,MAAM,UAAU,GAAG,IAAI,CAAC,KAAK,CAAC,WAAW,CAAC,CAAC;YAE3C,0DAA0D;YAC1D,IAAI,IAAI,CAAC,cAAc,CAAC,UAAU,CAAC,EAAE,CAAC;gBACpC,OAAO;oBACL,IAAI,EAAE,QAAQ;oBACd,KAAK,EAAE,4BAA4B,CAAC,UAAU,CAAC;oBAC/C,GAAG,EAAE,IAAI;iBACV,CAAC;YACJ,CAAC;YAED,iFAAiF;YACjF,IAAI,IAAI,CAAC,cAAc,CAAC,UAAU,CAAC,EAAE,CAAC;gBACpC,OAAO;oBACL,IAAI,EAAE,QAAQ;oBACd,KAAK,EAAE,UAA8B;oBACrC,GAAG,EAAE,IAAI;iBACV,CAAC;YACJ,CAAC;YAED,OAAO,IAAI,CAAC,CAAC,yBAAyB;QACxC,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YACf,gCAAgC;YAChC,OAAO,IAAI,CAAC;QACd,CAAC;IACH,CAAC;IAED;;OAEG;IACH,wBAAwB,CAAC,KAAuB;QAC9C,OAAO,IAAI,CAAC,SAAS,CAAC,0BAA0B,CAAC,KAAK,CAAC,CAAC,CAAC;IAC3D,CAAC;IAED;;OAEG;IACK,cAAc,CAAC,IAAS;QAC9B,mDAAmD;QACnD,OAAO,IAAI,IAAI,OAAO,IAAI,CAAC,IAAI,KAAK,QAAQ,IAAI,IAAI,CAAC,uBAAuB,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;IAC1F,CAAC;IAED;;OAEG;IACK,cAAc,CAAC,IAAS;QAC9B,mDAAmD;QACnD,OAAO,IAAI,IAAI,OAAO,IAAI,CAAC,IAAI,KAAK,QAAQ,IAAI,IAAI,CAAC,uBAAuB,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;IAC1F,CAAC;IAED;;OAEG;IACK,uBAAuB,CAAC,IAAY;QAC1C,oEAAoE;QACpE,MAAM,gBAAgB,GAAG;YACvB,iBAAiB;YACjB,OAAO;YAEP,qBAAqB;YACrB,iBAAiB;YACjB,iBAAiB;YACjB,2BAA2B;YAE3B,sBAAsB;YACtB,8BAA8B;YAC9B,4BAA4B;YAC5B,mCAAmC;YACnC,mCAAmC;YAEnC,2BAA2B;YAC3B,2BAA2B;YAC3B,6BAA6B;YAC7B,2BAA2B;YAC3B,6BAA6B;YAC7B,uDAAuD;YACvD,oDAAoD;YACpD,mDAAmD;YAEnD,4BAA4B;YAC5B,kBAAkB;YAClB,eAAe;YACf,4BAA4B;YAC5B,2BAA2B;YAE3B,2BAA2B;YAC3B,6BAA6B;YAC7B,4BAA4B;YAC5B,qBAAqB;YACrB,oBAAoB;YAEpB,yBAAyB;YACzB,iCAAiC;YACjC,gCAAgC;YAChC,sBAAsB;YACtB,qBAAqB;YAErB,mBAAmB;YACnB,qCAAqC;YACrC,oCAAoC;YACpC,iCAAiC;YACjC,gCAAgC;YAEhC,mBAAmB;YACnB,gCAAgC;YAChC,+BAA+B;YAE/B,uBAAuB;YACvB,wCAAwC;YACxC,uCAAuC;SACxC,CAAC;QAEF,OAAO,gBAAgB,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC;IACzC,CAAC;IAED;;OAEG;IACK,uBAAuB,CAAC,IAAY;QAC1C,oEAAoE;QACpE,MAAM,gBAAgB,GAAG;YACvB,qBAAqB;YACrB,gBAAgB;YAChB,wBAAwB;YAExB,mBAAmB;YACnB,wBAAwB;YACxB,yBAAyB;YACzB,sBAAsB;YACtB,yBAAyB;YAEzB,qBAAqB;YACrB,2BAA2B;YAC3B,2BAA2B;YAC3B,0BAA0B;YAE1B,gBAAgB;YAChB,mBAAmB;YAEnB,0BAA0B;YAC1B,0BAA0B;YAC1B,4BAA4B;YAC5B,0BAA0B;YAC1B,4BAA4B;YAE5B,sBAAsB;YACtB,iBAAiB;YACjB,iBAAiB;SAClB,CAAC;QAEF,OAAO,gBAAgB,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC;IACzC,CAAC;CACF","sourcesContent":["// Copyright (c) Microsoft Corporation.\n// Licensed under the MIT License.\n\nimport type { ServerEventUnion, ClientEventUnion } from \"../models/index.js\";\nimport { clientEventUnionSerializer, serverEventUnionDeserializer } from \"../models/models.js\";\n\n/**\n * Parsed message containing event data and metadata\n */\nexport interface ParsedMessage {\n  /** Type of event: client or server */\n  type: \"client\" | \"server\";\n  /** Parsed event data */\n  event: ClientEventUnion | ServerEventUnion;\n  /** Original raw data for debugging */\n  raw: string | ArrayBuffer;\n}\n\n/**\n * Parses and serializes Voice Live protocol messages\n */\nexport class VoiceLiveMessageParser {\n  /**\n   * Parses incoming WebSocket message data into typed events\n   */\n  parseIncomingMessage(data: string | ArrayBuffer): ParsedMessage | null {\n    try {\n      let messageText: string;\n\n      if (data instanceof ArrayBuffer) {\n        messageText = new TextDecoder().decode(data);\n      } else {\n        messageText = data;\n      }\n\n      const parsedData = JSON.parse(messageText);\n\n      // Validate and type the message based on the 'type' field\n      if (this._isServerEvent(parsedData)) {\n        return {\n          type: \"server\",\n          event: serverEventUnionDeserializer(parsedData),\n          raw: data,\n        };\n      }\n\n      // Handle unexpected client events (shouldn't normally receive these from server)\n      if (this._isClientEvent(parsedData)) {\n        return {\n          type: \"client\",\n          event: parsedData as ClientEventUnion,\n          raw: data,\n        };\n      }\n\n      return null; // Unknown message format\n    } catch (error) {\n      // Invalid JSON or parsing error\n      return null;\n    }\n  }\n\n  /**\n   * Serializes outgoing client events for WebSocket transmission\n   */\n  serializeOutgoingMessage(event: ClientEventUnion): string {\n    return JSON.stringify(clientEventUnionSerializer(event));\n  }\n\n  /**\n   * Checks if the parsed data represents a server event\n   */\n  private _isServerEvent(data: any): boolean {\n    // Check if data matches ServerEventUnion structure\n    return data && typeof data.type === \"string\" && this._isValidServerEventType(data.type);\n  }\n\n  /**\n   * Checks if the parsed data represents a client event\n   */\n  private _isClientEvent(data: any): boolean {\n    // Check if data matches ClientEventUnion structure\n    return data && typeof data.type === \"string\" && this._isValidClientEventType(data.type);\n  }\n\n  /**\n   * Validates if the type string represents a valid server event type\n   */\n  private _isValidServerEventType(type: string): boolean {\n    // Based on the comprehensive analysis in EXISTING_TYPES_ANALYSIS.md\n    const validServerTypes = [\n      // Error handling\n      \"error\",\n\n      // Session management\n      \"session.created\",\n      \"session.updated\",\n      \"session.avatar.connecting\",\n\n      // Audio buffer events\n      \"input_audio_buffer.committed\",\n      \"input_audio_buffer.cleared\",\n      \"input_audio_buffer.speech_started\",\n      \"input_audio_buffer.speech_stopped\",\n\n      // Conversation item events\n      \"conversation.item.created\",\n      \"conversation.item.truncated\",\n      \"conversation.item.deleted\",\n      \"conversation.item.retrieved\",\n      \"conversation.item.input_audio_transcription.completed\",\n      \"conversation.item.input_audio_transcription.failed\",\n      \"conversation.item.input_audio_transcription.delta\",\n\n      // Response lifecycle events\n      \"response.created\",\n      \"response.done\",\n      \"response.output_item.added\",\n      \"response.output_item.done\",\n\n      // Content streaming events\n      \"response.content_part.added\",\n      \"response.content_part.done\",\n      \"response.text.delta\",\n      \"response.text.done\",\n\n      // Audio streaming events\n      \"response.audio_transcript.delta\",\n      \"response.audio_transcript.done\",\n      \"response.audio.delta\",\n      \"response.audio.done\",\n\n      // Animation events\n      \"response.animation.blendshape.delta\",\n      \"response.animation.blendshape.done\",\n      \"response.animation.viseme.delta\",\n      \"response.animation.viseme.done\",\n\n      // Timestamp events\n      \"response.audio.timestamp.delta\",\n      \"response.audio.timestamp.done\",\n\n      // Function call events\n      \"response.function_call_arguments.delta\",\n      \"response.function_call_arguments.done\",\n    ];\n\n    return validServerTypes.includes(type);\n  }\n\n  /**\n   * Validates if the type string represents a valid client event type\n   */\n  private _isValidClientEventType(type: string): boolean {\n    // Based on the comprehensive analysis in EXISTING_TYPES_ANALYSIS.md\n    const validClientTypes = [\n      // Session management\n      \"session.update\",\n      \"session.avatar.connect\",\n\n      // Turn-based audio\n      \"input_audio_turn.start\",\n      \"input_audio_turn.append\",\n      \"input_audio_turn.end\",\n      \"input_audio_turn.cancel\",\n\n      // Buffer-based audio\n      \"input_audio_buffer.append\",\n      \"input_audio_buffer.commit\",\n      \"input_audio_buffer.clear\",\n\n      // Audio control\n      \"input_audio.clear\",\n\n      // Conversation management\n      \"conversation.item.create\",\n      \"conversation.item.truncate\",\n      \"conversation.item.delete\",\n      \"conversation.item.retrieve\",\n\n      // Response management\n      \"response.create\",\n      \"response.cancel\",\n    ];\n\n    return validClientTypes.includes(type);\n  }\n}\n"]}

package/dist/esm/voiceLiveClient.d.ts

@@ -0,0 +1,65 @@
+import type { KeyCredential, TokenCredential } from "@azure/core-auth";
+import type { RequestSession } from "./models/index.js";
+import type { CreateSessionOptions, StartSessionOptions, VoiceLiveSessionOptions } from "./voiceLiveSession.js";
+import { VoiceLiveSession } from "./voiceLiveSession.js";
+export interface VoiceLiveClientOptions {
+    /** API version to use for the Voice Live service */
+    apiVersion?: string;
+    /** Default session options to apply to all created sessions */
+    defaultSessionOptions?: VoiceLiveSessionOptions;
+}
+/**
+ * The VoiceLive client provides session management for real-time conversational AI capabilities.
+ *
+ * This client acts as a factory for creating VoiceLiveSession instances, which handle the actual
+ * WebSocket connections and real-time interactions with the service.
+ */
+export declare class VoiceLiveClient {
+    private readonly _endpoint;
+    private readonly _credential;
+    private readonly _options;
+    /**
+     * Creates an instance of VoiceLiveClient with endpoint and credential.
+     *
+     * @param endpoint - The Voice Live service endpoint URL
+     * @param credential - Azure TokenCredential or KeyCredential for authentication
+     * @param options - Optional configuration for the client
+     */
+    constructor(endpoint: string, credential: TokenCredential | KeyCredential, options?: VoiceLiveClientOptions);
+    /**
+     * Creates a new VoiceLiveSession for real-time voice communication.
+     *
+     * @param model - The model name to use for the session (e.g., "gpt-4o-realtime-preview")
+     * @param sessionOptions - Optional configuration specific to this session
+     * @returns A new VoiceLiveSession instance ready to connect
+     */
+    createSession(model: string, sessionOptions?: CreateSessionOptions): VoiceLiveSession;
+    /**
+     * Creates a new VoiceLiveSession for real-time voice communication with session configuration.
+     *
+     * @param sessionConfig - Session configuration including model and other settings
+     * @returns A new VoiceLiveSession instance ready to connect
+     */
+    createSession(sessionConfig: RequestSession, sessionOptions?: CreateSessionOptions): VoiceLiveSession;
+    /**
+     * Creates and immediately connects a new VoiceLiveSession.
+     *
+     * @param model - The model name to use for the session (e.g., "gpt-4o-realtime-preview")
+     * @param sessionOptions - Optional configuration specific to this session
+     * @returns A connected VoiceLiveSession instance
+     */
+    startSession(model: string, sessionOptions?: StartSessionOptions): Promise<VoiceLiveSession>;
+    /**
+     * Creates and immediately connects a new VoiceLiveSession with session configuration.
+     *
+     * @param sessionConfig - Session configuration including model and other settings
+     * @param sessionOptions - Optional configuration specific to this session
+     * @returns A connected VoiceLiveSession instance
+     */
+    startSession(sessionConfig: RequestSession, sessionOptions?: StartSessionOptions): Promise<VoiceLiveSession>;
+    get endpoint(): string;
+    get apiVersion(): string;
+    private _buildDefaultOptions;
+    private _normalizeEndpoint;
+}
+//# sourceMappingURL=voiceLiveClient.d.ts.map

package/dist/esm/voiceLiveClient.js

@@ -0,0 +1,81 @@
+// Copyright (c) Microsoft Corporation.
+// Licensed under the MIT License.
+import { VoiceLiveSession } from "./voiceLiveSession.js";
+import { logger } from "./logger.js";
+/**
+ * The VoiceLive client provides session management for real-time conversational AI capabilities.
+ *
+ * This client acts as a factory for creating VoiceLiveSession instances, which handle the actual
+ * WebSocket connections and real-time interactions with the service.
+ */
+export class VoiceLiveClient {
+    _endpoint;
+    _credential;
+    _options;
+    /**
+     * Creates an instance of VoiceLiveClient with endpoint and credential.
+     *
+     * @param endpoint - The Voice Live service endpoint URL
+     * @param credential - Azure TokenCredential or KeyCredential for authentication
+     * @param options - Optional configuration for the client
+     */
+    constructor(endpoint, credential, options = {}) {
+        this._endpoint = this._normalizeEndpoint(endpoint);
+        this._credential = credential;
+        this._options = this._buildDefaultOptions(options);
+        logger.info("VoiceLiveClient created", {
+            endpoint: this._endpoint,
+            apiVersion: this._options.apiVersion,
+        });
+    }
+    createSession(modelOrConfig, sessionOptions) {
+        // Extract model name from the parameter
+        const model = typeof modelOrConfig === "string" ? modelOrConfig : modelOrConfig.model;
+        if (!model) {
+            throw new Error("Model name is required. Provide either a model string or RequestSession with model property.");
+        }
+        // Merge default session options with provided options
+        const mergedOptions = {
+            ...this._options.defaultSessionOptions,
+            ...sessionOptions,
+        };
+        const session = new VoiceLiveSession(this._endpoint, this._credential, this._options.apiVersion, model, mergedOptions);
+        // If full session config was provided, store it for later use
+        if (typeof modelOrConfig !== "string") {
+            session._initialSessionConfig = modelOrConfig;
+        }
+        logger.info("VoiceLiveSession created", { model });
+        return session;
+    }
+    async startSession(modelOrConfig, sessionOptions) {
+        const session = this.createSession(modelOrConfig, sessionOptions);
+        await session.connect();
+        // If full session config was provided, send it after connection
+        if (typeof modelOrConfig !== "string") {
+            await session.updateSession(modelOrConfig);
+        }
+        return session;
+    }
+    // Properties
+    get endpoint() {
+        return this._endpoint;
+    }
+    get apiVersion() {
+        return this._options.apiVersion;
+    }
+    _buildDefaultOptions(options) {
+        return {
+            apiVersion: options.apiVersion || "2025-10-01",
+            defaultSessionOptions: options.defaultSessionOptions || {},
+        };
+    }
+    _normalizeEndpoint(endpoint) {
+        // Ensure endpoint has proper protocol
+        if (!endpoint.startsWith("http://") && !endpoint.startsWith("https://")) {
+            endpoint = `https://${endpoint}`;
+        }
+        // Remove trailing slash
+        return endpoint.replace(/\/$/, "");
+    }
+}
+//# sourceMappingURL=voiceLiveClient.js.map

package/dist/esm/voiceLiveClient.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"voiceLiveClient.js","sourceRoot":"","sources":["../../src/voiceLiveClient.ts"],"names":[],"mappings":"AAAA,uCAAuC;AACvC,kCAAkC;AASlC,OAAO,EAAE,gBAAgB,EAAE,MAAM,uBAAuB,CAAC;AACzD,OAAO,EAAE,MAAM,EAAE,MAAM,aAAa,CAAC;AASrC;;;;;GAKG;AACH,MAAM,OAAO,eAAe;IACT,SAAS,CAAS;IAClB,WAAW,CAAkC;IAC7C,QAAQ,CAAmC;IAE5D;;;;;;OAMG;IACH,YACE,QAAgB,EAChB,UAA2C,EAC3C,UAAkC,EAAE;QAEpC,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC,kBAAkB,CAAC,QAAQ,CAAC,CAAC;QACnD,IAAI,CAAC,WAAW,GAAG,UAAU,CAAC;QAC9B,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,oBAAoB,CAAC,OAAO,CAAC,CAAC;QAEnD,MAAM,CAAC,IAAI,CAAC,yBAAyB,EAAE;YACrC,QAAQ,EAAE,IAAI,CAAC,SAAS;YACxB,UAAU,EAAE,IAAI,CAAC,QAAQ,CAAC,UAAU;SACrC,CAAC,CAAC;IACL,CAAC;IAsBD,aAAa,CACX,aAAsC,EACtC,cAAqC;QAErC,wCAAwC;QACxC,MAAM,KAAK,GAAG,OAAO,aAAa,KAAK,QAAQ,CAAC,CAAC,CAAC,aAAa,CAAC,CAAC,CAAC,aAAa,CAAC,KAAK,CAAC;QACtF,IAAI,CAAC,KAAK,EAAE,CAAC;YACX,MAAM,IAAI,KAAK,CACb,8FAA8F,CAC/F,CAAC;QACJ,CAAC;QAED,sDAAsD;QACtD,MAAM,aAAa,GAA4B;YAC7C,GAAG,IAAI,CAAC,QAAQ,CAAC,qBAAqB;YACtC,GAAG,cAAc;SAClB,CAAC;QAEF,MAAM,OAAO,GAAG,IAAI,gBAAgB,CAClC,IAAI,CAAC,SAAS,EACd,IAAI,CAAC,WAAW,EAChB,IAAI,CAAC,QAAQ,CAAC,UAAU,EACxB,KAAK,EACL,aAAa,CACd,CAAC;QAEF,8DAA8D;QAC9D,IAAI,OAAO,aAAa,KAAK,QAAQ,EAAE,CAAC;YACrC,OAAe,CAAC,qBAAqB,GAAG,aAAa,CAAC;QACzD,CAAC;QAED,MAAM,CAAC,IAAI,CAAC,0BAA0B,EAAE,EAAE,KAAK,EAAE,CAAC,CAAC;QACnD,OAAO,OAAO,CAAC;IACjB,CAAC;IA0BD,KAAK,CAAC,YAAY,CAChB,aAAsC,EACtC,cAAoC;QAEpC,MAAM,OAAO,GAAG,IAAI,CAAC,aAAa,CAAC,aAAoB,EAAE,cAAc,CAAC,CAAC;QACzE,MAAM,OAAO,CAAC,OAAO,EAAE,CAAC;QAExB,gEAAgE;QAChE,IAAI,OAAO,aAAa,KAAK,QAAQ,EAAE,CAAC;YACtC,MAAM,OAAO,CAAC,aAAa,CAAC,aAAa,CAAC,CAAC;QAC7C,CAAC;QAED,OAAO,OAAO,CAAC;IACjB,CAAC;IAED,aAAa;IACb,IAAI,QAAQ;QACV,OAAO,IAAI,CAAC,SAAS,CAAC;IACxB,CAAC;IAED,IAAI,UAAU;QACZ,OAAO,IAAI,CAAC,QAAQ,CAAC,UAAU,CAAC;IAClC,CAAC;IAEO,oBAAoB,CAAC,OAA+B;QAC1D,OAAO;YACL,UAAU,EAAE,OAAO,CAAC,UAAU,IAAI,YAAY;YAC9C,qBAAqB,EAAE,OAAO,CAAC,qBAAqB,IAAI,EAAE;SAC3D,CAAC;IACJ,CAAC;IAEO,kBAAkB,CAAC,QAAgB;QACzC,sCAAsC;QACtC,IAAI,CAAC,QAAQ,CAAC,UAAU,CAAC,SAAS,CAAC,IAAI,CAAC,QAAQ,CAAC,UAAU,CAAC,UAAU,CAAC,EAAE,CAAC;YACxE,QAAQ,GAAG,WAAW,QAAQ,EAAE,CAAC;QACnC,CAAC;QAED,wBAAwB;QACxB,OAAO,QAAQ,CAAC,OAAO,CAAC,KAAK,EAAE,EAAE,CAAC,CAAC;IACrC,CAAC;CACF","sourcesContent":["// Copyright (c) Microsoft Corporation.\n// Licensed under the MIT License.\n\nimport type { KeyCredential, TokenCredential } from \"@azure/core-auth\";\nimport type { RequestSession } from \"./models/index.js\";\nimport type {\n  CreateSessionOptions,\n  StartSessionOptions,\n  VoiceLiveSessionOptions,\n} from \"./voiceLiveSession.js\";\nimport { VoiceLiveSession } from \"./voiceLiveSession.js\";\nimport { logger } from \"./logger.js\";\n\nexport interface VoiceLiveClientOptions {\n  /** API version to use for the Voice Live service */\n  apiVersion?: string;\n  /** Default session options to apply to all created sessions */\n  defaultSessionOptions?: VoiceLiveSessionOptions;\n}\n\n/**\n * The VoiceLive client provides session management for real-time conversational AI capabilities.\n *\n * This client acts as a factory for creating VoiceLiveSession instances, which handle the actual\n * WebSocket connections and real-time interactions with the service.\n */\nexport class VoiceLiveClient {\n  private readonly _endpoint: string;\n  private readonly _credential: TokenCredential | KeyCredential;\n  private readonly _options: Required<VoiceLiveClientOptions>;\n\n  /**\n   * Creates an instance of VoiceLiveClient with endpoint and credential.\n   *\n   * @param endpoint - The Voice Live service endpoint URL\n   * @param credential - Azure TokenCredential or KeyCredential for authentication\n   * @param options - Optional configuration for the client\n   */\n  constructor(\n    endpoint: string,\n    credential: TokenCredential | KeyCredential,\n    options: VoiceLiveClientOptions = {},\n  ) {\n    this._endpoint = this._normalizeEndpoint(endpoint);\n    this._credential = credential;\n    this._options = this._buildDefaultOptions(options);\n\n    logger.info(\"VoiceLiveClient created\", {\n      endpoint: this._endpoint,\n      apiVersion: this._options.apiVersion,\n    });\n  }\n\n  /**\n   * Creates a new VoiceLiveSession for real-time voice communication.\n   *\n   * @param model - The model name to use for the session (e.g., \"gpt-4o-realtime-preview\")\n   * @param sessionOptions - Optional configuration specific to this session\n   * @returns A new VoiceLiveSession instance ready to connect\n   */\n  createSession(model: string, sessionOptions?: CreateSessionOptions): VoiceLiveSession;\n\n  /**\n   * Creates a new VoiceLiveSession for real-time voice communication with session configuration.\n   *\n   * @param sessionConfig - Session configuration including model and other settings\n   * @returns A new VoiceLiveSession instance ready to connect\n   */\n  createSession(\n    sessionConfig: RequestSession,\n    sessionOptions?: CreateSessionOptions,\n  ): VoiceLiveSession;\n\n  createSession(\n    modelOrConfig: string | RequestSession,\n    sessionOptions?: CreateSessionOptions,\n  ): VoiceLiveSession {\n    // Extract model name from the parameter\n    const model = typeof modelOrConfig === \"string\" ? modelOrConfig : modelOrConfig.model;\n    if (!model) {\n      throw new Error(\n        \"Model name is required. Provide either a model string or RequestSession with model property.\",\n      );\n    }\n\n    // Merge default session options with provided options\n    const mergedOptions: VoiceLiveSessionOptions = {\n      ...this._options.defaultSessionOptions,\n      ...sessionOptions,\n    };\n\n    const session = new VoiceLiveSession(\n      this._endpoint,\n      this._credential,\n      this._options.apiVersion,\n      model,\n      mergedOptions,\n    );\n\n    // If full session config was provided, store it for later use\n    if (typeof modelOrConfig !== \"string\") {\n      (session as any)._initialSessionConfig = modelOrConfig;\n    }\n\n    logger.info(\"VoiceLiveSession created\", { model });\n    return session;\n  }\n\n  /**\n   * Creates and immediately connects a new VoiceLiveSession.\n   *\n   * @param model - The model name to use for the session (e.g., \"gpt-4o-realtime-preview\")\n   * @param sessionOptions - Optional configuration specific to this session\n   * @returns A connected VoiceLiveSession instance\n   */\n  async startSession(\n    model: string,\n    sessionOptions?: StartSessionOptions,\n  ): Promise<VoiceLiveSession>;\n\n  /**\n   * Creates and immediately connects a new VoiceLiveSession with session configuration.\n   *\n   * @param sessionConfig - Session configuration including model and other settings\n   * @param sessionOptions - Optional configuration specific to this session\n   * @returns A connected VoiceLiveSession instance\n   */\n  async startSession(\n    sessionConfig: RequestSession,\n    sessionOptions?: StartSessionOptions,\n  ): Promise<VoiceLiveSession>;\n\n  async startSession(\n    modelOrConfig: string | RequestSession,\n    sessionOptions?: StartSessionOptions,\n  ): Promise<VoiceLiveSession> {\n    const session = this.createSession(modelOrConfig as any, sessionOptions);\n    await session.connect();\n\n    // If full session config was provided, send it after connection\n    if (typeof modelOrConfig !== \"string\") {\n      await session.updateSession(modelOrConfig);\n    }\n\n    return session;\n  }\n\n  // Properties\n  get endpoint(): string {\n    return this._endpoint;\n  }\n\n  get apiVersion(): string {\n    return this._options.apiVersion;\n  }\n\n  private _buildDefaultOptions(options: VoiceLiveClientOptions): Required<VoiceLiveClientOptions> {\n    return {\n      apiVersion: options.apiVersion || \"2025-10-01\",\n      defaultSessionOptions: options.defaultSessionOptions || {},\n    };\n  }\n\n  private _normalizeEndpoint(endpoint: string): string {\n    // Ensure endpoint has proper protocol\n    if (!endpoint.startsWith(\"http://\") && !endpoint.startsWith(\"https://\")) {\n      endpoint = `https://${endpoint}`;\n    }\n\n    // Remove trailing slash\n    return endpoint.replace(/\\/$/, \"\");\n  }\n}\n"]}

package/dist/esm/voiceLiveSession.d.ts

@@ -0,0 +1,138 @@
+import type { KeyCredential, TokenCredential } from "@azure/core-auth";
+import type { AbortSignalLike } from "@azure/abort-controller";
+import type { RequestSession, ClientEventUnion, ConversationRequestItem } from "./models/index.js";
+import { ConnectionState } from "./websocket/connectionManager.js";
+import type { VoiceLiveSessionHandlers, VoiceLiveSubscription } from "./handlers/sessionHandlers.js";
+export interface VoiceLiveSessionOptions {
+    /** Connection timeout in milliseconds */
+    connectionTimeoutInMs?: number;
+    /** Enable debug logging for development */
+    enableDebugLogging?: boolean;
+}
+export interface CreateSessionOptions extends VoiceLiveSessionOptions {
+}
+export interface StartSessionOptions extends VoiceLiveSessionOptions {
+}
+export interface ConnectOptions {
+    /** Abort signal to cancel connection attempt */
+    abortSignal?: AbortSignalLike;
+    /** Override connection timeout for this operation */
+    timeoutInMs?: number;
+}
+export interface SendEventOptions {
+    /** Abort signal to cancel send operation */
+    abortSignal?: AbortSignalLike;
+    /** Timeout for send operation */
+    timeoutInMs?: number;
+}
+export interface AudioStreamOptions extends SendEventOptions {
+    /** Turn ID for turn-based audio (if not provided, uses buffer mode) */
+    turnId?: string;
+}
+export interface TurnOptions extends SendEventOptions {
+    /** Custom turn ID (if not provided, one will be generated) */
+    turnId?: string;
+}
+/**
+ * Represents a WebSocket-based session for real-time voice communication with the Azure VoiceLive service.
+ *
+ * This class manages the connection, handles real-time communication, and provides access to all
+ * interactive features including audio streaming, conversation management, and avatar control.
+ */
+export declare class VoiceLiveSession {
+    private readonly _endpoint;
+    private readonly _credentialHandler;
+    private readonly _options;
+    private readonly _messageParser;
+    private readonly _model;
+    private _connectionManager?;
+    private _sessionId?;
+    private _activeTurnId?;
+    private _disposed;
+    private readonly _subscriptionManager;
+    /**
+     * Creates an instance of VoiceLiveSession.
+     *
+     * @param endpoint - The Voice Live service endpoint URL
+     * @param credential - Azure TokenCredential or KeyCredential for authentication
+     * @param apiVersion - API version to use for the Voice Live service
+     * @param model - The model name to use for the session
+     * @param options - Optional configuration for the session
+     */
+    constructor(endpoint: string, credential: TokenCredential | KeyCredential, apiVersion: string, model: string, options?: VoiceLiveSessionOptions);
+    /**
+     * Establishes connection to the Voice Live service with authentication.
+     */
+    connect(options?: ConnectOptions): Promise<void>;
+    /**
+     * Disconnects from the Voice Live service and cleans up resources.
+     */
+    disconnect(): Promise<void>;
+    /**
+     * Subscribe to VoiceLive session events using strongly-typed handlers.
+     * This follows the Azure SDK pattern used by EventHub, Service Bus, etc.
+     *
+     * @param handlers - Handler functions for different types of events
+     * @returns A subscription object that can be used to stop receiving events
+     */
+    subscribe(handlers: VoiceLiveSessionHandlers): VoiceLiveSubscription;
+    /**
+     * Sends a custom client event to the service.
+     */
+    sendEvent(event: ClientEventUnion, options?: SendEventOptions): Promise<void>;
+    /**
+     * Updates the session configuration with the service.
+     */
+    updateSession(session: RequestSession, options?: SendEventOptions): Promise<void>;
+    /**
+     * Sends audio data to the service using turn-based or buffer-based approach.
+     */
+    sendAudio(audioData: ArrayBuffer | Uint8Array, options?: AudioStreamOptions): Promise<void>;
+    /**
+     * Starts a new audio turn for turn-based audio input.
+     */
+    startAudioTurn(options?: TurnOptions): Promise<string>;
+    /**
+     * Ends the current audio turn.
+     */
+    endAudioTurn(turnId?: string, options?: SendEventOptions): Promise<void>;
+    /**
+     * Adds a conversation item (message) to the conversation.
+     */
+    addConversationItem(item: ConversationRequestItem, options?: SendEventOptions): Promise<void>;
+    /**
+     * Indicates whether the session is currently connected to the Voice Live service.
+     */
+    get isConnected(): boolean;
+    /**
+     * Gets the current connection state of the session.
+     */
+    get connectionState(): ConnectionState;
+    /**
+     * Gets the current session ID.
+     */
+    get sessionId(): string | undefined;
+    /**
+     * Gets the current active audio turn ID.
+     */
+    get activeTurnId(): string | undefined;
+    /**
+     * Disposes the session and cleans up resources.
+     */
+    dispose(): Promise<void>;
+    private _buildDefaultOptions;
+    private _normalizeEndpoint;
+    private _setupConnectionEventHandlers;
+    private _markSessionAsDead;
+    private _handleIncomingMessage;
+    private _handleServerEvent;
+    private _notifyConnectionEvent;
+    private _notifyServerEvent;
+    private _sendEvent;
+    private _ensureConnected;
+    private _ensureNotDisposed;
+    private _generateEventId;
+    private _generateTurnId;
+    private _arrayBufferToBase64;
+}
+//# sourceMappingURL=voiceLiveSession.d.ts.map